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

Learn More →

Intelligent Obstacle Avoidance Algorithm for Mobile Robots in Uncertain Environment

Intelligent Obstacle Avoidance Algorithm for Mobile Robots in Uncertain Environment Hindawi Journal of Robotics Volume 2022, Article ID 8954060, 9 pages https://doi.org/10.1155/2022/8954060 Research Article Intelligent Obstacle Avoidance Algorithm for Mobile Robots in Uncertain Environment 1,2,3 1,2,3 1,2,3 1,2,3 Liwei Guan , Yu Lu, Zhijie He, and Xi Chen College of Physics and Energy, Fujian Normal University, Fujian Provincial Key Laboratory of Quantum Manipulation and New Energy Materials, Fuzhou, Fujian 350117, China Fujian Provincial Collaborative Innovation Center for Optoelectronic Semiconductors and Efficient Devices, Xiamen, Fujian 361005, China Fujian Provincial Engineering Technology Research Center of Solar Energy Conversion and Energy Storage, Fuzhou, Fujian 350117, China Correspondence should be addressed to Liwei Guan; guanlw@fjnu.edu.cn Received 30 January 2022; Revised 7 March 2022; Accepted 9 March 2022; Published 30 March 2022 Academic Editor: Shan Zhong Copyright © 2022 Liwei Guan et al. +is is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited. +e application of mobile robots and artificial intelligence technology has shown great application prospects in many fields. +e ability of intelligent obstacle avoidance is the basis for the deep application of mobile robots. However, there are often more or less uncertain factors in the actual operating environment of the robot, such as people or objects that are not updated in time or temporarily appear. +erefore, it is an important step to complete the automatic learning of obstacle avoidance for mobile robots. In a nondeterministic environment, a mobile robot intelligent obstacle avoidance algorithm based on an improved fuzzy neural network with self-learning is firstly proposed. +e mobile robot intelligent obstacle avoidance system is constructed through the reaction layer, the deliberation layer, and the supervision layer. +rough the analysis of sensor performance, model accuracy, path obstacle avoidance optimization, and obstacle avoidance simulation, the following conclusions are drawn. First, through network training, the accuracy rate of the test set is stable at 98%, and the loss of the function value has also been reduced from the original 0.79 to 0.08, which is 10 times smaller. Second, the traditional single sensor cannot meet the obstacle avoidance requirements of robots, and mobile robots must combine multipurpose technology. +ird, the algorithm in this paper encounters the following. When there are obstacles, the path is dominated by straight lines, obstacle avoidance planning is optimal, and the distance is shorter. Fourth, the larger N : M, the larger the solution space, indicating that this algorithm gradually improves the search efficiency to the greatest extent and can handle any form of medium and large scale task allocation problem. On the basis of the continuous improvement of artificial 1. Introduction intelligence, the development of robot technology is be- Robot technology is an important manifestation and symbol coming more and more intelligent, and the ability of mobile of the degree of industrial automation and the national high- robots to intelligently avoid obstacles is an important in- dicator of their intelligence [2]. +ey not only reflect the tech level. Robot technology is a high-tech formed on the basis of interdisciplinary. It is one of the hot spots of current efficiency, feasibility and energy consumption of mobile scientific research [1]. Mobile robots are widely used in robot motion but also reflect the way the robot detects various fields, especially in harsh and dangerous environ- obstacles, processes obstacle information, and avoids ob- ments such as military reconnaissance, antinuclear pollu- stacles. Today’s mobile robots have moved from structured tion, mine clearance, and material handling in civil use. work spaces to uncertain environments [3]. In an unknown +erefore, mobile robots greatly facilitate people’s pro- and uncertain environment, it is an important step for duction and life. +e research on robots has received mobile robots to learn to avoid obstacles autonomously. It widespread attention. enables mobile robots to have human-like behaviour 2 Journal of Robotics strategies and avoid obstacles, thereby realizing intelligent out that based on global path planning, navigation and navigation of mobile robots [4]. How to intelligently avoid obstacle avoidance are performed according to the planned path method [13]. Jin J. combined fuzzy logic and grid graph these obstacles unknown to the environment map in an uncertain environment is a key link of mobile robots and one in the target navigation of mobile robots, using the mini- of the research hot spots of robotics. mum risk criterion as the evaluation function to improve the At present, great