Get 20M+ Full-Text Papers For Less Than $1.50/day. Start a 14-Day Trial for You or Your Team.

Learn More →

A GPS-based real-time avalanche path warning and navigation system

A GPS-based real-time avalanche path warning and navigation system Geomatics, Natural Hazards and Risk, 2014 Vol. 5, No. 1, 56–80, http://dx.doi.org/10.1080/19475705.2012.762429 SANJAY KUMAR DEWALI*, JAGDISH CHANDRA JOSHI, ASHWAGOSHA GANJU and SNEHMANI Research and Development Centre, Snow and Avalanche Study Establishment, Chandigarh 160036, India (Received 14 June 2012; final version received 21 December 2012) Frequent avalanche activity and poor route visibility due to bad weather make snow bound mountainous regions quite unsafe for travel during the winter months. Hence, there is a need for an accurate navigation device that can help in the safe movement of mountain travellers in avalanche prone areas. This paper presents the design and implementation of a portable GPS-based warning and navigation system, capable of providing safe navigation to mountain travellers in avalanche-prone regions. The system uses a hand-held GPS with a positional accuracy of 2–5 m. A customized application has been developed for the visuali- zation of maps, navigation, positioning, tracking and issuing avalanche path warning. On the basis of registered avalanche path data, this application updates the traveller at predefined time intervals on whether the current position of the traveller is inside or outside an avalanche path. When the traveller enters an avalanche prone area, a warning in the form of text and voice messages is generat- ed by the device. For the testing and accuracy analysis of this application, a part of the Manali-Dhundi highway was designated as the study area. In numerous track tests, the system has demonstrated a high level of accuracy and a repetition in locating registered avalanche sites in open slopes/bare lands but accuracy dete- riorated in narrow valley/forested areas. An analysis of the trial results shows that the system can help travellers in snowbound mountain regions as a warning and navigation tool, using accurately registered avalanche sites and an appropriate buffer size around these sites. 1. Introduction The movement of persons in populated high mountain snow-bound regions of the Indian Himalayas is frequently affected by avalanche hazards and missing route problems due to snow deposition and bad weather conditions. Avalanche occur- rences cause misfortune, damage and adverse effects on human beings and property. Recognizing a hazard and taking preventive measures goes a long way in minimizing the losses due to any such disaster. For safe and correct movement in such hilly ter- rain, the traveller needs accurate information about both missing routes and spatial and temporal patterns of avalanches, in order to avoid avalanche sites and selection of wrong routes during the move. A scientific understanding of avalanches, as well as a knowledge of the local patterns of avalanche activity (gained through experience), is crucial for avalanche forecasting (McCollister et al. 2002), hazard mapping and *Corresponding author. Email: sk.dewali@sase.drdo.in 2013 Taylor & Francis A GPS-based real-time avalanche path warning 57 safe navigation. For decades, researchers have been developing physical, statistical and empirical models describing interaction between the snow cover, atmosphere and terrestrial surface. Although these models have improved the understanding of avalanche phenomena, still predicting the avalanche and assessing hazard is very dif- ficult due to the complex interrelationship of various contributing factors (snowpack parameters, climatic variables and terrain parameters). Various avalanche hazard zonation schemes have been implemented in the past for mapping, monitoring and assessing the hazards of various snow bound regions. A number of studies related to avalanche hazard susceptibility zonation based on the terrain parameters using remote sensing and GIS have been conducted (Gleason 1994; Gruber 2001; Tracy 2001; Barbolini & Keylock 2002; Maggioni & Gruber 2003). Using these schemes, avalanche hazard maps and avalanche atlases have been prepared for different regions. These maps and atlases contain all the relevant details of tracks and avalanche prone sites, but this information cannot be used efficiently and effectively in standalone mode for real time applications, because most of the ground features and control points of the map are not visible and are buried under the snow. A GPS-based navigation device using these hazard maps and atlases will be a useful tool for such real time navigation applications. With the increasing avail- ability of low-cost GPS technology, various navigation devices and location-based services (LBS) are readily available for the users, but their use in the Himalayan mountainous terrain is restricted due to one, the unavailability of high-resolution ac- curately registered avalanche sites and mountain tracks network data, two, the un- availability of cell phone networks in these regions and three, these LBS require user input interactively while this kind of user interaction must be minimized to avoid dis- tracting travellers in avalanche-prone areas. This work presents the development of an application and its implementation in a portable GPS device for automatic navigation, positioning, tracking and informing the user about an avalanche site in its proximity. The system is capable of generating different text and voice messages and warnings to the traveller depending on the po- sition of the traveller and avalanche path location. Users can also capture the data of any new avalanche path by measuring its outlines using this GPS system and similar- ly, new track/route data can be captured. After validation, this data may be used in future for navigation purposes. This device provides a potentially valuable means of identification of avalanche hazard areas and safe routes over large mountainous regions and provides relevant information for prevention of path loss and avalanche hazard incidents through identification and avoidance of avalanche terrain. Further- more, this provides an efficient way to warn and educate people about avalanche hazard areas. It can be used by land use planners and decision-makers for sustainable growth in the snowbound regions of the Himalayas. 1.1 Global positioning system-based navigation systems Most of the popular navigation systems employ GPS to locate position. GPS is a worldwide, portable and easy positioning system. Due to the evolution of this sys- tem, it is now possible to obtain precise and continuous data in both the horizontal and vertical planes. It helps in all aerial, underwater and terrestrial modes of navigation. 58 S.K. Dewali et al. In the past, a lot of work has been done by researchers on developing GPS as a stand-alone precise Global Navigation System. Hasan et al. (2009) reviewed all the significant developments and technical trends in the area of navigation systems. The improvement of the positioning accuracy of a GPS receiver is of prime impor- tance for accurate navigation. Various techniques/models have been attempted inde- pendently or in combination to achieve better accuracy, some of these use the combination of GPS software, Kalman filter and modified Kalman filter (Sato et al. 2000; Ladetto et al. 2001), modification in the measurement model associated with the GPS signal processing and statistical modelling (Krali et al. 2000; Mladen et al. 2006). Xuchu et al. (2000) developed a non-linear model using unscented Kalman filter for position and velocity estimation. The objective of this research is to address the GPS positioning problems in vehicle navigation under circumstances where visible satellites are frequently variable or less than four. This approach provides better estimates than those provided by other solutions. Ma et al. (2001) and Klukas et al. (2003) utilized an Urban Three-State Fade Model (UTSFM) to describe the GPS signal fading distribution according to satellite elevation angles, based on the analysis of signal power fading measurements gathered over different environments. Li and Xi (2007) discussed the method of combining FFT and circular convolution. This new method can be widely used in a highly dynamic situation with a high requirement of position precision, because it reduces the cost time of acquisition. Single frequency precise point positioning is already used in the GNSS communi- ty. The performance of the single frequency GPS receivers for precise point position- ing evaluated using different empirical ionospheric models and filters. The research by Øvstedal (2002) used the corrections from the International GNSS Service, which was newly available at the time. The standard deviations reached around 1 m. Le and Tiberius (2007) investigated single-frequency precise point positioning static as well as kinematic performance with the use of final products for all corrections that are available weeks after recording. van Bree and Tiberius (2012) presented the performance of real-time single-fre- quency precise point positioning demonstrated in terms of position accuracy. This precise point positioning technique relies on predicted satellite orbits, predicted glob- al ionospheric maps and in particular on real-time satellite clock estimates. Results are presented using only the measurements from a user receiver on the L1-frequency, for almost three months of data. As much as 95% errors are about 0.30 m in the hor- izontal direction and 0.65 m in the vertical direction. A significant improvement of position accuracy with real-time single-frequency precise point positioning can be gained when a good high-end receiver is used. 1.1.1 Pedestrian navigation systems. The most known navigation systems are vehi- cle-based navigation systems. However, fewer advances have been made in pedestri- an navigation systems (PNSs). There are some remarkable achievements, such as the PNS for visiting large delimitated areas (a museum or an institutional building) etc. But the most challenging task is to develop PNSs for guiding people in different areas, especially in hilly snow-covered mountainous areas. The main difference be- tween the vehicle-based navigation system and the PNS is the degree of freedom of motion as the pedestrian motion has a greater degree of freedom than the vehicular motion. Another difference is the availability of data for navigation. In vehicular A GPS-based real-time avalanche path warning 59 navigation systems the high-resolution road/street network data and base maps are available for navigation but this is not so in the case of the PNS. Pedestrian navigation systems have been reported in different research papers. Inoue et al. (2009) described a positioning system for indoor pedestrian navigation services using mobile phones. The system is made of smart phones and license-free radio beacon devices. In this system, the user’s device receives wireless beacon signals from the environment and can detect a user’s position independently from the mobile terminal. An exploratory study of a guiding system that uses geo-tagged photo col- lections from mobile phones for navigation is proposed by Beeharee and Steed (2006) and Hile et al. (2008). A user of the system sees a route description as text and a map that refers to a series of photographs. The experiment shows that presenting the right photographs helps particular types of routing instructions for users not fa- miliar with an area. Golledge et al. (1991) reported progress towards the development of a GIS data- base and analytical tools in a personal guidance system for blind travellers. Loomis et al. (1994) proposed the design for a navigation system for the visually impaired and described the progress made towards a portable, self-contained system that will allow visually impaired individuals to travel through familiar and unfamiliar envi- ronment without the assistance of guides. The proposed system has three compo- nents: a GPS receiver, a GIS with database and analysis tools along with a user interface. Golledge et al. (1998) describes the process of the building of a GIS for use in real time by blind travellers. In this work the various components of personal guid- ance system were identified and implemented and the limitations of GPS/GIS system were also discussed. Hile et al. (2008) and Bessho et al. (2008) describe the implementation and deploy- ment of a pedestrian navigation system which realizes a timely navigation by present- ing landmark-based instruction of guidance using high-level reasoning to influence the selection of landmarks along a navigation path and lower-level reasoning to se- lect appropriate images of those landmarks. Stark et al. (2007) describe a field study comparing four different navigational con- cepts i.e. auditory instructions plus digital, dynamic route (audio method), digital, dynamic route (route method), map with position and direction (direction method) and textual description by street names (description method) for pedestrians. All of these systems are based on recent state-of-the-art approaches and have been evaluat- ed by real users. The study ends with a recommendation for designing mobile pedes- trian navigation systems. Toth et al. (2007) present an artificial neural network and fuzzy logic-based theo- retical foundation and implementation algorithms, which integrate GPS, micro-elec- tro-mechanical inertial measurement unit (MEMS IMU), digital barometer, electronic compass and human pedometry to provide navigation and tracking of mil- itary and rescue ground personnel. Chen et al. (2009) propose an integrated GPS and multi-sensor pedestrian posi- tioning system to bridge the gaps of GPS signal outages. It includes an OEM GPS re- ceiver, a MEMS 3-axis accelerometer and a 2-axis digital compass. The positioning algorithm is a loosely coupled integration of the GPS and Pedestrian Dead Reckon- ing sensors via a Kalman filter. Martin et al. (2006) presented an approach to provide a software-based solution for a more correct position to a mobile pedestrian using a consumer-grade GPS receiver. A near-complete pocket PC implementation of a mobile multi-modal interaction 60 S.K. Dewali et al. (M3I) platform for pedestrian navigation is described in Wasinger et al. (2003).The platform easily supports indoor and outdoor navigation. 