progress has been made in the research effect of path planning [14]. Zafar M. proposed a learning of autonomous obstacle avoidance algorithms for mobile algorithm based on neural network error backpropagation robots in the world, and many algorithms are applied in the to adjust the membership function parameters of the fuzzy autonomous obstacle avoidance systems of mobile robots. logic system to improve the trajectory smoothness of the +ese intelligent obstacle avoidance algorithms mainly in- mobile robot [15]. Cui Min applied the improved neural clude fuzzy control, grid map, artificial potential field, and network algorithm to the path planning of mobile robots, neural network [5]. +e fuzzy control algorithm has the which improved the operation efficiency of the algorithm advantages of simple algorithm, easy to understand, and [16]. Chen D. proposed a reinforcement learning mecha- nism based on fuzzy neural network, which utilizes the strong robustness, but there are also problems such as the need for experience in design, low control accuracy, and no residual in the learning algorithm and obtains a better al- learning ability; neural network algorithms can use a large gorithm convergence speed in the navigation process of the amount of data to train models, which can automatically mobile robot [17]. learn parameters to obtain the final end-to-end network In the previously mentioned research, different algo- model, but there are problems of high network complexity rithms are used in mobile robot path planning and navi- and practicability. +is paper fully considers the practica- gation and obstacle avoidance, and the performance of the bility and operability of intelligent obstacle avoidance of system has been improved to a certain extent. However, in mobile robots, combines fuzzy control algorithm with the nondeterministic environment, the previously men- tioned obstacle avoidance and planning algorithms have neural network algorithm, and takes the lead in proposing an intelligent obstacle avoidance algorithm for mobile robots some deficiencies in flexibility, real-time, humanization, and intelligent performance. +is paper focuses on the problems based on improved fuzzy neural network with self-learning. existing in the previously mentioned algorithms. In a nondeterministic environment, an intelligent obstacle 2. Related Work avoidance algorithm for mobile robots based on an im- A mobile robot is a robot system that completes certain work proved fuzzy neural network with self-learning is firstly functions. It can sense the external environment and its own proposed, which improves the flexibility, autonomy, and situation through sensors and realize its autonomous stability of mobile robots. movement in an environment with obstacles [6]. Mobile robots have become a research hot spot in robotics because 3. Implementation of Intelligent Obstacle they have shown more and more extensive application Avoidance Algorithm prospects in various aspects of agriculture, industry, aero- space, medicine, and human life [7]. +is paper fully considers the importance of obstacle +e earliest research on mobile robots was from 1966 to avoidance for mobile robots in nondeterministic environ- 1972, when Nils Nilssen and Charles Rosen of Stanford ments and proposes an intelligent obstacle avoidance al- Research Institute developed an autonomous mobile robot gorithm for mobile robots based on improved fuzzy neural [8]. In the 1980s, a number of universities such as Stanford network autonomous learning. First, in view of the high and MIT established a special scientific research team for complexity and poor real-time performance of the fuzzy ALV research [9]. In the 1990s, Japan developed working algorithm, the necessary simplification of the fuzzy neural robots that can operate in extreme environments and suc- network is carried out to perform image preprocessing and cessfully developed humanoid robots [10]. In 2003, Tsinghua rough positioning of obstacles. Second, the parameters of the University in China independently developed the robot fuzzy membership function are automatically and dynam- automatic navigation system [11]. All over the world, Japan ically adjusted by the neural network with self-learning has always been at the forefront of the world in the field of ability, so that the fuzzy control rules have stronger object humanoid robots and wheeled robots. +e ASIMO biped adaptability. Furthermore, for dynamic obstacles in au- walking robot developed by Honda represents the highest tonomous obstacle avoidance, edges are introduced. +e level in the world. ASIMO uses a variety of sensors in the detection operator implements the dynamic obstacle realization of its functions, including cameras on the head, avoidance strategy of the mobile robot. ultrasonic sensors around, and ground pressure sensors on the bottom of the feet. In terms of functional realization, ASIMO not only realizes the functions of walking, posi- 3.1. Improved Fuzzy Neural Network Algorithm. +is paper tioning, and navigation but also adds functions such as face applies the improved fuzzy neural network algorithm to the recognition and communication by voice or gesture [12]. field of image processing, analyzes the image to detect ob- In recent years, breakthroughs have been made in the stacles by counting the neurons in the visible range, and uses research on the theory and algorithm of navigation control the knowledge of distance images to convert the detection of mobile robots in unknown environments. Yan Z. pointed results to finally determine the location of the obstacles. Journal of Robotics 3 3.1.1. Image Preprocessing and Coarse Positioning of Among them, D(X) represents the unit neuron of the Obstacles. Image preprocessing includes image greyscale, sample space X. Various colour components of an image can image denouncing, and effective image region selection. be used as the basic measure of the amount of information, Image greyscale is to complete the conversion from colour to and grayscale representation is also a method of expressing greyscale. In the system, the video image format conversion information. Each grayscale level is the basic element of the is completed through the acquisition of the camera. +e composition. +e probability of occurrence is used to value of the three colour components of R, G, and B in the characterize the amount of information in the area. RGB space of the image is stored in the image, and greyscale Use the knowledge of neurons to analyze the image and can be achieved by using locate obstacles. +e unit information entropy expression formula of the image is applied to a certain frame of gray � 0.5 × Red + 0.23 × Green + 0.27 × Blue. (1) greyscale image and can be written as +e size of the redundancy is related to the probability of 255 M � + 􏽘 kqi + log qi. (4) occurrence of each basic element in the information. Its i�0 expression is as follows: Among them, qi represents the probability of the oc- M(X) � E[log q(ai) × 10] � 􏽘 qi(ai)log qi(ai). (2) currence of a pixel whose gray level is i; k is the number of i�0 gray levels whose qi is not zero in all gray levels. In the actual experiment, the horizontal image analysis process is as Among them, ai is the event in the sample population follows: select a row of pixels as the data source for statistics X. +e previously mentioned formula indicates that the qi; that is, count the gray level distribution of the pixels in more average the number of occurrences of each sample this row, and obtain the corresponding pixel points in each event, the greater the number of different events in the gray level i. +e number of pixels, qi, is the ratio of the sample population X, and the greater the amount of in- number of pixels contained in the gray level to the total formation. If there are N groups of samples, X , X . . .X , the 1 2 n number of pixels in the row; then, the unit information corresponding sample numbers are k ,k . . .k , and the in- 1 2 n entropy is calculated for all gray levels whose qi is not 0. formation entropy is M ,M . . .M . +e unit information 1 2 n entropy is the average value of the information contained in the unit sample in the sample population, which is embodied 3.1.2. Accurate Positioning of Obstacles in the Image. as the ratio of the population neuron to the number of spatial Since the experimental environment is relatively simple, the samples, namely, relatively simple Roberts edge detection operator can be used for detection to obtain edge point information, which D(X) � M(X)∗ n � 􏽘 qi(ai)log qi(ai) × . (3) can effectively improve the real-time performance of the i�0 system. +e Roberts operator in an image pixel array looks like this 􏽱������������������������������������������� 2 2 (5) k(a, b) � [t(a, b) + t(a + 1, b + 1)] − [t(a, b + 1) + t(a + 1, b)] . α � arctan h × y1, (7) +e Roberts operator is a diagonal derivative computed in a 2×2 neighbourhood. In practical applications, the β � arctan h × (y1 − y2), (8) previously mentioned formula can be replaced by a sim- plified calculation form, which is expressed as follows; that λ � arctan(x1 × y1), (9) is, the absolute value of Roberts is replaced. where h, y1, y2, and x1 are data that can be measured. After k(a, b) � |t(a, b) + t(a + 1, b + 1)| + |t(a, b + 1) + t(a + 1, b)|. (6) obtaining α, β, and c, x and y can be obtained from the trigonometric function relationship, and the derivation formula is as follows: 3.1.3. Distance Image Obstacle Location. +e range image y � h × tan􏽨(45 + α) − 􏼐s − u􏼑s × (α + β)􏽩, y y stores the depth information of the ray associated with each pixel and the first focus of the scene observed by the camera. x � y × tan􏼂s + v􏼃 × s + c, (10) x x According to the idea of interpolation, using the geometric 􏽱������ 􏽱������ 2 2 2 2 principle of camera imaging, using Manlius’s theorem, es- B � x − y × x + y . tablish a three-dimensional distance function, calculate the Among them, B is the distance between