2. Proposed system description The schematic diagram of the proposed system solution for establishing an accurate and safe navigation for mountain travellers through avalanche-prone regions is shown in figure 1. The mountain traveller receives information regularly at a prede- fined interval, about the route status in the form of one of the following three condi- tions: (a) safe route (update) condition, (b) close to avalanche path condition (alert) and (c) inside the avalanche path condition (warning). The system essentially uses a hand-held GPS (GIS eXplorist Pro 10 GPS) unit (figure 3 (a)) for position tracking, application execution for analysis in real-time mode and displaying the results in the form of current position and route status. The technical details of the complete unit which includes the GPS receiver, processing and display unit are given in table 1. The workflow of the implementation of the system includes three main steps: one, data preparation; two, application development and three, implementing the applica- tion and data in a GPS device. 2.1 Study area and data preparation The study area is a part of the national highway between the locations of Manali and Dhundi in Himachal Pradesh (India). The overview map of the region and Cartosat-1 imagery along with the registered avalanche sites are shown in figure 2. The area falls in the Pir Panjal range of Indian western Himalayas and the altitude of the national highway in this region varies from 2000 m to 3000 m from the mean sea level. Most Figure 1. A schematic diagram of the navigation using the proposed system, the traveller is informed regularly about the route status in the form of one of these three conditions (a) safe route (update), (b) close to avalanche path (alert) and (c) inside the avalanche path (warning). A GPS-based real-time avalanche path warning 61 Table 1. Details of the GPS, processing and display unit. GPS 1. Receiver Integrated high-sensitivity, 20-channel GPS/SBAS receiver– SiRFstar III 2. Antenna Multidirectional GPS patch antenna 3. Accuracy 3 to 5 meter real-time accuracy with SBAS Sub-meter real-time accuracy using supported Bluetooth GPS w/SBAS, DGPS/Beacon receiver or OmniSTAR receiver 4. Data transfer protocol NMEA 5. Update rate 1 Hz Processing & Display Unit 1. OS Windows Mobile 2. Processor Samsung 533 MHz processor 3. RAM 128 MB 4. Display 16-bit WQVGA display (400  240) 5. Audio device Integrated speaker/microphone Physical 00 00 00 1. Size 2.57  5.04  1.45 2. Weight 195 grams Environmental 1. Operating temperature 10 Cto þ60 C 2. Humidity Waterproof IPX7 of the slopes in the Pir Panjal range are forested and heavy snowfall and mild ambi- ent temperatures are characteristics of this region. The road length taken for the trial and validation is approximately 19 km. There are 10 major avalanche sites affecting the traffic on this axis between Manali and Dhundi. These avalanche sites are trig- gered off during winter due to changes in snow conditions and cause hazard along the highway. A photograph of the field test and navigation in the study area using the proposed system is shown in figure 3 (b). Ortho-rectified Cartosat-1 satellite pan- chromatic image is used as the base map for navigation. Figure 2. Overview of the map region along with Cartosat-1 orthorectified imagery of the study area, road stretch and registered avalanche sites used for trial. 62 S.K. Dewali et al. Figure 3. (a) Hand-held GPS receiver used for position estimation, running the application and displaying results. (b) Photograph of the field trial and navigation using proposed system. 2.1.1 Generation of ortho image. The correct orthorectification of Cartosat data is the first requirement of the application. Cloud-free panchromatic stereo pair data of the study area, acquired on 29 Sept 2006 from Cartosat-1 (IRS-P5), are used for this purpose. The spatial resolution of data is 2.5 m in the horizontal plane and the swathe about 27 km. The ortho image generation requires ground control points (GCPs) and high-resolution DEM of the study. Hence, the methodology adopted to produce the Cartosat DEM involved stereo-strip triangulation of stereo pairs using high precise ground control points and automatic dense conjugate pair generation using a matching approach. The generated DEM is further evaluated for quality and editing to remove anomalies. The evaluation of DEM is done in two modes: first, in point mode (Kay et al. 2003; Nadeem et al. 2007), the accuracy is tested at GCP locations and RMSE is calculated and second, in surface mode, DEM is compared with the reference DEM generated from toposheets (1:25000 Scale). The Cartosat-1 stereo system is designed to provide stereo images. Two images of the same area have been taken from different angles. Stereo correlation has been applied to extract the matching point in two stereo images and a sensor geometric model is used to cal- culate elevations. Rational polynomial coefficients (RPC) are supplied with imagery product. These RPCs and GCP are used by Photogrammetric software to transform the ground-to-image geometric correction. Nine GCPs were used for geometric cor- rection and six were used for the evaluation of the orthorectified Cartosat-1 image. Figure 4 shows the flowchart of the various steps involved in ortho image generation. The ERDAS Imagine (LPS 9.3) software package is used for the generation of the or- tho image. Spatial resolution of the generated DEM is 10 m and this DEM dataset is resampled at 2.5 to generate the orthorectified image at 2.5 m spatial resolution. 2.1.2 Registered avalanche sites, buffer and road/track data. There are 10 major registered avalanche sites affecting the road track in the study area. All these sites are already mapped in the Snow and Avalanche Study Establishment (SASE) internal A GPS-based real-time avalanche path warning 63 Figure 4. Functional flow of ortho-image generation from Cartosat stereo pair data. report on avalanche hazard mitigation scheme 2010, using a hybrid approach. In this approach, data collected from various sources are used for manual delineation of the avalanche outline on a topographic map. Avalanche occurrence data of the past 22 years, manual measurement of each avalanche site using the GPS taken during the ground reconnaissance, observations from aerial reconnaissance and analyzed digital terrain data are taken as input for avalanche site delineation. Most of the ava- lanches in this region are frequent and are triggered annually. A vector polygon data layer of all mapped avalanche sites is generated with all the relevant attributes, in- cluding the physical dimensions of the avalanche sites and history of their past occur- rences (table 2). This attribute information will be displayed as avalanche site details during the navigation. Another vector layer using a buffer of 5 m around each Table 2. Avalanche sites of the study area with all attributes (SASE report on avalanche hazard mitigation scheme, 2010). Avalanche Length of Formation Road-affected Past occurrences site avalanche path (m) Zone Area (ha) Length (m) (number of times) Avl Site 1 1610 5.7 600 4 Avl Site 2 889 3.0 50 Data not available Avl Site 3 130 42.2 100 3 Avl Site 4 783 8.6 500 1 Avl Site 5 1530 19.4 350 3 Avl Site 6 659 2.8 80 Data not available Avl Site 7 3385 58.7 120 4 Avl Site 8 705 4.1 50 Data not available Avl Site 9 1350 10.9 50 2 Avl Site 10 730 16.0 120 40 64 S.K. Dewali et al. Table 3. Details of messages and information as per route conditions. Type of Sl. Route message (text/ Content No. status Criteria voice) of message 1. Safe Current position is outside Text Route is safe at of any avalanche site/ present buffer polygon 2. Close to Current position is inside Text þ voice Be careful you are avalanche any buffer polygon approaching an path avalanche site 3. Inside the Current position is inside Text þ voice þ all Move cautiously avalanche any avalanche site attributes of the you are inside an path polygon avalanche site avalanche path 4. GPS not active GPS data not received Text þ voice GPS is not active avalanche site is generated. This buffer layer is used to identify the traveller approaching an avalanche site. Finally, the road/track data layer is generated. 2.2 Application development An application is developed to (a) receive the GPS position coordinate at 1 Hz rate and calculate the average position at the user-defined interval, (b) query the current average position for the route status (safe route condition or close to avalanche path condition or inside the avalanche path condition) on the basis of avalanche polygon or buffer polygon data, (c) display the position on map (d) popup the text messages on display and play voice messages according to route condition. Different route conditions and corresponding messages and updates are shown in table 3. The appli- cation is developed using ArcPad and VB script. The functional flow for identifica- tion of the route status is shown in figure 5. Besides this, the application can capture new features during the navigation with all the required attributes in the GIS envi- ronment. These new features after validation can be added to the database for future navigation. 2.2.1 Implementing application and data in GPS device. After developing the appli- cation, it is tested for different route conditions using the laptop connected to a GPS unit and also by providing fake position (x, y) data. Once the test results are found satisfactory, the application and related data are finally deployed in a windows mo- bile-based GPS unit for real-time navigation testing and analysis. 3. Performance of the system The errors in this system are mainly divided into three categories; these are GPS errors in position measurement, errors in the GIS database generation and errors due to the update rate of the system and movement of the person. A GPS-based real-time avalanche path warning 65 Figure 5. A schematic of workflow for route condition identification. 3.1 GPS Error in position measurement A comprehensive overview of the GPS system and various processes involved in the estimation of position and factors affecting the accuracy and availability of position are given by Kaplan (1996) and Grewal et al. (2002). The accuracy of the position/time solution determined by GPS is expressed as the product of a geometry factor and a pseudo range error (PR) factor (Kaplan 1996; Aloi et al. 2007): s ¼ s  DOP ð1Þ position UERE where s is the standard deviation of the position accuracy, s is the stan- position UERE dard deviation of the user equivalent range error (UERE) and DOP is the position dilution of precision of the satellites used in the position solution. s , comprises individual standard deviations of all the error sources that de- UERE grade PR accuracy. The major error sources are: satellite clock errors, satellite ephemeris errors, ionospheric delay, tropospheric delay, receiver noise and 66 S.K. Dewali et al. multipath. Assuming these error sources to be independent of each other, s can UERE be defined as the root-sum-square (RSS) of these components (2), qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 2 2 2 2 2 s ¼ ðs þ s þ s þ s þ s þ s Þ ð2Þ UERE tropo rec clock ephemeris iono multipath noise where s is the standard deviation of the satellite clock error, s is the stan- clock ephemeris dard deviation of the satellite ephemeris error, s is the standard deviation of the iono ionospheric delay, s is the standard deviation of the tropospheric delay, s tropo rec noise is the standard deviation of the receiver thermal noise and s the standard de- multipath viation of the multipath error. The error budget (s ), estimated by GPS Joint UERE Program Offices for standard positioning service (SPS) under open sky conditions (NAVSTAR GPS User Equipment Introduction 1996) is 8.0 m. The five measures used in characterizing the accuracy of the position/time solution are (Kaplan 1996): (i) geometric dilution of precision (GDOP), (ii) position dilution of precision (PDOP), (iii) horizontal dilution of precision (HDOP), (iv) vertical dilu- tion of precision (VDOP) and (v) time dilution of precision (TDOP). These parame- ters are calculated as follows using equations (3) to (7): qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 2 2 2 2 GDOP ¼ ðs þ s þ s þ s Þ=s ; ð3Þ UERE x y z t qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 2 2 2 PDOP ¼ ðs þ s þ s Þ=s ; ð4Þ UERE x y z qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 2 2 HDOP ¼ ðs þ s Þ=s ; ð5Þ UERE x y qffiffiffiffiffi VDOP ¼ s =s ; ð6Þ UERE qffiffiffiffiffi TDOP ¼ s =s : ð7Þ UERE Here s ; s and s are standard deviations of the error in the x-axis component, x y z y-axis component and z-axis component of the calculated position, respectively, and s is the standard deviation of the receiver’s time bias error. The HDOP is the most relevant parameter in the current application since a two-dimensional (2-D) position of the traveller is desired. When GPS positions are logged at a fixed location over time, the positions are scattered over an area due to measurement errors. The dispersion of these points results in a scatter plot. The area within which the measurements are likely to be pres- ent is called the confidence region. The confidence region is then analyzed to quantify the GPS performance statistically. The confidence region with a radius describes the probability that the solution will be present within the specified accuracy (Novatel report 2003). Besides the HDOP and VDOP, the following three positional error measures are also used for evaluating the GPS accuracy. qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 2 2 Distance root mean squareðDRMSÞ¼ ðs þ s Þ; ð8Þ x y 2DRMS ¼ 2  DRMS; ð9Þ Circular error probability CEP ¼ð0:62s þ 0:56s Þ: ð10Þ x y A GPS-based real-time avalanche path warning 67 CEP refers to the radius of a circle in which 50% of the measurements lie. Similarly, DRMS and 2DRMS refer to the area within which 65% and 95% measurements are confined. 