the vertical distance from any position in the monitoring area to the projection points of the camera coordinates; S and S are the camera, and establish a distance image about the distance x y total number of pixels in the x and y directions in the image between any point in the monitoring area and the camera: 4 Journal of Robotics platform of the ideal hybrid architecture. In this chapter, we plane; u and v represent the horizontal and vertical imaging planes, respectively. are based on the mobile robot intelligent obstacle avoidance architecture, follow the general design principles of intel- +e position of the target point in ABCD can be obtained by formulas (7)–(9), that is, the coordinates of the target ligent system structure, and add supervision management point in the world coordinate system. By analyzing the image and learning functions on the basis of the original hybrid through image neurons and edge information in the image, structure. +rough the improved fuzzy neural network al- the characteristic edge points of obstacles can be obtained. gorithm, the robot can learn behavior autonomously in the +rough the knowledge in this section, these correspon- dynamic unknown environment so as to adapt to the dences with the world coordinate system are established to changes of the environment, as shown in Figure 2. obtain the position of the obstacle in the actual situation. Reactive behaviour control layer contains reflective behaviour and reactive behaviour, as well as a structure for coordinating reactive behaviour. Reflective behaviour is 3.2. Intelligent Obstacle Avoidance System. A very important used to react to emergencies in the environment, or it can be task area for mobile robots is navigation, which is to find a a single control rule. For the navigation tasks of mobile route so that it can move safely without collisions around all robots, reactive behaviors are designed for obstacle avoid- obstacles. +is planning technology that can autonomously ance, goal orientation, and roaming. +ese behaviors can be avoid obstacles and complete tasks is the frontier technology pare-designed by expert experience or acquired through for intelligent mobile robots to achieve autonomous be- learning and evolution. All behavior can directly respond to haviour control. +erefore, it is the most basic problem to the information felt from the outside world and can also apply the improved fuzzy neural network algorithm to the receive control signals from the deliberation layer and the intelligent obstacle avoidance system of mobile robots. +e supervision layer to perform actions. mobile robot intelligent obstacle avoidance system designed Regarding the deliberate behaviour control layer, the in this paper integrates different functions in the system, as task for the navigation class includes five basic modules, and shown in Figure 1. other functions can be added on this basis. +e modules are Much of the structure of the intelligent obstacle environment model knowledge base, task planning module, avoidance system involves planning, using a reactive planner positioning module, navigation module, and sequencer. +e called PRS-L. PRS-L can accept human natural language environmental model knowledge base can directly receive instructions, and then, start to run navigation tasks and the environmental model transmitted from the outside perceptual recognition routines [18]. Both planning and world and store it in the knowledge base through human- execution rely on a locally aware spatial-centrist model of the computer interaction and can also generate a map based on environment. +e reactive components of an intelligent the collected sensor information, maintain the map infor- architecture consist of behavior. +ese behavior extract mation in time, and provide orientation data to other virtual sensor input and output fuzzy rules from the local modules. +e task planning module receives the instructions perception space of the central environment model and then input by the user and transmits it to the robot for task synthesize control instructions through fuzzy logic. planning. +e positioning module uses the shaft angle en- From the intelligent obstacle avoidance system, it can be coder combined with the feature quantity extracted from the seen that a hybrid architecture should have the following external sensor data to determine the position of the robot in modules and objects. the environment at each moment. Navigation accepts the Sequencer agent is used to generate the set of tasks task from the task planning module and calculates the path required to complete the subtasks and to determine all and decomposes it into subtasks. +e sequencer is used to timing and activation conditions. Timing is usually repre- generate the task set required to complete the subtasks and sented as a correlation network or finite state machine, and determine all the timing and activation conditions. When sequencers can generate these structures and modify them