3.2 Error in GIS database generation Accuracy of the registration of GIS database is another important parameter on which the performance of the system depends. In this application, the 2-D accuracy (in the XY-plane) is desired, so the accuracy of the orthorectified image is evaluated at the known six GCP locations and the RMSE is calculated. The error in the XY-position of registered images is found within one pixel (2.5 M). This registered image is used only as a base map for navigation, so that the error in the image does not affect the ac- curacy of navigation directly, but it affects visualization. All the navigation-based queries are executed on the registered avalanche path data, so errors in avalanche path data will degrade the performance of the system. All the avalanche path polygons were created using the methodology given in ‘study area and data preparation section’. The geometric error in avalanche path (EAP) polygon data is below 1 m. 3.3 Error due to update rate (UPER) of the system and movement of the person In addition to the above-mentioned errors in the position, the update rate of the sys- tem at which it disseminates information to the user introduces an additional error. If the update interval of the system is Dt s and the speed of the traveller is v m/s, the maximum update error (MaxUPER) is given by equation (11): MaxUPER ¼ v  Dt ð11Þ The error due to MaxUPER is very high at higher speed of the traveler and slower up- date rates of system, as compared to the GPS position measurement error. This may in- crease the total error beyond the tolerance limit and make the system unusable. Limits of the total error (TE) of the system in a two-dimensional (2-D) position is given by equation (12): HDOP  EAP  TE  HDOP þ EAP þ MaxUPER: ð12Þ . Hence, the analysis and minimization of MaxUPER is essentially required for the better and usable performance of the system. In the present study, various Dt (1 s, 5 s, 10 s, 15 s, 20 s, 25 s and 30 s) are attempted with different tracking speeds (0.2 m/s to 12 m/s) for the analysis of the MaxUPER (figure 14). 4. Field experiments The experiments were performed in the Manali-Dhundi area in Himachal Pradesh, India (figure 3) from 15 Feb 2012 to 25 Feb 2012 during different times from morn- ing to evening. Experiments were conducted in two parts, in the first part of the ex- periment from 15 Feb 2012 to 18 Feb 2012, well-distributed ground control points were collected in the study area using a Leica SR 510 single frequency (12 L1 chan- nels) GPS receiver with AT501 antenna in the kinematic mode (full phase, C/A 68 S.K. Dewali et al. Figure 6. Road position initally and after geometric correction. Available in colour online. narrow code, position update rate 10 Hz, time for each measurement 15–20 min) and post-processing was done using SKI-Pro L1 software. The positional accuracy after post-processing was found to be 20–40 cm. These points were then used for orthorec- tification of the Cartosat-1 images, geometric correction of the road and avalanche sites of the study area. Figure 6 shows the initial and corrected road network. In the horizontal plane, a shift of 30 to 70 m was found in the data. This initial error in data was corrected by applying geometric correction using commercial GIS/image processing software. Using this corrected dataset the second part of the experiment was conducted. In this part, a total of 12 track tests at different locations of the study area were conducted from 19 Feb 2012 to 25 Feb 2012 (figure 7). Details of each trial are given in table 4. During the trials, most of the study area was snow covered, a snow depth of 2–3 m was observed in the navigation area and air temperatures were mostly found below 0 C. The working of the GPS system was found satisfactory under such environmen- tal conditions. 5. Trial results and analysis The experimental results demonstrate the performance of the proposed system. To understand the stability and consistency of the GPS measurement in the study re- gion, navigation trajectories of the complete track during the trials 10 and 11 are shown in figure 8. GPS positions are found consistent throughout the navigation and follow the reference road given in figure 7. Figures 9(a) and (b) show the results of the various trials (trial 6, 7, 8, 9, 11, 12) for the two different parts of the experimental track of approximate lengths of 1.5 km and 2.5 km, respectively. All the trials are following the same path within the range of 0.5–2.5 m along the straight/smooth parts of the test track, while along the curved and zigzag part of the test track the de- viation among different trials is observed within the range of 1.0 to 5.5 m. The speed of the traveller for different trials was (0.25 m/s to 1.4 m/s) and the update rate of system was 5 s. A GPS-based real-time avalanche path warning 69 Figure 7. Expermental plan of field trials. For the analysis of the GPS position accuracy, single point positions at a fixed location were collected in an open area for five hours using the proposed GPS receiver. Figure 10 shows the scatter plot of these single point positions. The CEP of the dataset is foundtobe1.48 m andDRMS and 2DRMSare 1.83 and3.66, respectively. This Table 4. Field experiment details. Mode of Number of Trial Time Approximate path movement/speed avalanche no. Date (hr) length covered (km) (km/hr) sites crossed 1. 19 Feb 12 0900–1300 5 (Manali and nearby) Onfoot (1–3 km/hr) 0 2. 20 Feb 12 1600–1800 5 (Manali and nearby) Onfoot (1–3 km/hr) 0 3. 25 Feb 12 1600–1830 5 (Manali and nearby) Onfoot (1–3 km/hr) 0 4. 19 Feb 12 1400–1800 7 (Palchan and nearby) Onfoot (2–3 km/hr) 1 5. 21 Feb 12 1600–1830 7 (Palchan and nearby) Onfoot (1–3 km/hr) 1 6. 22 Feb 12 0900–1300 7 (Palchan and nearby) Onfoot (1–3 km/hr) 1 7. 20 Feb 12 0830–1500 12 (Manali to Solang) Onfoot (2–4 km/hr) 3 8. 22 Feb 12 1400–1830 12 (Manali to Solang) Onfoot (2–4 km/hr) 3 9. 21 Feb 12 0830–1430 12 (Manali to Solang) Onfoot and vehicle 3 (2–10 km/hr) 10. 23 Feb 12 0830–1630 19 (Manali to Dhundi) Onfoot (2–5 km/hr) 10 11. 24 Feb 12 0900–1500 19 (Manali to Dhundi) Onfoot and vehicle 10 (2–10 km/hr) 12. 25 Feb 12 0900–1400 19 (Manali to Dhundi) Onfoot and vehicle 10 (2–30 km/hr) 70 S.K. Dewali et al. Figure 8. Trajectories followed during full study track navigated during trials 10 and 12. indicated that 95% of the position measurements are within 3.66 m from the fixed location. The three measures used in characterizing the accuracy of the position, i.e. (i) hori- zontal dilution of precision (HDOP), (ii) vertical dilution of precision (VDOP) and (iii) position dilution of precision (PDOP) were calculated for all the trials. The histo- grams of these error measures are shown in figures 11(a)and (b). The analysis of the histograms showed that for more than 90% of the position measurements, the values of HDOP and VDOP were found in the range of 1.0 m–1.5 m and 1.5 m to 3.5 m, re- spectively. This consistent lower value of HDOP indicates the satisfactory perfor- mance of the system in the study area. Variations of these measures with different track conditions were also studied (figure 12). An instantaneous value of these meas- ures was found highly fluctuating, so the simple moving averages (using 5 points be- fore and 5 points after the target point) of these measures were calculated to find out the spatial average value of these measures (figure 12). The average values of HDOP, VDOP and PDOP for the open areas were found between 1.0 m–1.5 m, 1.75 m– 2.75 m and 2.0 m–3.0 m, respectively. The values of these measures were higher for narrow (HDOP 1.0 m–2.25 m, VDOP 2.0 m–5.0 m, PDOP 3.0 m–5.5 m) and forest- ed (HDOP 1.5 m–3.25 m, VDOP 3.0 m–7.0 m, PDOP 3.5 m–7.5 m) regions. The accuracy of the GPS position depends on the number of satellites available for position estimation. A variation in different error measures and their averages along with the satellites’ availability number is shown in figure 13. Low and variable satel- lite availability was found in narrow and forested regions, leading to a high value of all error measures. For a variable satellite availability number < 6, the different error measure averages were high (HDOP 1.5 m–3.25 m, VDOP 3.25 m–7.5 m, PDOP 3.5 m–8.0 m) and for variable satellite availability number 6 the different error measure averages were relatively low (HDOP 1.25 m–1.75 m, VDOP 1.75 m–3.0 m, PDOP 2.0 m–4.0 m). The dependence of HDOP and VDOP on the number of satel- lites was analyzed. Figures 11(c) and (d) show the variation of these measures with A GPS-based real-time avalanche path warning 71 Figure 9. (a) and (b) Comparison of path followed by the traveller during different trials of the two different parts of the study area. the number of satellites available for position estimation. The mean value of both HDOP and VDOP was found decreasing with an increasing number of available sat- ellites. As the number of satellites increased from three to six, the HDOP and VDOP decreased significantly from 4 m to 1.5 m and 6 m to 3.25 m, respectively and after satellite availability number 6 the variation in these measures was not prominent. The box plot of HDOP and VDOP with satellite number also showed the same trend. In this figure on each box the central mark is median, edges of the box are the 25th and 75th percentiles and the whiskers extend to the most extreme data points. The range of the data was found to reduce significantly with increasing satellite numbers. With increasing satellite numbers the skewness of the data was found to be 72 S.K. Dewali et al. Figure 10. Scatter plot of the single point positions at a fixed location in an open area along with CEP (50% probability), DRMS (65% probability) and 2DRMS (95% probability). Figure 11. Statistical analysis of the error measures and their dependence on the number of satellites available for estimating position and speed of the traveller. (a) and (b) Histogram of the HDOP and VDOP of test data. (c) and (d) Variation of the mean and range of HDOP and VDOP with number of satellites available. (e) and (f) Variation of the mean and range of HDOP and VDOP with the speed of the traveller. A GPS-based real-time avalanche path warning 73 Figure 12. Three measures used in characterizing the accuracy of the position: (i) horizontal dilution of precision (HDOP), (ii) vertical dilution of precision (VDOP), (iii) position dilution of precision (PDOP) and their moving averages, along with the track conditions. Available in colour online. decreasing and the mean and median values were coinciding. So the measurements with six or more satellites were found mostly consistent and accurate. The effect of the speed of the traveller on the accuracy of the GPS position was an- alyzed (figure 14). A variation in different error measures and their averages along with different traveller speeds is shown here. Average error measures didn’t show sig- nificant variations in their values with the increasing speed of the traveller, but the fluctuations in the instantaneous values of the error measures increases with increas- ing speed. To understand the dependence of HDOP and VDOP on the speed of the traveller, the values of these measures were grouped into 12 equal speed intervals of 1 m/s each, starting from 0 m/s to 12 m/s (figures 11(e) and (f)). The mean of HDOP and VDOP is calculated for each group. The mean of these measures didn’t show Figure 13. Variation in HDOP, VDOP and PDOP and their moving averages with the number of satellites available for position estimation. Available in colour online. 