the purposeful planning is required, the reaction layer ob- dynamically [19]. tains the ordered actions through the sequencer. Resource manager is used to allocate resources for be- Supervise and manage the behaviour control layer: havior including selection of schema libraries. used to monitor the execution of the deliberation layer Cartographer is used to generate, store, and maintain and the behaviour layer, so that the robot can notice map or spatial information and provide a means of accessing whether it is making progress. At present, five modules are data. set up, namely, a cross module for coordinating and Mission planner interacts with the user, passes in- reflecting deliberation behavior, a fault supervision structions to the robot, and generates mission plans. module, a supervision module for the execution of be- Performance supervision and problem solving agents are havior at the reaction layer, a supervision module for used to let the robot notice if it is making progress. behaviour planning at the deliberation layer, and a learning evolution unit. +e learning evolution unit has 3.3. Intelligent System Hierarchy. Because the mobile robot two functions. One is to learn the local optimal path planning behaviour in a static environment of delibera- obstacle avoidance system adopts the modular design method based on mufti-agent, the functions are independent tion behaviour, and obtain the design of reactive be- of each other. +erefore, it is very easy to realize the function haviour autonomously. +e second is to use the collected expansion and can be used as the design and experiment sample data for training and learning in a dynamic Supervisory layer Deliberate layer Reaction control layer Actuator Navigation task Topology planner Environmental Model Environmental classifier Knowledge Journal of Robotics 5 Resource manager Mission planner Cartographer Smart logic Software agent Performance Detection Agent People tracking Planning agent (PRS-L) Object recognition Local perception space Surface build Reactive behavior Sensor Orientation map maintenance Sequencer virtual sensor Figure 1: Intelligent obstacle avoidance system diagram. Prior knowledge base HMI Task set Learning/Evolution React layer Deliberate layer Fault monitoring Unit monitoring monitoring Mission planning Task sequencer Map maintenance Map main i i i ten Process monitoring Navigation Posi Pi Position i iti tion i i Path sequencer module Rout Rt Rt R RRt R R R R Route plan t t t teplan Avoidance Behavioral Tracking fusion Pass through Reactive behavior Along the wall ON/OFF Towards the goal Task fusion Rotation Reflective behavior Kinetics/ Emergency stop Behavior Kinematics controller Other behavior Figure 2: Mobile robot hierarchy diagram. environment to establish a prediction model for dynamic 3.4. Obstacle Avoidance Control Process. In robot route obstacle avoidance. Learning the functions of evolution planning, various information obtained from the robot itself could enable robots to autonomously adapt to environ- and its environment is synthesized, enabling the robot to mental changes and enhance their intelligence. understand its environment and make decisions through 6 Journal of Robotics controller processing. So as to avoid obstacles, find the gradient learning. +is is a supervised learning process. optimal path and move autonomously [20]. Figure 3 shows Before adding the 1∗ 1 constitutional layer, after multiple the flowchart of the mobile robot path planning. straining, the final test set accuracy rate was stable at about +e mobile robot is equipped with a positioning system 95%. After adding the 1∗ 1 constitutional layer, the network that can detect the global position and heading of the mobile is retrained, the test set accuracy rate is stable at 98%, and the robot. Six ultrasonic sensors are used to detect local obstacle loss function value is also reduced from the original 0.79 to information. +e mobile robot performs detection every 1s, the current 0.08, which is 10 times smaller. Since the added and the acquired sensor data is fused as the input of the 1∗ 1 constitutional layer contains nonlinear units, the controller. By verifying the effectiveness of the previously nonlinear expression ability of the model is improved. After mentioned improved fuzzy neural network algorithm, a adding 1∗ 1 to two different curves, the fluctuation is ob- system simulation model is established in Mat lab with viously reduced compared with the previous curve, and it is Simulation, the control rules are simulated, and the fuzzy relatively stable. logic toolbox software is used to simulate the neural network algorithm. In this simulation system, the steps of path 4.2. Sensor Property Index Analysis. +e sensor control planning simulation are as follows. system and the information processing system constitute the (1) Establish environmental information: environmen- robot’s obstacle avoidance system. Among them, the sensor tal information is to