74 S.K. Dewali et al. Figure 14. Variation in HDOP, VDOP and PDOP and their moving averages with the speed of the traveller. Available in colour online. any significant dependence on the speed of the traveller. Also, the box plot of these groups didn’t show any trend with speed. For the validation of the GPS measured position coordinates, a total of 15 ground control points, which were previously iden- tified all along the track, were used for comparison. Using the measured GPS coordi- nates of known control points and reference coordinates of these points, X-residual, Y-residual and RMSE were calculated for all the known points (figure 15). The mean RMSE of the observed GPS positions with respect to reference position was found 2.99 m with a minimum of 1.12 m and a maximum of 5.92 m. This small RMSE indicated that GPS measured positions of control points are close to the Figure 15. Comparison of the GPS-measured position coordinates and reference coordinates at different ground control points (a) X-Residual, (b) Y-Residual and (C) RMSE. A GPS-based real-time avalanche path warning 75 Figure 16. Variation of the update rate error (UPER) with different update rate (t) of the system and the speed of the traveller. reference coordinates and hence the GPS position can be used for navigation in the study region within this error range. Although the effect of the speed of the traveller on the accuracy of the GPS posi- tion was not found very significant (figure 14), but it significantly affected the update error (UPER) of the system as given in equation (11). In figure 16, a variation in UPER with varying speed and varying update rates (t) is demonstrated. Suppose the update rate ‘t’, is 1 s and if the speed is 0.5 m/s, 1 m/s and 10 m/s, then the UPER will be 0.5 m, 1 m and 10 m, respectively, because the system updates the informa- tion after 1 s only. In another case, suppose the speed is 2 m/s and if the update rates are 1 s, 5 s and 10 s, then the UPER will be 2 m, 10 m and 20 m, respectively. There- fore, to make the system usable for the present application, the speed and update rate should be adjusted in such a way, so that MaxUPER  2:0m. 5.1 Avalanche path navigation and warning After the analysis of the GPS measured position, trajectories of different trials and MaxUPER introduced by the system, performance of the system is analyzed for real-time identification avalanche sites and generating various text and audio mes- sages. Screen shots of system performance in predicting various route conditions (table 3) are given in figure 17(a), (b), (c) and (d). For this purpose, 10 avalanche site polygons and their buffers of the study area were taken. Most of the avalanche sites affecting the track have already been marked by the SASE and signboards have been placed at the point where the route crosses the avalanche paths (i.e. at both the ends of track length affected by each avalanche path). These points were used in the map for navigation and also as reference points in the field during trials. The field trials 10 and 11 were used for this analysis. Throughout the navigation, updates and the mes- sages generated by the system, were monitored carefully. Figure 17(a) is the case of safe route condition, and the prediction of this condition was done satisfactorily 76 S.K. Dewali et al. most of the time by the system. Whenever the system generated the warning for ‘inside an avalanche path’ condition (figure 17(c)), the position on ground was marked and the distance of this location from the point where the route crosses the corresponding avalanche path (i.e. location of nearest signboard) was measured dur- ing the navigation. Results of the navigation are given in table 5. The mean and max- imum distance between the predicted and actual starting point of various avalanche Figure 17. (a) ‘Safe route’ condition update generated by system. (b). ‘Approaching to ava- lanche site’ condition alert with details of avalanche site. (c). ‘Entering in an avalanche path’ condition warning with details of avalanche site. (d). ‘Out of the avalanche path’ condition up- date with details of the avalanche site. A GPS-based real-time avalanche path warning 77 Figure 17. Continued. sites ranges between 2 m to 3 m and 5 m to 7 m, respectively. This is the total error (TE) of the system and it is the sum of error in GPS position, error in registration of avalanche site (EAP) and error due to update rate (UPER) of the system. Although all these error components have been calculated in the present study, further analysis is required for the quantification of each error component. From this analysis, it is suggested that a buffer of 7 m around the avalanche polygon be used instead of a 5 m buffer, so that the ‘approaching to avalanche site’ condition alert (figure 17(b)) is generated by the system before entering the avalanche zone, even if there is an error of 7 m in predicting an avalanche site. 78 S.K. Dewali et al. Table 5. Mean and maximum error distances in predicting avalanche site polygons with different speeds and update rates. Distance from reference point at which warning generated (m) Total number of Avalanche Update Speed measurements in site rate (s) (m/s) Mean Maximum trials 10, 11 Avl Site 1 1, 5 1.0–0.3 No reference available 5 Avl Site 2 1, 5 1.0–0.3 No reference available 4 Avl Site 3 1, 5 1.0–0.3 2.1 6.2 6 Avl Site 4 1, 5 2.0–0.2 2.2 5.7 4 Avl Site 5 1, 5 4.0–0.3 2.4 5.6 4 Avl Site 6 1, 5 1.0–0.3 1.9 4.8 6 Avl Site 7 1, 5 3.0–0.1 3.2 10.6 4 Avl Site 8 1, 5 1.0–0.3 2.9 6.5 5 Avl Site 9 1, 5 4.0–0.2 3.0 6.2 6 Avl Site 10 1, 5 1.0–0.3 2.8 5.6 4 6. Conclusions and future scope This paper proposes the design and development of an avalanche warning and nav- igation system by implementing a customized user-friendly application in a hand- held GPS (eXplorist Pro 10). The performance of the GPS under various track conditions was evaluated. The reduction in positional accuracy was observed in narrow valley and forested areas (average value of HDOP 1.0 m–1.5 m for open areas and 2.0 m–3.5 m for narrow and forested regions). The dependence of HDOP and VDOP on the number of satellites available for position estimation was monitored. With increasing satellite numbers, the value of these measures was found to be decreasing and the measurements with the six or more satellites were found more consistent and accurate. No significant effect of the speed of traveller on the GPS positional accuracy was found. Also the effect of the UPER in the overall accuracy of the system was analyzed. The update rate and speed of the trav- eller was found to be very crucial in detecting the avalanche polygons. During the numerous field trials, the system was able to locate avalanche polygons with a mean error of 2 to 3 m and a maximum error of 5 to 7 m. Therefore, by taking the appropriate buffer size, this system can guide the travellers in snowbound mountain regions as a warning and navigation tool. In future, the system can be developed in the following ways: one, the proposed system works in a standalone mode, in case of any emergency there is no way to report avalanche accident or call for help if a traveller falls victim. So, in future, the system can be integrated with a radio com- munication device to establish a radio link for short-range communication between the proposed system and a central station. Two, as the use of GNSS-based applica- tions in mobile phones is gaining popularity due to fast development of the mobile phone technology, so the implementation of this application on the mobile phone platform will be more useful. The current application can be deployed on any win- dows-based mobile phone. In future, the application can be developed for other mobile phone operating systems. A GPS-based real-time avalanche path warning 79 Acknowledgements The authors are grateful to Dr M.R. Bhutiyani for providing continuous support and encouragement during the present study. The authors would like to thank all the SASE personnel who have helped during this study. References ALOI DN, ALSLIETY M, AKOS DM. 2007. Methodology for the evaluation of a GPS receiver performance in telematics applications. IEEE Trans Instrum Meas. 56:11–24. BARBOLINI M, KEYLOCK CJ. 2002. A new method for avalanche hazard mapping using a combi- nation of statistical and deterministic models. Nat Hazards Earth Syst Sci. 2:239–245. BEEHAREE AK, STEED A. 2006. A natural wayfinding exploiting photos in pedestrian naviga- tion systems. ACM International Conference Proceedings Series, Proceedings of the 8th Conference on Human-computer Interaction with Mobile Devices and Services; Helsinki, Finland, 159:81–88. BESSHO M, KOBAYASHI S, KOSHIZUKA N, SAKAMURA K. 2008. Implementation and deployment of a place-based pedestrian navigation system. Proceedings of the 32nd Annual IEEE International Computer Software and Applications Conference (COMPSAC 2008); Turku, Finland. CHEN W, FU Z, CHEN R, CHEN Y, ANDREI O, TUOMO,WANG J. 2009. An integrated GPS and multi-sensor pedestrian positioning system for 3D urban navigation. Proceedings of IEEE (Urban Remote Sensing Joint Event). DEPARTMENT OF DEFENSE. 1996. NAVSTAR GPS user equipment introduction (public release version) GPS joint program office, Space and Missiles System Centre, Los Angeles, CA. GLEASON JA. 1994. Terrain parameters of avalanche starting zones and their effects on ava- lanche frequency. International Snow Science Workshop (ISSW): Snowbird, Utah, USA, p. 393–404. GOLLEDGE RG, LOOMIS JM, KLATZKY RL, FLURY A, YANG XL. 1991. Designing a personal guidance system to aid navigation without sight: Progress on the GIS component. Int J Geogr Inf Syst. 5:373–396. GOLLEDGE RG, KLATZKY RL, LOOMIS JM, SPEIGLE J, TIETZ J. 1998. A geographic information system for a GPS-based personal guidance system. Int J Geogr Inf Sci. 12:727–749. GREWAL MS, WEILL LR, ANDEREWS AP. 2002. Global positioning systems, inertial navigation and integration. New York: Wiley. GRUBER U. 2001. Using GIS for avalanche hazard mapping in Switzerland. Proceedings of the 2001 ESRI International User Conference, San Diego. HASAN AM, SAMSUDIN K, RAMLI AR, AZMIR RS, ISMAEEL SA. 2009. A review of navigation systems (integration and algorithms). Aust J Basic Appl Sci. 3:943–959. HILE H, VEDANTHAM R, LIU A, GELFAND N, CUELLAR G, GRZESZCZUK R, BORRIELLO G. 2008. Landmark-based pedestrian navigation from collections of geotagged photos. Pro- ceedings of ACM International Conference on Mobile and Ubiquitous Multimedia (MUM 2008). INOUE Y, SASHIMA A, KURUMATANI K. 2009. Indoor positioning system using beacon devices for practical pedestrian navigation on mobile phone. Lecture Notes In Computer Sci- ence, Vol. 5585, Proceedings of the 6th International Conference on Ubiquitous Intel- ligence and Computing; Brisbane, Australia, p. 251–265. KAPLAN ED. 1996. Understanding GPS principles and applications. Norwood, MA: Artech House. KAY S, SPRUYT P, ALEXANDROU K. 2003. Geometric quality assessment of orthorectified VHR space image data. Photogramm Eng Remote Sens. 484–491. KLUKAS R, LACHAPELLE G, MA C, JEE G. 2003. GPS signal fading model for urban centers. IEE Proc, Microw Antennas Propag. 150:245–252. KRALI D, HSU T, CARIN L. 2000. On GPS positioning and integrity monitoring. IEEE Trans Aerosp Electron Syst. 36:327–336. 80 S.K. Dewali et al. LADETTO Q, GABAGLIO V, MERMINOD B. 2001. Two different approaches for augmented GPS pedestrian navigation. International Symposium on Location Based Services for Cellu- lar Users; Locellus. LE AQ, TIBERIUS CCJM. 2007. Single-frequency precise point positioning with optimal filter- ing. GPS Solut. 11:61–69. LI Y, XI X. 2007. Combining FFT and circular convolution method for high dynamic GPS sig- nal acquisition. Eighth International Conference on Electronic Measurement and Instruments; ICEMI, 2, p. 159–162. LOOMIS JM, GOLLEDGE RG, KLATZKY RL, SPEIGLE JM, TIETZ J. 1994. Personal guidance sys- tem for the visually impaired. Proceedings of the First Annual International ACM/ SIGCAPH Conference on Assistive Technologies (Assets ‘94); Marina Del Rey, Cali- fornia, p. 85–91. MA C, JEE G, MACGOUGAN G, LACHAPELLE G, BLOEBAUM S. 2001. GPS signal degradation modeling. ION GPS 2001, 11–14 September, Salt Lake City, UT, p. 882–893. MAGGIONI M, GRUBER U. 2003. The influence of topographic parameters on avalanche release dimension and frequency. Cold Reg Sci Technol. 37:407–419. MARTIN JD, KROSCHE J, BOLL S. 2006. Dynamic GPS-position correction for mobile pedestrian navigation and orientation. Proceedings of the 3rd Workshop on Positioning, Naviga- tion and Communication (WPNC’06), p. 199–208. MCCOLLISTER C, BIRKELAND K, HANSEN K, ASPINALL R, COMEY R. 2002. A probabilistic tech- nique for exploring multi-scale spatial patterns in historical avalanche data by combin- ing GIS and meteorological nearest neighbors with an example from the Jackson Hole Ski Area Wyoming. International Snow Science Workshop (ISSW) Proceedings. MLADEN M, ZORAN S, STEVAN B. 2006. A comparison of accuracy using a GPS and low-cost DGPS. IEEE Trans Instrum Meas. 55:1677–1683. NADEEM A, ANJUM M, RITESH A, JAYAPRASAD P, PATHAN SK, AJAI,SINGH,DK, SINGH AK. 2007. Extraction and Validation of Cartosat-1 DEM. J Indian Soc Remote Sens. 35:121–127. NOVATEL REPORT. 2003. GPS position accuracy measures, APN-029 Rev 1. Available from: http://www.novatel.com. ØVSTEDAL O. 2002. Absolute positioning with single frequency GPS receivers. GPS Solut. 5:33–44. SNOW AND AVALANCHE STUDY ESTABLISHMENT (SASE). 2010. Internal report on avalanche haz- ard mitigation scheme, 16. SATO G, ASAI T, SAKAMOTO T, HASE T. 2000. Improvement of the positioning of a software- based GPS receiver using a 32-bit embedded microprocessor. IEEE Trans Consum Electron. 46:521–530. STARK A, RIEBECK M, KAWALEK J. 2007. How to Design an Advanced Pedestrian Navigation System: Field Trial Results. Proceedings of IEEE International Workshop on Intelli- gent Data Acquisition and Advanced Computing Systems: Technology and Applica- tions; Dortmund, Germany, p. 690–694. TOTH C, GREJNER-BRZEZINSKA DA, MOAFIPOOR S. 2007. Pedestrian tracking and navigation using neural networks and fuzzy logic. Proceedings of IEEE International Symposium on Intelligent Signal Processing. TRACY L. 2001. Using GIS in avalanche hazard management. Proceedings of the 2001 ESRI International User Conference; San Diego. VAN BREE RJP, TIBERIUS CCJM. 2012. Real-time single-frequency precise point positioning: Accuracy Assessment. GPS Solut. 16:259–266. WASINGER R, STAHL C, KRUGER A. 2003. M3I in a pedestrian navigation & exploration system. Mobile HCI, Springer, p. 481–485. XUCHU M, MASSAKI W, HIDEKI H. 2000. Nonlinear GPS models for position estimate using low-cost GPS receiver; p. 637–642. http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png "Geomatics, Natural Hazards and Risk" Taylor & Francis

A GPS-based real-time avalanche path warning and navigation system

Loading next page...