establish the coordinates, di- control system is composed of sensors and microprocessors mensions of obstacles, and the starting point and that obtain environmental information. It mainly collects target point of the robot. information from unknown environments, which is the only way for robots to understand environmental information. (2) Establish a simulated robot, including some pa- +erefore, the reasonable configuration of the sensor de- rameters such as robot size and moving speed. termines the accuracy of the system’s acquisition of the According to the kinematic model of the robot in the external environment. actual system, it is assumed that the travelling speed ° As can be seen from Figure 5, the radar has the best of the mobile robot is 0.6 m/s, and it can realize 360 performance index and high cost. Ultrasonic and infrared in situ. have the lowest cost. In view of the need for obstacle (3) Establish a simulated sensor: used to perceive the avoidance, combined with the cost of the sensor, the simulated environment information, that is, to ob- performance index of the collected information, the tain the value of the obstacle distance information d hardware implementation circuit, volume, and other and the target direction angle ?. comprehensive factors, the sensor control system in this (4) +e robot controller is designed by an improved paper mainly uses ultrasonic and infrared sensors to collect fuzzy neural network algorithm, and the driving the obstacle distance of the external environment and instructions of the robot are obtained by analyzing electronic compass to obtain the target object. Considering the obstacle and target data obtained by the sensor. the unknowns and complexity of the external environment, (5) Transmission of control instructions: the driving the multipurpose technology is applied to the mobile robot obstacle avoidance system, and the hardware and algo- instructions are transmitted to the simulated robot, the mobile robot moves according to the instruc- rithms are processed accordingly to enhance the intelli- gence of the robot. tions, and then the map coordinates of the robot after executing the driving instructions are calculated; repeat steps (3) to (5) until the robot reaches the 4.3. Path Obstacle Avoidance Optimization Analysis. In a predetermined target point. complex environment with obstacles, the ultimate goal of path planning is to solve an optimal route, so that the robot can move smoothly and avoid all obstacles without collision 4. Intelligent Obstacle Avoidance Algorithm [21]. In practical applications, the environment is unknown Results and Analysis to the mobile robot. +e unknown environment, the de- 4.1. Algorithm Accuracy Analysis. In this experiment, the tection accuracy of the obstacle detector and the difference of two network models before and after the improvement were the path planning algorithm have a great influence on used for training, respectively, and the training results were whether the path planning can be successfully implemented. compared after 300 times of effective training. Before +is paper compares and analyzes the experimental paths of training, you need to import the saved network model data the BUG algorithm and the improved fuzzy neural network in advance, and use the training data set to continue training algorithm. the network model before the improvement. At the same It can be seen from Figure 6 that when the BUG algo- time, the loss value and the accuracy curve are obtained, and rithm is used, the robot moves along the edge contour of the the accuracy curve of the test set is obtained, as shown in obstacle. When it can move to the target point, it will directly Figure 4. leave the edge of the obstacle and move beyond the target It can be seen from Figure 4 that since the training, the point. It is not limited to the distance judgment between the loss value has been continuously reduced and the accuracy obstacle and the target point. H1 and H2 are the arrival has been continuously improved in the process of reverse points, and L1 and L2 are the separation points. When using Journal of Robotics 7 Accept ambient mode Start Determine the minimum distance Initialization Obstacle Type Robot state selection Capture the location of e location of the obstacles acquisition target Determine the minimum Are there any obstacles around? With wall type distance Determine the Into the dead zone minimum distance Output speed U-turn in place Target-oriented behavior Figure 3: Flowchart of intelligent obstacle avoidance path. 100 Before improvement Long distance Aer improvement 80 5 Target Dust 0 20 40 60 80 100 120 140 160 Cost 0 Exclude Before improvement Aer improvement Temperature All day 0 20 40 60 80 100 120 140 160 Night Figure 4: Model loss value and accuracy curve. Ultrasound Video Infrared Radar Laser the intelligent algorithm, the robot initially moves along the Figure 5: Sensor performance graph. line connecting the starting point and the target