 
/lp/taylor-francis/a-gps-based-real-time-avalanche-path-warning-and-navigation-system-oxgVwI9KrF

References (38)

Publisher
Taylor & Francis
Copyright
Copyright Taylor & Francis Group, LLC
ISSN
1947-5713
eISSN
1947-5705
DOI
10.1080/19475705.2012.762429
Publisher site
See Article on Publisher Site

Abstract

Geomatics, Natural Hazards and Risk, 2014 Vol. 5, No. 1, 56–80, http://dx.doi.org/10.1080/19475705.2012.762429 SANJAY KUMAR DEWALI*, JAGDISH CHANDRA JOSHI, ASHWAGOSHA GANJU and SNEHMANI Research and Development Centre, Snow and Avalanche Study Establishment, Chandigarh 160036, India (Received 14 June 2012; final version received 21 December 2012) Frequent avalanche activity and poor route visibility due to bad weather make snow bound mountainous regions quite unsafe for travel during the winter months. Hence, there is a need for an accurate navigation device that can help in the safe movement of mountain travellers in avalanche prone areas. This paper presents the design and implementation of a portable GPS-based warning and navigation system, capable of providing safe navigation to mountain travellers in avalanche-prone regions. The system uses a hand-held GPS with a positional accuracy of 2–5 m. A customized application has been developed for the visuali- zation of maps, navigation, positioning, tracking and issuing avalanche path warning. On the basis of registered avalanche path data, this application updates the traveller at predefined time intervals on whether the current position of the traveller is inside or outside an avalanche path. When the traveller enters an avalanche prone area, a warning in the form of text and voice messages is generat- ed by the device. For the testing and accuracy analysis of this application, a part of the Manali-Dhundi highway was designated as the study area. In numerous track tests, the system has demonstrated a high level of accuracy and a repetition in locating registered avalanche sites in open slopes/bare lands but accuracy dete- riorated in narrow valley/forested areas. An analysis of the trial results shows that the system can help travellers in snowbound mountain regions as a warning and navigation tool, using accurately registered avalanche sites and an appropriate buffer size around these sites. 1. Introduction The movement of persons in populated high mountain snow-bound regions of the Indian Himalayas is frequently affected by avalanche hazards and missing route problems due to snow deposition and bad weather conditions. Avalanche occur- rences cause misfortune, damage and adverse effects on human beings and property. Recognizing a hazard and taking preventive measures goes a long way in minimizing the losses due to any such disaster. For safe and correct movement in such hilly ter- rain, the traveller needs accurate information about both missing routes and spatial and temporal patterns of avalanches, in order to avoid avalanche sites and selection of wrong routes during the move. A scientific understanding of avalanches, as well as a knowledge of the local patterns of avalanche activity (gained through experience), is crucial for avalanche forecasting (McCollister et al. 2002), hazard mapping and *Corresponding author. Email: sk.dewali@sase.drdo.in 2013 Taylor & Francis A GPS-based real-time avalanche path warning 57 safe navigation. For decades, researchers have been developing physical, statistical and empirical models describing interaction between the snow cover, atmosphere and terrestrial surface. Although these models have improved the understanding of avalanche phenomena, still predicting the avalanche and assessing hazard is very dif- ficult due to the complex interrelationship of various contributing factors (snowpack parameters, climatic variables and terrain parameters). Various avalanche hazard zonation schemes have been implemented in the past for mapping, monitoring and assessing the hazards of various snow bound regions. A number of studies related to avalanche hazard susceptibility zonation based on the terrain parameters using remote sensing and GIS have been conducted (Gleason 1994; Gruber 2001; Tracy 2001; Barbolini & Keylock 2002; Maggioni & Gruber 2003). Using these schemes, avalanche hazard maps and avalanche atlases have been prepared for different regions. These maps and atlases contain all the relevant details of tracks and avalanche prone sites, but this information cannot be used efficiently and effectively in standalone mode for real time applications, because most of the ground features and control points of the map are not visible and are buried under the snow. A GPS-based navigation device using these hazard maps and atlases will be a useful tool for such real time navigation applications. With the increasing avail- ability of low-cost GPS technology, various navigation devices and location-based services (LBS) are readily available for the users, but their use in the Himalayan mountainous terrain is restricted due to one, the unavailability of high-resolution ac- curately registered avalanche sites and mountain tracks network data, two, the un- availability of cell phone networks in these regions and three, these LBS require user input interactively while this kind of user interaction must be minimized to avoid dis- tracting travellers in avalanche-prone areas. This work presents the development of an application and its implementation in a portable GPS device for automatic navigation, positioning, tracking and informing the user about an avalanche site in its proximity. The system is capable of generating different text and voice messages and warnings to the traveller depending on the po- sition of the traveller and avalanche path location. Users can also capture the data of any new avalanche path by measuring its outlines using this GPS system and similar- ly, new track/route data can be captured. After validation, this data may be used in future for navigation purposes. This device provides a potentially valuable means of identification of avalanche hazard areas and safe routes over large mountainous regions and provides relevant information for prevention of path loss and avalanche hazard incidents through identification and avoidance of avalanche terrain. Further- more, this provides an efficient way to warn and educate people about avalanche hazard areas. It can be used by land use planners and decision-makers for sustainable growth in the snowbound regions of the Himalayas. 1.1 Global positioning system-based navigation systems Most of the popular navigation systems employ GPS to locate position. GPS is a worldwide, portable and easy positioning system. Due to the evolution of this sys- tem, it is now possible to obtain precise and continuous data in both the horizontal and vertical planes. It helps in all aerial, underwater and terrestrial modes of navigation. 58 S.K. Dewali et al. In the past, a lot of work has been done by researchers on developing GPS as a stand-alone precise Global Navigation System. Hasan et al. (2009) reviewed all the significant developments and technical trends in the area of navigation systems. The improvement of the positioning accuracy of a GPS receiver is of prime impor- tance for accurate navigation. Various techniques/models have been attempted inde- pendently or in combination to achieve better accuracy, some of these use the combination of GPS software, Kalman filter and modified Kalman filter (Sato et al. 2000; Ladetto et al. 2001), modification in the measurement model associated with the GPS signal processing and statistical modelling (Krali et al. 2000; Mladen et al. 2006). Xuchu et al. (2000) developed a non-linear model using unscented Kalman filter for position and velocity estimation. The objective of this research is to address the GPS positioning problems in vehicle navigation under circumstances where visible satellites are frequently variable or less than four. This approach provides better estimates than those provided by other solutions. Ma et al. (2001) and Klukas et al. (2003) utilized an Urban Three-State Fade Model (UTSFM) to describe the GPS signal fading distribution according to satellite elevation angles, based on the analysis of signal power fading measurements gathered over different environments. Li and Xi (2007) discussed the method of combining FFT and circular convolution. This new method can be widely used in a highly dynamic situation with a high requirement of position precision, because it reduces the cost time of acquisition. Single frequency precise point positioning is already used in the GNSS communi- ty. The performance of the single frequency GPS receivers for precise point position- ing evaluated using different empirical ionospheric models and filters. The research by Øvstedal (2002) used the corrections from the International GNSS Service, which was newly available at the time. The standard deviations reached around 1 m. Le and Tiberius (2007) investigated single-frequency precise point positioning static as well as kinematic performance with the use of final products for all corrections that are available weeks after recording. van Bree and Tiberius (2012) presented the performance of real-time single-fre- quency precise point positioning demonstrated in terms of position accuracy. This precise point positioning technique relies on predicted satellite orbits, predicted glob- al ionospheric maps and in particular on real-time satellite clock estimates. Results are presented using only the measurements from a user receiver on the L1-frequency, for almost three months of data. As much as 95% errors are about 0.30 m in the hor- izontal direction and 0.65 m in the vertical direction. A significant improvement of position accuracy with real-time single-frequency precise point positioning can be gained when a good high-end receiver is used. 1.1.1 Pedestrian navigation systems. The most known navigation systems are vehi- cle-based navigation systems. However, fewer advances have been made in pedestri- an navigation systems (PNSs). There are some remarkable achievements, such as the PNS for visiting large delimitated areas (a museum or an institutional building) etc. But the most challenging task is to develop PNSs for guiding people in different areas, especially in hilly snow-covered mountainous areas. The main difference be- tween the vehicle-based navigation system and the PNS is the degree of freedom of motion as the pedestrian motion has a greater degree of freedom than the vehicular motion. Another difference is the availability of data for navigation. In vehicular A GPS-based real-time avalanche path warning 59 navigation systems the high-resolution road/street network data and base maps are available for navigation but this is not so in the case of the PNS. Pedestrian navigation systems have been reported in different research papers. Inoue et al. (2009) described a positioning system for indoor pedestrian navigation services using mobile phones. The system is made of smart phones and license-free radio beacon devices. In this system, the user’s device receives wireless beacon signals from the environment and can detect a user’s position independently from the mobile terminal. An exploratory study of a guiding system that uses geo-tagged photo col- lections from mobile phones for navigation is proposed by Beeharee and Steed (2006) and Hile et al. (2008). A user of the system sees a route description as text and a map that refers to a series of photographs. The experiment shows that presenting the right photographs helps particular types of routing instructions for users not fa- miliar with an area. Golledge et al. (1991) reported progress towards the development of a GIS data- base and analytical tools in a personal guidance system for blind travellers. Loomis et al. (1994) proposed the design for a navigation system for the visually impaired and described the progress made towards a portable, self-contained system that will allow visually impaired individuals to travel through familiar and unfamiliar envi- ronment without the assistance of guides. The proposed system has three compo- nents: a GPS receiver, a GIS with database and analysis tools along with a user interface. Golledge et al. (1998) describes the process of the building of a GIS for use in real time by blind travellers. In this work the various components of personal guid- ance system were identified and implemented and the limitations of GPS/GIS system were also discussed. Hile et al. (2008) and Bessho et al. (2008) describe the implementation and deploy- ment of a pedestrian navigation system which realizes a timely navigation by present- ing landmark-based instruction of guidance using high-level reasoning to influence the selection of landmarks along a navigation path and lower-level reasoning to se- lect appropriate images of those landmarks. Stark et al. (2007) describe a field study comparing four different navigational con- cepts i.e. auditory instructions plus digital, dynamic route (audio method), digital, dynamic route (route method), map with position and direction (direction method) and textual description by street names (description method) for pedestrians. All of these systems are based on recent state-of-the-art approaches and have been evaluat- ed by real users. The study ends with a recommendation for designing mobile pedes- trian navigation systems. Toth et al. (2007) present an artificial neural network and fuzzy logic-based theo- retical foundation and implementation algorithms, which integrate GPS, micro-elec- tro-mechanical inertial measurement unit (MEMS IMU), digital barometer, electronic compass and human pedometry to provide navigation and tracking of mil- itary and rescue ground personnel. Chen et al. (2009) propose an integrated GPS and multi-sensor pedestrian posi- tioning system to bridge the gaps of GPS signal outages. It includes an OEM GPS re- ceiver, a MEMS 3-axis accelerometer and a 2-axis digital compass. The positioning algorithm is a loosely coupled integration of the GPS and Pedestrian Dead Reckon- ing sensors via a Kalman filter. Martin et al. (2006) presented an approach to provide a software-based solution for a more correct position to a mobile pedestrian using a consumer-grade GPS receiver. A near-complete pocket PC implementation of a mobile multi-modal interaction 60 S.K. Dewali et al. (M3I) platform for pedestrian navigation is described in Wasinger et al. (2003).The platform easily supports indoor and outdoor navigation. 