point. When encountering an obstacle, it moves along the tangential direction of the obstacle until no obstacle is detected and then updates the current point to the target point. It extends 4.4. Obstacle Avoidance Algorithm Simulation. In order to the updated main line and repeat this process until the target verify the performance of the algorithm, a lot of simulation point is reached. After comparison, it is found that the research has been done in this paper. Let the values of N and improved fuzzy neural network algorithm in this paper, the M be 4, 10, 30, and 60, respectively, the selection probability path distance is dominated by straight lines, the distance is is 0.1, the cross-mutation probability is 0.08, the a and ß shorter, and the distance is shortened by 5–10m. values are both 1, and the population evolution stops at 200 Loss rate Accuracy 8 Journal of Robotics 0 5 10 15 20 25 0 5 10 15 20 25 H1 H1 L1 L1 L2 H2 L2 H2 Distance (m) Distance (m) Figure 6: Comparison and analysis diagram of path obstacle avoidance optimization. 0.75 0.80 0.60 0.64 0.45 0.48 0.30 0.32 0.15 0.16 0 10 20 30 40 50 60 0 10 20 30 40 50 60 First group (time /ms) Second Group (time /ms) Figure 7: Obstacle avoidance algorithm simulation curve. or 1000 generations. Research the convergence performance 5. Conclusion of the algorithm evolution under the same parameters: the evolution algebra when the algorithm converges and the +e ability to autonomously avoid obstacles is the main execution time is consumed by the evolution calculation. indicator to measure the intelligence of mobile robots, and it +e changes of the maximum fitness value during the is also an important condition for intelligent robots to drive evolution of each group of experiments are shown in safely [22]. +is paper is the first to propose an intelligent Figure 7. obstacle avoidance algorithm for mobile robots based on It can be seen from the figure that, with the increase of improved fuzzy neural network, which achieves precise N : M, the method will gradually improve the search effi- positioning through adaptive learning. Following the general ciency. +e larger N : M is, the larger the solution space is, principles of intelligent control system design, on the basis of but the convergence speed does not change, which shows response and deliberation, a monitoring layer is added. that the search efficiency of this method is the largest as the Supervise and coordinate the implementation of deliberative value increases. When the solution space of the problem is layer behaviors to learn adaptive behaviors in unknown large, the speed of convergence is obviously slowed down, environments. By collecting sample data for training, it the algebra of convergence is improved, and the convergence learns to build predictive models to avoid dynamic obstacles. time is prolonged. +is further verifies that the method is +rough the obstacle avoidance experiment simulation, the suitable for medium and large scale problem spaces. In accuracy and real-time performance of the system are summary, the algorithm can handle any form of medium-to- further verified, and finally, a good obstacle avoidance effect large-scale task assignment problem. is achieved. Journal of Robotics 9 Journal of Electrical Engineering and Technology, vol. 12, Data Availability no. 11, pp. 918–925, 2017. [12] G. Lee and D. Chwa, “Decentralized behaviour-based for- +e data used to support the findings of this study are mation control of multiple robots considering obstacle available from the corresponding author upon request. avoidance,” Intelligent Service Robotics, vol. 11, no. 9, pp. 127–138, 2018. Conflicts of Interest [13] Z. Yan, J. Li, and G. Zhang, “A real-time reaction obstacle avoidance algorithm for autonomous underwater vehicles in +e authors declare that there are no conflicts of interest or unknown environments,” Sensors, vol. 18, no. 12, pp. 438–448, ethics in this article. 2018. [14] J. Jin and W. Chung, “Obstacle avoidance of two-wheel differential robots considering the uncertainty of robot mo- Acknowledgments tion on the basis of encoder odometer information,” Sensors, vol. 19, no. 17, pp. 289–376, 2019. +is study was supported by the Key Project of Industry [15] M. N. Zafar and J. C. Mohanta, “Methodology for path Guidance of Science and Technology Department of Fujian planning and optimization of mobile robots: a review,” Province“R and Dand application of large space automatic Procedia Computer Science, vol. 133, no. 32, pp. 141–152, 2018. tracking and positioningJet Fire Control System for Intel- [16] M. Cui, H. Liu, and W. Liu, “An adaptive unscented kalman ligent Security”(no.2020Y0021). filter-based controller for simultaneous obstacle avoidance and tracking of wheeled mobile robots with unknown slipping parameters,” Journal of Intelligent and Robotic Systems, References vol. 92, no. 46, pp. 489–504, 2018. [17] D. Chen, “Fuzzy obstacle avoidance optimization of soccer [1] S. M. H. Rostami, A. K. Sangaiah, and