2. Proposed system description The schematic diagram of the proposed system solution for establishing an accurate and safe navigation for mountain travellers through avalanche-prone regions is shown in figure 1. The mountain traveller receives information regularly at a prede- fined interval, about the route status in the form of one of the following three condi- tions: (a) safe route (update) condition, (b) close to avalanche path condition (alert) and (c) inside the avalanche path condition (warning). The system essentially uses a hand-held GPS (GIS eXplorist Pro 10 GPS) unit (figure 3 (a)) for position tracking, application execution for analysis in real-time mode and displaying the results in the form of current position and route status. The technical details of the complete unit which includes the GPS receiver, processing and display unit are given in table 1. The workflow of the implementation of the system includes three main steps: one, data preparation; two, application development and three, implementing the applica- tion and data in a GPS device. 2.1 Study area and data preparation The study area is a part of the national highway between the locations of Manali and Dhundi in Himachal Pradesh (India). The overview map of the region and Cartosat-1 imagery along with the registered avalanche sites are shown in figure 2. The area falls in the Pir Panjal range of Indian western Himalayas and the altitude of the national highway in this region varies from 2000 m to 3000 m from the mean sea level. Most Figure 1. A schematic diagram of the navigation using the proposed system, the traveller is informed regularly about the route status in the form of one of these three conditions (a) safe route (update), (b) close to avalanche path (alert) and (c) inside the avalanche path (warning). A GPS-based real-time avalanche path warning 61 Table 1. Details of the GPS, processing and display unit. GPS 1. Receiver Integrated high-sensitivity, 20-channel GPS/SBAS receiver– SiRFstar III 2. Antenna Multidirectional GPS patch antenna 3. Accuracy 3 to 5 meter real-time accuracy with SBAS Sub-meter real-time accuracy using supported Bluetooth GPS w/SBAS, DGPS/Beacon receiver or OmniSTAR receiver 4. Data transfer protocol NMEA 5. Update rate 1 Hz Processing & Display Unit 1. OS Windows Mobile 2. Processor Samsung 533 MHz processor 3. RAM 128 MB 4. Display 16-bit WQVGA display (400  240) 5. Audio device Integrated speaker/microphone Physical 00 00 00 1. Size 2.57  5.04  1.45 2. Weight 195 grams Environmental 1. Operating temperature 10 Cto þ60 C 2. Humidity Waterproof IPX7 of the slopes in the Pir Panjal range are forested and heavy snowfall and mild ambi- ent temperatures are characteristics of this region. The road length taken for the trial and validation is approximately 19 km. There are 10 major avalanche sites affecting the traffic on this axis between Manali and Dhundi. These avalanche sites are trig- gered off during winter due to changes in snow conditions and cause hazard along the highway. A photograph of the field test and navigation in the study area using the proposed system is shown in figure 3 (b). Ortho-rectified Cartosat-1 satellite pan- chromatic image is used as the base map for navigation. Figure 2. Overview of the map region along with Cartosat-1 orthorectified imagery of the study area, road stretch and registered avalanche sites used for trial. 62 S.K. Dewali et al. Figure 3. (a) Hand-held GPS receiver used for position estimation, running the application and displaying results. (b) Photograph of the field trial and navigation using proposed system. 2.1.1 Generation of ortho image. The correct orthorectification of Cartosat data is the first requirement of the application. Cloud-free panchromatic stereo pair data of the study area, acquired on 29 Sept 2006 from Cartosat-1 (IRS-P5), are used for this purpose. The spatial resolution of data is 2.5 m in the horizontal plane and the swathe about 27 km. The ortho image generation requires ground control points (GCPs) and high-resolution DEM of the study. Hence, the methodology adopted to produce the Cartosat DEM involved stereo-strip triangulation of stereo pairs using high precise ground control points and automatic dense conjugate pair generation using a matching approach. The generated DEM is further evaluated for quality and editing to remove anomalies. The evaluation of DEM is done in two modes: first, in point mode (Kay et al. 2003; Nadeem et al. 2007), the accuracy is tested at GCP locations and RMSE is calculated and second, in surface mode, DEM is compared with the reference DEM generated from toposheets (1:25000 Scale). The Cartosat-1 stereo system is designed to provide stereo images. Two images of the same area have been taken from different angles. Stereo correlation has been applied to extract the matching point in two stereo images and a sensor geometric model is used to cal- culate elevations. Rational polynomial coefficients (RPC) are supplied with imagery product. These RPCs and GCP are used by Photogrammetric software to transform the ground-to-image geometric correction. Nine GCPs were used for geometric cor- rection and six were used for the evaluation of the orthorectified Cartosat-1 image. Figure 4 shows the flowchart of the various steps involved in ortho image generation. The ERDAS Imagine (LPS 9.3) software package is used for the generation of the or- tho image. Spatial resolution of the generated DEM is 10 m and this DEM dataset is resampled at 2.5 to generate the orthorectified image at 2.5 m spatial resolution. 2.1.2 Registered avalanche sites, buffer and road/track data. There are 10 major registered avalanche sites affecting the road track in the study area. All these sites are already mapped in the Snow and Avalanche Study Establishment (SASE) internal A GPS-based real-time avalanche path warning 63 Figure 4. Functional flow of ortho-image generation from Cartosat stereo pair data. report on avalanche hazard mitigation scheme 2010, using a hybrid approach. In this approach, data collected from various sources are used for manual delineation of the avalanche outline on a topographic map. Avalanche occurrence data of the past 22 years, manual measurement of each avalanche site using the GPS taken during the ground reconnaissance, observations from aerial reconnaissance and analyzed digital terrain data are taken as input for avalanche site delineation. Most of the ava- lanches in this region are frequent and are triggered annually. A vector polygon data layer of all mapped avalanche sites is generated with all the relevant attributes, in- cluding the physical dimensions of the avalanche sites and history of their past occur- rences (table 2). This attribute information will be displayed as avalanche site details during the navigation. Another vector layer using a buffer of 5 m around each Table 2. Avalanche sites of the study area with all attributes (SASE report on avalanche hazard mitigation scheme, 2010). Avalanche Length of Formation Road-affected Past occurrences site avalanche path (m) Zone Area (ha) Length (m) (number of times) Avl Site 1 1610 5.7 600 4 Avl Site 2 889 3.0 50 Data not available Avl Site 3 130 42.2 100 3 Avl Site 4 783 8.6 500 1 Avl Site 5 1530 19.4 350 3 Avl Site 6 659 2.8 80 Data not available Avl Site 7 3385 58.7 120 4 Avl Site 8 705 4.1 50 Data not available Avl Site 9 1350 10.9 50 2 Avl Site 10 730 16.0 120 40 64 S.K. Dewali et al. Table 3. Details of messages and information as per route conditions. Type of Sl. Route message (text/ Content No. status Criteria voice) of message 1. Safe Current position is outside Text Route is safe at of any avalanche site/ present buffer polygon 2. Close to Current position is inside Text þ voice Be careful you are avalanche any buffer polygon approaching an path avalanche site 3. Inside the Current position is inside Text þ voice þ all Move cautiously avalanche any avalanche site attributes of the you are inside an path polygon avalanche site avalanche path 4. GPS not active GPS data not received Text þ voice GPS is not active avalanche site is generated. This buffer layer is used to identify the traveller approaching an avalanche site. Finally, the road/track data layer is generated. 2.2 Application development An application is developed to (a) receive the GPS position coordinate at 1 Hz rate and calculate the average position at the user-defined interval, (b) query the current average position for the route status (safe route condition or close to avalanche path condition or inside the avalanche path condition) on the basis of avalanche polygon or buffer polygon data, (c) display the position on map (d) popup the text messages on display and play voice messages according to route condition. Different route conditions and corresponding messages and updates are shown in table 3. The appli- cation is developed using ArcPad and VB script. The functional flow for identifica- tion of the route status is shown in figure 5. Besides this, the application can capture new features during the navigation with all the required attributes in the GIS envi- ronment. These new features after validation can be added to the database for future navigation. 2.2.1 Implementing application and data in GPS device. After developing the appli- cation, it is tested for different route conditions using the laptop connected to a GPS unit and also by providing fake position (x, y) data. Once the test results are found satisfactory, the application and related data are finally deployed in a windows mo- bile-based GPS unit for real-time navigation testing and analysis. 3. Performance of the system The errors in this system are mainly divided into three categories; these are GPS errors in position measurement, errors in the GIS database generation and errors due to the update rate of the system and movement of the person. A GPS-based real-time avalanche path warning 65 Figure 5. A schematic of workflow for route condition identification. 3.1 GPS Error in position measurement A comprehensive overview of the GPS system and various processes involved in the estimation of position and factors affecting the accuracy and availability of position are given by Kaplan (1996) and Grewal et al. (2002). The accuracy of the position/time solution determined by GPS is expressed as the product of a geometry factor and a pseudo range error (PR) factor (Kaplan 1996; Aloi et al. 2007): s ¼ s  DOP ð1Þ position UERE where s is the standard deviation of the position accuracy, s is the stan- position UERE dard deviation of the user equivalent range error (UERE) and DOP is the position dilution of precision of the satellites used in the position solution. s , comprises individual standard deviations of all the error sources that de- UERE grade PR accuracy. The major error sources are: satellite clock errors, satellite ephemeris errors, ionospheric delay, tropospheric delay, receiver noise and 66 S.K. Dewali et al. multipath. Assuming these error sources to be independent of each other, s can UERE be defined as the root-sum-square (RSS) of these components (2), qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 2 2 2 2 2 s ¼ ðs þ s þ s þ s þ s þ s Þ ð2Þ UERE tropo rec clock ephemeris iono multipath noise where s is the standard deviation of the satellite clock error, s is the stan- clock ephemeris dard deviation of the satellite ephemeris error, s is the standard deviation of the iono ionospheric delay, s is the standard deviation of the tropospheric delay, s tropo rec noise is the standard deviation of the receiver thermal noise and s the standard de- multipath viation of the multipath error. The error budget (s ), estimated by GPS Joint UERE Program Offices for standard positioning service (SPS) under open sky conditions (NAVSTAR GPS User Equipment Introduction 1996) is 8.0 m. The five measures used in characterizing the accuracy of the position/time solution are (Kaplan 1996): (i) geometric dilution of precision (GDOP), (ii) position dilution of precision (PDOP), (iii) horizontal dilution of precision (HDOP), (iv) vertical dilu- tion of precision (VDOP) and (v) time dilution of precision (TDOP). These parame- ters are calculated as follows using equations (3) to (7): qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 2 2 2 2 GDOP ¼ ðs þ s þ s þ s Þ=s ; ð3Þ UERE x y z t qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 2 2 2 PDOP ¼ ðs þ s þ s Þ=s ; ð4Þ UERE x y z qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 2 2 HDOP ¼ ðs þ s Þ=s ; ð5Þ UERE x y qffiffiffiffiffi VDOP ¼ s =s ; ð6Þ UERE qffiffiffiffiffi TDOP ¼ s =s : ð7Þ UERE Here s ; s and s are standard deviations of the error in the x-axis component, x y z y-axis component and z-axis component of the calculated position, respectively, and s is the standard deviation of the receiver’s time bias error. The HDOP is the most relevant parameter in the current application since a two-dimensional (2-D) position of the traveller is desired. When GPS positions are logged at a fixed location over time, the positions are scattered over an area due to measurement errors. The dispersion of these points results in a scatter plot. The area within which the measurements are likely to be pres- ent is called the confidence region. The confidence region is then analyzed to quantify the GPS performance statistically. The confidence region with a radius describes the probability that the solution will be present within the specified accuracy (Novatel report 2003). Besides the HDOP and VDOP, the following three positional error measures are also used for evaluating the GPS accuracy. qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 2 2 Distance root mean squareðDRMSÞ¼ ðs þ s Þ; ð8Þ x y 2DRMS ¼ 2  DRMS; ð9Þ Circular error probability CEP ¼ð0:62s þ 0:56s Þ: ð10Þ x y A GPS-based real-time avalanche path warning 67 CEP refers to the radius of a circle in which 50% of the measurements lie. Similarly, DRMS and 2DRMS refer to the area within which 65% and 95% measurements are confined. 3.2 Error in GIS database generation Accuracy of the registration of GIS database is another important parameter on which the performance of the system depends. In this application, the 2-D accuracy (in the XY-plane) is desired, so the accuracy of the orthorectified image is evaluated at the known six GCP locations and the RMSE is calculated. The error in the XY-position of registered images is found within one pixel (2.5 M). This registered image is used only as a base map for navigation, so that the error in the image does not affect the ac- curacy of navigation directly, but it affects visualization. All the navigation-based queries are executed on the registered avalanche path data, so errors in avalanche path data will degrade the performance of the system. All the avalanche path polygons were created using the methodology given in ‘study area and data preparation section’. The geometric error in avalanche path (EAP) polygon data is below 1 m. 3.3 Error due to update rate (UPER) of the system and movement of the person In addition to the above-mentioned errors in the position, the update rate of the sys- tem at which it disseminates information to the user introduces an additional error. If the update interval of the system is Dt s and the speed of the traveller is v m/s, the maximum update error (MaxUPER) is given by equation (11): MaxUPER ¼ v  Dt ð11Þ The error due to MaxUPER is very high at higher speed of the traveler and slower up- date rates of system, as compared to the GPS position measurement error. This may in- crease the total error beyond the tolerance limit and make the system unusable. Limits of the total error (TE) of the system in a two-dimensional (2-D) position is given by equation (12): HDOP  EAP  TE  HDOP þ EAP þ MaxUPER: ð12Þ . Hence, the analysis and minimization of MaxUPER is essentially required for the better and usable performance of the system. In the present study, various Dt (1 s, 5 s, 10 s, 15 s, 20 s, 25 s and 30 s) are attempted with different tracking speeds (0.2 m/s to 12 m/s) for the analysis of the MaxUPER (figure 14). 4. Field experiments The experiments were performed in the Manali-Dhundi area in Himachal Pradesh, India (figure 3) from 15 Feb 2012 to 25 Feb 2012 during different times from morn- ing to evening. Experiments were conducted in two parts, in the first part of the ex- periment from 15 Feb 2012 to 18 Feb 2012, well-distributed ground control points were collected in the study area using a Leica SR 510 single frequency (12 L1 chan- nels) GPS receiver with AT501 antenna in the kinematic mode (full phase, C/A 68 S.K. Dewali et al. Figure 6. Road position initally and after geometric correction. Available in colour online. narrow code, position update rate 10 Hz, time for each measurement 15–20 min) and post-processing was done using SKI-Pro L1 software. The positional accuracy after post-processing was found to be 20–40 cm. These points were then used for orthorec- tification of the Cartosat-1 images, geometric correction of the road and avalanche sites of the study area. Figure 6 shows the initial and corrected road network. In the horizontal plane, a shift of 30 to 70 m was found in the data. This initial error in data was corrected by applying geometric correction using commercial GIS/image processing software. Using this corrected dataset the second part of the experiment was conducted. In this part, a total of 12 track tests at different locations of the study area were conducted from 19 Feb 2012 to 25 Feb 2012 (figure 7). Details of each trial are given in table 4. During the trials, most of the study area was snow covered, a snow depth of 2–3 m was observed in the navigation area and air temperatures were mostly found below 0 C. The working of the GPS system was found satisfactory under such environmen- tal conditions. 5. Trial results and analysis The experimental results demonstrate the performance of the proposed system. To understand the stability and consistency of the GPS measurement in the study re- gion, navigation trajectories of the complete track during the trials 10 and 11 are shown in figure 8. GPS positions are found consistent throughout the navigation and follow the reference road given in figure 7. Figures 9(a) and (b) show the results of the various trials (trial 6, 7, 8, 9, 11, 12) for the two different parts of the experimental track of approximate lengths of 1.5 km and 2.5 km, respectively. All the trials are following the same path within the range of 0.5–2.5 m along the straight/smooth parts of the test track, while along the curved and zigzag part of the test track the de- viation among different trials is observed within the range of 1.0 to 5.5 m. The speed of the traveller for different trials was (0.25 m/s to 1.4 m/s) and the update rate of system was 5 s. A GPS-based real-time avalanche path warning 69 Figure 7. Expermental plan of field trials. For the analysis of the GPS position accuracy, single point positions at a fixed location were collected in an open area for five hours using the proposed GPS receiver. Figure 10 shows the scatter plot of these single point positions. The CEP of the dataset is foundtobe1.48 m andDRMS and 2DRMSare 1.83 and3.66, respectively. This Table 4. Field experiment details. Mode of Number of Trial Time Approximate path movement/speed avalanche no. Date (hr) length covered (km) (km/hr) sites crossed 1. 19 Feb 12 0900–1300 5 (Manali and nearby) Onfoot (1–3 km/hr) 0 2. 20 Feb 12 1600–1800 5 (Manali and nearby) Onfoot (1–3 km/hr) 0 3. 25 Feb 12 1600–1830 5 (Manali and nearby) Onfoot (1–3 km/hr) 0 4. 19 Feb 12 1400–1800 7 (Palchan and nearby) Onfoot (2–3 km/hr) 1 5. 21 Feb 12 1600–1830 7 (Palchan and nearby) Onfoot (1–3 km/hr) 1 6. 22 Feb 12 0900–1300 7 (Palchan and nearby) Onfoot (1–3 km/hr) 1 7. 20 Feb 12 0830–1500 12 (Manali to Solang) Onfoot (2–4 km/hr) 3 8. 22 Feb 12 1400–1830 12 (Manali to Solang) Onfoot (2–4 km/hr) 3 9. 21 Feb 12 0830–1430 12 (Manali to Solang) Onfoot and vehicle 3 (2–10 km/hr) 10. 23 Feb 12 0830–1630 19 (Manali to Dhundi) Onfoot (2–5 km/hr) 10 11. 24 Feb 12 0900–1500 19 (Manali to Dhundi) Onfoot and vehicle 10 (2–10 km/hr) 12. 25 Feb 12 0900–1400 19 (Manali to Dhundi) Onfoot and vehicle 10 (2–30 km/hr) 70 S.K. Dewali et al. Figure 8. Trajectories followed during full study track navigated during trials 10 and 12. indicated that 95% of the position measurements are within 3.66 m from the fixed location. The three measures used in characterizing the accuracy of the position, i.e. (i) hori- zontal dilution of precision (HDOP), (ii) vertical dilution of precision (VDOP) and (iii) position dilution of precision (PDOP) were calculated for all the trials. The histo- grams of these error measures are shown in figures 11(a)and (b). The analysis of the histograms showed that for more than 90% of the position measurements, the values of HDOP and VDOP were found in the range of 1.0 m–1.5 m and 1.5 m to 3.5 m, re- spectively. This consistent lower value of HDOP indicates the satisfactory perfor- mance of the system in the study area. Variations of these measures with different track conditions were also studied (figure 12). An instantaneous value of these meas- ures was found highly fluctuating, so the simple moving averages (using 5 points be- fore and 5 points after the target point) of these measures were calculated to find out the spatial average value of these measures (figure 12). The average values of HDOP, VDOP and PDOP for the open areas were found between 1.0 m–1.5 m, 1.75 m– 2.75 m and 2.0 m–3.0 m, respectively. The values of these measures were higher for narrow (HDOP 1.0 m–2.25 m, VDOP 2.0 m–5.0 m, PDOP 3.0 m–5.5 m) and forest- ed (HDOP 1.5 m–3.25 m, VDOP 3.0 m–7.0 m, PDOP 3.5 m–7.5 m) regions. The accuracy of the GPS position depends on the number of satellites available for position estimation. A variation in different error measures and their averages along with the satellites’ availability number is shown in figure 13. Low and variable satel- lite availability was found in narrow and forested regions, leading to a high value of all error measures. For a variable satellite availability number < 6, the different error measure averages were high (HDOP 1.5 m–3.25 m, VDOP 3.25 m–7.5 m, PDOP 3.5 m–8.0 m) and for variable satellite availability number 6 the different error measure averages were relatively low (HDOP 1.25 m–1.75 m, VDOP 1.75 m–3.0 m, PDOP 2.0 m–4.0 m). The dependence of HDOP and VDOP on the number of satel- lites was analyzed. Figures 11(c) and (d) show the variation of these measures with A GPS-based real-time avalanche path warning 71 Figure 9. (a) and (b) Comparison of path followed by the traveller during different trials of the two different parts of the study area. the number of satellites available for position estimation. The mean value of both HDOP and VDOP was found decreasing with an increasing number of available sat- ellites. As the number of satellites increased from three to six, the HDOP and VDOP decreased significantly from 4 m to 1.5 m and 6 m to 3.25 m, respectively and after satellite availability number 6 the variation in these measures was not prominent. The box plot of HDOP and VDOP with satellite number also showed the same trend. In this figure on each box the central mark is median, edges of the box are the 25th and 75th percentiles and the whiskers extend to the most extreme data points. The range of the data was found to reduce significantly with increasing satellite numbers. With increasing satellite numbers the skewness of the data was found to be 72 S.K. Dewali et al. Figure 10. Scatter plot of the single point positions at a fixed location in an open area along with CEP (50% probability), DRMS (65% probability) and 2DRMS (95% probability). Figure 11. Statistical analysis of the error measures and their dependence on the number of satellites available for estimating position and speed of the traveller. (a) and (b) Histogram of the HDOP and VDOP of test data. (c) and (d) Variation of the mean and range of HDOP and VDOP with number of satellites available. (e) and (f) Variation of the mean and range of HDOP and VDOP with the speed of the traveller. A GPS-based real-time avalanche path warning 73 Figure 12. Three measures used in characterizing the accuracy of the position: (i) horizontal dilution of precision (HDOP), (ii) vertical dilution of precision (VDOP), (iii) position dilution of precision (PDOP) and their moving averages, along with the track conditions. Available in colour online. decreasing and the mean and median values were coinciding. So the measurements with six or more satellites were found mostly consistent and accurate. The effect of the speed of the traveller on the accuracy of the GPS position was an- alyzed (figure 14). A variation in different error measures and their averages along with different traveller speeds is shown here. Average error measures didn’t show sig- nificant variations in their values with the increasing speed of the traveller, but the fluctuations in the instantaneous values of the error measures increases with increas- ing speed. To understand the dependence of HDOP and VDOP on the speed of the traveller, the values of these measures were grouped into 12 equal speed intervals of 1 m/s each, starting from 0 m/s to 12 m/s (figures 11(e) and (f)). The mean of HDOP and VDOP is calculated for each group. The mean of these measures didn’t show Figure 13. Variation in HDOP, VDOP and PDOP and their moving averages with the number of satellites available for position estimation. Available in colour online. 74 S.K. Dewali et al. Figure 14. Variation in HDOP, VDOP and PDOP and their moving averages with the speed of the traveller. Available in colour online. any significant dependence on the speed of the traveller. Also, the box plot of these groups didn’t show any trend with speed. For the validation of the GPS measured position coordinates, a total of 15 ground control points, which were previously iden- tified all along the track, were used for comparison. Using the measured GPS coordi- nates of known control points and reference coordinates of these points, X-residual, Y-residual and RMSE were calculated for all the known points (figure 15). The mean RMSE of the observed GPS positions with respect to reference position was found 2.99 m with a minimum of 1.12 m and a maximum of 5.92 m. This small RMSE indicated that GPS measured positions of control points are close to the Figure 15. Comparison of the GPS-measured position coordinates and reference coordinates at different ground control points (a) X-Residual, (b) Y-Residual and (C) RMSE. A GPS-based real-time avalanche path warning 75 Figure 16. Variation of the update rate error (UPER) with different update rate (t) of the system and the speed of the traveller. reference coordinates and hence the GPS position can be used for navigation in the study region within this error range. Although the effect of the speed of the traveller on the accuracy of the GPS posi- tion was not found very significant (figure 14), but it significantly affected the update error (UPER) of the system as given in equation (11). In figure 16, a variation in UPER with varying speed and varying update rates (t) is demonstrated. Suppose the update rate ‘t’, is 1 s and if the speed is 0.5 m/s, 1 m/s and 10 m/s, then the UPER will be 0.5 m, 1 m and 10 m, respectively, because the system updates the informa- tion after 1 s only. In another case, suppose the speed is 2 m/s and if the update rates are 1 s, 5 s and 10 s, then the UPER will be 2 m, 10 m and 20 m, respectively. There- fore, to make the system usable for the present application, the speed and update rate should be adjusted in such a way, so that MaxUPER  2:0m. 5.1 Avalanche path navigation and warning After the analysis of the GPS measured position, trajectories of different trials and MaxUPER introduced by the system, performance of the system is analyzed for real-time identification avalanche sites and generating various text and audio mes- sages. Screen shots of system performance in predicting various route conditions (table 3) are given in figure 17(a), (b), (c) and (d). For this purpose, 10 avalanche site polygons and their buffers of the study area were taken. Most of the avalanche sites affecting the track have already been marked by the SASE and signboards have been placed at the point where the route crosses the avalanche paths (i.e. at both the ends of track length affected by each avalanche path). These points were used in the map for navigation and also as reference points in the field during trials. The field trials 10 and 11 were used for this analysis. Throughout the navigation, updates and the mes- sages generated by the system, were monitored carefully. Figure 17(a) is the case of safe route condition, and the prediction of this condition was done satisfactorily 76 S.K. Dewali et al. most of the time by the system. Whenever the system generated the warning for ‘inside an avalanche path’ condition (figure 17(c)), the position on ground was marked and the distance of this location from the point where the route crosses the corresponding avalanche path (i.e. location of nearest signboard) was measured dur- ing the navigation. Results of the navigation are given in table 5. The mean and max- imum distance between the predicted and actual starting point of various avalanche Figure 17. (a) ‘Safe route’ condition update generated by system. (b). ‘Approaching to ava- lanche site’ condition alert with details of avalanche site. (c). ‘Entering in an avalanche path’ condition warning with details of avalanche site. (d). ‘Out of the avalanche path’ condition up- date with details of the avalanche site. A GPS-based real-time avalanche path warning 77 Figure 17. Continued. sites ranges between 2 m to 3 m and 5 m to 7 m, respectively. This is the total error (TE) of the system and it is the sum of error in GPS position, error in registration of avalanche site (EAP) and error due to update rate (UPER) of the system. Although all these error components have been calculated in the present study, further analysis is required for the quantification of each error component. From this analysis, it is suggested that a buffer of 7 m around the avalanche polygon be used instead of a 5 m buffer, so that the ‘approaching to avalanche site’ condition alert (figure 17(b)) is generated by the system before entering the avalanche zone, even if there is an error of 7 m in predicting an avalanche site. 78 S.K. Dewali et al. Table 5. Mean and maximum error distances in predicting avalanche site polygons with different speeds and update rates. Distance from reference point at which warning generated (m) Total number of Avalanche Update Speed measurements in site rate (s) (m/s) Mean Maximum trials 10, 11 Avl Site 1 1, 5 1.0–0.3 No reference available 5 Avl Site 2 1, 5 1.0–0.3 No reference available 4 Avl Site 3 1, 5 1.0–0.3 2.1 6.2 6 Avl Site 4 1, 5 2.0–0.2 2.2 5.7 4 Avl Site 5 1, 5 4.0–0.3 2.4 5.6 4 Avl Site 6 1, 5 1.0–0.3 1.9 4.8 6 Avl Site 7 1, 5 3.0–0.1 3.2 10.6 4 Avl Site 8 1, 5 1.0–0.3 2.9 6.5 5 Avl Site 9 1, 5 4.0–0.2 3.0 6.2 6 Avl Site 10 1, 5 1.0–0.3 2.8 5.6 4 6. Conclusions and future scope This paper proposes the design and development of an avalanche warning and nav- igation system by implementing a customized user-friendly application in a hand- held GPS (eXplorist Pro 10). The performance of the GPS under various track conditions was evaluated. The reduction in positional accuracy was observed in narrow valley and forested areas (average value of HDOP 1.0 m–1.5 m for open areas and 2.0 m–3.5 m for narrow and forested regions). The dependence of HDOP and VDOP on the number of satellites available for position estimation was monitored. With increasing satellite numbers, the value of these measures was found to be decreasing and the measurements with the six or more satellites were found more consistent and accurate. No significant effect of the speed of traveller on the GPS positional accuracy was found. Also the effect of the UPER in the overall accuracy of the system was analyzed. The update rate and speed of the trav- eller was found to be very crucial in detecting the avalanche polygons. During the numerous field trials, the system was able to locate avalanche polygons with a mean error of 2 to 3 m and a maximum error of 5 to 7 m. Therefore, by taking the appropriate buffer size, this system can guide the travellers in snowbound mountain regions as a warning and navigation tool. In future, the system can be developed in the following ways: one, the proposed system works in a standalone mode, in case of any emergency there is no way to report avalanche accident or call for help if a traveller falls victim. So, in future, the system can be integrated with a radio com- munication device to establish a radio link for short-range communication between the proposed system and a central station. Two, as the use of GNSS-based applica- tions in mobile phones is gaining popularity due to fast development of the mobile phone technology, so the implementation of this application on the mobile phone platform will be more useful. The current application can be deployed on any win- dows-based mobile phone. In future, the application can be developed for other mobile phone operating systems. A GPS-based real-time avalanche path warning 79 Acknowledgements The authors are grateful to Dr M.R. Bhutiyani for providing continuous support and encouragement during the present study. The authors would like to thank all the SASE personnel who have helped during this study. References ALOI DN, ALSLIETY M, AKOS DM. 2007. Methodology for the evaluation of a GPS receiver performance in telematics applications. IEEE Trans Instrum Meas. 56:11–24. BARBOLINI M, KEYLOCK CJ. 2002. A new method for avalanche hazard mapping using a combi- nation of statistical and deterministic models. Nat Hazards Earth Syst Sci. 2:239–245. BEEHAREE AK, STEED A. 2006. A natural wayfinding exploiting photos in pedestrian naviga- tion systems. ACM International Conference Proceedings Series, Proceedings of the 8th Conference on Human-computer Interaction with Mobile Devices and Services; Helsinki, Finland, 159:81–88. BESSHO M, KOBAYASHI S, KOSHIZUKA N, SAKAMURA K. 2008. Implementation and deployment of a place-based pedestrian navigation system. Proceedings of the 32nd Annual IEEE International Computer Software and Applications Conference (COMPSAC 2008); Turku, Finland. CHEN W, FU Z, CHEN R, CHEN Y, ANDREI O, TUOMO,WANG J. 2009. An integrated GPS and multi-sensor pedestrian positioning system for 3D urban navigation. Proceedings of IEEE (Urban Remote Sensing Joint Event). DEPARTMENT OF DEFENSE. 1996. NAVSTAR GPS user equipment introduction (public release version) GPS joint program office, Space and Missiles System Centre, Los Angeles, CA. GLEASON JA. 1994. Terrain parameters of avalanche starting zones and their effects on ava- lanche frequency. International Snow Science Workshop (ISSW): Snowbird, Utah, USA, p. 393–404. GOLLEDGE RG, LOOMIS JM, KLATZKY RL, FLURY A, YANG XL. 1991. Designing a personal guidance system to aid navigation without sight: Progress on the GIS component. Int J Geogr Inf Syst. 5:373–396. GOLLEDGE RG, KLATZKY RL, LOOMIS JM, SPEIGLE J, TIETZ J. 1998. A geographic information system for a GPS-based personal guidance system. Int J Geogr Inf Sci. 12:727–749. GREWAL MS, WEILL LR, ANDEREWS AP. 2002. Global positioning systems, inertial navigation and integration. New York: Wiley. GRUBER U. 2001. Using GIS for avalanche hazard mapping in Switzerland. Proceedings of the 2001 ESRI International User Conference, San Diego. HASAN AM, SAMSUDIN K, RAMLI AR, AZMIR RS, ISMAEEL SA. 2009. A review of navigation systems (integration and algorithms). Aust J Basic Appl Sci. 3:943–959. HILE H, VEDANTHAM R, LIU A, GELFAND N, CUELLAR G, GRZESZCZUK R, BORRIELLO G. 2008. Landmark-based pedestrian navigation from collections of geotagged photos. Pro- ceedings of ACM International Conference on Mobile and Ubiquitous Multimedia (MUM 2008). INOUE Y, SASHIMA A, KURUMATANI K. 2009. Indoor positioning system using beacon devices for practical pedestrian navigation on mobile phone. Lecture Notes In Computer Sci- ence, Vol. 5585, Proceedings of the 6th International Conference on Ubiquitous Intel- ligence and Computing; Brisbane, Australia, p. 251–265. KAPLAN ED. 1996. Understanding GPS principles and applications. Norwood, MA: Artech House. KAY S, SPRUYT P, ALEXANDROU K. 2003. Geometric quality assessment of orthorectified VHR space image data. Photogramm Eng Remote Sens. 484–491. KLUKAS R, LACHAPELLE G, MA C, JEE G. 2003. GPS signal fading model for urban centers. IEE Proc, Microw Antennas Propag. 150:245–252. KRALI D, HSU T, CARIN L. 2000. On GPS positioning and integrity monitoring. IEEE Trans Aerosp Electron Syst. 36:327–336. 80 S.K. Dewali et al. LADETTO Q, GABAGLIO V, MERMINOD B. 2001. Two different approaches for augmented GPS pedestrian navigation. International Symposium on Location Based Services for Cellu- lar Users; Locellus. LE AQ, TIBERIUS CCJM. 2007. Single-frequency precise point positioning with optimal filter- ing. GPS Solut. 11:61–69. LI Y, XI X. 2007. Combining FFT and circular convolution method for high dynamic GPS sig- nal acquisition. Eighth International Conference on Electronic Measurement and Instruments; ICEMI, 2, p. 159–162. LOOMIS JM, GOLLEDGE RG, KLATZKY RL, SPEIGLE JM, TIETZ J. 1994. Personal guidance sys- tem for the visually impaired. Proceedings of the First Annual International ACM/ SIGCAPH Conference on Assistive Technologies (Assets ‘94); Marina Del Rey, Cali- fornia, p. 85–91. MA C, JEE G, MACGOUGAN G, LACHAPELLE G, BLOEBAUM S. 2001. GPS signal degradation modeling. ION GPS 2001, 11–14 September, Salt Lake City, UT, p. 882–893. MAGGIONI M, GRUBER U. 2003. The influence of topographic parameters on avalanche release dimension and frequency. Cold Reg Sci Technol. 37:407–419. MARTIN JD, KROSCHE J, BOLL S. 2006. Dynamic GPS-position correction for mobile pedestrian navigation and orientation. Proceedings of the 3rd Workshop on Positioning, Naviga- tion and Communication (WPNC’06), p. 199–208. MCCOLLISTER C, BIRKELAND K, HANSEN K, ASPINALL R, COMEY R. 2002. A probabilistic tech- nique for exploring multi-scale spatial patterns in historical avalanche data by combin- ing GIS and meteorological nearest neighbors with an example from the Jackson Hole Ski Area Wyoming. International Snow Science Workshop (ISSW) Proceedings. MLADEN M, ZORAN S, STEVAN B. 2006. A comparison of accuracy using a GPS and low-cost DGPS. IEEE Trans Instrum Meas. 55:1677–1683. NADEEM A, ANJUM M, RITESH A, JAYAPRASAD P, PATHAN SK, AJAI,SINGH,DK, SINGH AK. 2007. Extraction and Validation of Cartosat-1 DEM. J Indian Soc Remote Sens. 35:121–127. NOVATEL REPORT. 2003. GPS position accuracy measures, APN-029 Rev 1. Available from: http://www.novatel.com. ØVSTEDAL O. 2002. Absolute positioning with single frequency GPS receivers. GPS Solut. 5:33–44. SNOW AND AVALANCHE STUDY ESTABLISHMENT (SASE). 2010. Internal report on avalanche haz- ard mitigation scheme, 16. SATO G, ASAI T, SAKAMOTO T, HASE T. 2000. Improvement of the positioning of a software- based GPS receiver using a 32-bit embedded microprocessor. IEEE Trans Consum Electron. 46:521–530. STARK A, RIEBECK M, KAWALEK J. 2007. How to Design an Advanced Pedestrian Navigation System: Field Trial Results. Proceedings of IEEE International Workshop on Intelli- gent Data Acquisition and Advanced Computing Systems: Technology and Applica- tions; Dortmund, Germany, p. 690–694. TOTH C, GREJNER-BRZEZINSKA DA, MOAFIPOOR S. 2007. Pedestrian tracking and navigation using neural networks and fuzzy logic. Proceedings of IEEE International Symposium on Intelligent Signal Processing. TRACY L. 2001. Using GIS in avalanche hazard management. Proceedings of the 2001 ESRI International User Conference; San Diego. VAN BREE RJP, TIBERIUS CCJM. 2012. Real-time single-frequency precise point positioning: Accuracy Assessment. GPS Solut. 16:259–266. WASINGER R, STAHL C, KRUGER A. 2003. M3I in a pedestrian navigation & exploration system. Mobile HCI, Springer, p. 481–485. XUCHU M, MASSAKI W, HIDEKI H. 2000. Nonlinear GPS models for position estimate using low-cost GPS receiver; p. 637–642.

Journal

"Geomatics, Natural Hazards and Risk"Taylor & Francis

Published: Mar 1, 2014

There are no references for this article.