J. Wang, “Obstacle robot based on an improved genetic algorithm,” Journal of avoidance of mobile robots using modified artificial potential Ambient Intelligence and Humanized Computing, vol. 11, field algorithm,” EURASIP Journal on Wireless Communi- no. 12, pp. 6187–6198, 2020. cations and Networking, vol. 22, no. 4, pp. 1–19, 2019. [18] Z. Yi, G. Li, and S. Chen, “A navigation method for mobile [2] B. K. Patle, A. Pandey, A. Jagadeesh, and D. R. Parhi, “Path robots using interval type-2 fuzzy neural network fitting planning in uncertain environment by using firefly algo- Q-learning in unknown environments,” Journal of Intelligent rithm,” Defence technology, vol. 14, no. 6, pp. 691–701, 2018. and Fuzzy Systems, vol. 37, no. 35, pp. 1113–1121, 2019. [3] M. Radmanesh, M. Kumar, and P. H. Guentert, “Overview of [19] W. Zheng, H. B. Wang, and Z. M. Zhang, “Multi-layer feed- path-planning and obstacle avoidance algorithms for UAVs: a forward neural network deep learning control with hybrid comparative study,” Unmanned Systems, vol. 6, no. 12, position and virtual-force algorithm for mobile robot obstacle pp. 95–118, 2018. avoidance[J],” International Journal of Control, Automation [4] A. Nasrinahar and J. H. Chuah, “Intelligent motion planning and Systems, vol. 17, no. 8, pp. 1007–1018, 2019. of a mobile robot with dynamic obstacle avoidance,” Journal [20] D. Q. Bao and I. Zelinka, “Obstacle avoidance for swarm robot on Vehicle Routing Algorithms, vol. 1, no. 2, pp. 89–104, 2018. based on self-organizing migrating algorithm,” Procedia [5] A. Adib and B. Masoumi, “Mobile robots navigation in un- Computer Science, vol. 150, no. 22, pp. 425–432, 2019. known environments by using fuzzy logic and learning [21] A. K. Rath, D. R. Parhi, H. C. Das, M. K. Muni, and automata,” Artificial Intelligence and Robotics, vol. 56, no. 15, P. B. Kumar, “Analysis and use of fuzzy intelligent technique pp. 111–123, 2017. for navigation of humanoid robot in obstacle prone zone,” [6] Q. Wu, H. Lin, Y. Jin, Z. Chen, S. Li, and D. Chen, “A new Defence Technology, vol. 14, no. 6, pp. 677–682, 2018. fallback beetle antennae search algorithm for path planning of [22] D. C. Rao, M. R. Kabat, and P. K. Das, “Hybrid IWD-DE: a mobile robots with collision-free capability,” Soft Computing, novel approach to model cooperative navigation planning for vol. 24, no. 3, pp. 2369–2380, 2020. multi-robot in unknown dynamic environment,” Journal of [7] L. Zhang, J. Wang, and Z. Lin, “Distributed cooperative Bionics Engineering, vol. 16, no. 10, pp. 235–252, 2019. obstacle avoidance for mobile robots using independent virtual centre points,” Journal of Intelligent and Robotic Systems, vol. 98, no. 9, pp. 791–805, 2020. [8] J. Y. Jhang, C. J. Lin, and C. T. Lin, “Navigation control of mobile robots using an interval type-2 fuzzy controller based on dynamic-group particle swarm optimization,” Interna- tional Journal of Control, Automation and Systems, vol. 16, no. 10, pp. 2446–2457, 2018. [9] F. Shamsfakhr and B. Sadeghibigham, “A neural network approach to navigation of a mobile robot and obstacle avoidance in dynamic and unknown environments,” Turkish Journal of Electrical Engineering and Computer Sciences, vol. 25, no. 23, pp. 1629–1642, 2017. [10] S. Muthukumaran and R. Sivaramakrishnan, “Optimal path planning for an autonomous mobile robot using dragonfly algorithm,” International Journal of Simulation Modelling, vol. 18, no. 4, pp. 397–407, 2019. [11] R. Zhao and H. K. Lee, “Fuzzy-based path planning for multiple mobile robots in unknown dynamic environment,” http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png Journal of Robotics Hindawi Publishing Corporation

Intelligent Obstacle Avoidance Algorithm for Mobile Robots in Uncertain Environment

Guan, Liwei; Lu, Yu; He, Zhijie; Chen, Xi
Journal of Robotics , Volume 2022 – Mar 30, 2022
Download PDF
9 pages

Loading next page...
 
/lp/hindawi-publishing-corporation/intelligent-obstacle-avoidance-algorithm-for-mobile-robots-in-kAwXXjaL3f

References (23)

Publisher
Hindawi Publishing Corporation
Copyright
Copyright © 2022 Liwei Guan et al. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
ISSN
1687-9600
eISSN
1687-9619
DOI
10.1155/2022/8954060
Publisher site
See Article on Publisher Site

Abstract

Hindawi Journal of Robotics Volume 2022, Article ID 8954060, 9 pages https://doi.org/10.1155/2022/8954060 Research Article Intelligent Obstacle Avoidance Algorithm for Mobile Robots in Uncertain Environment 1,2,3 1,2,3 1,2,3 1,2,3 Liwei Guan , Yu Lu, Zhijie He, and Xi Chen College of Physics and Energy, Fujian Normal University, Fujian Provincial Key Laboratory of Quantum Manipulation and New Energy Materials, Fuzhou, Fujian 350117, China Fujian Provincial Collaborative Innovation Center for Optoelectronic Semiconductors and Efficient Devices, Xiamen, Fujian 361005, China Fujian Provincial Engineering Technology Research Center of Solar Energy Conversion and Energy Storage, Fuzhou, Fujian 350117, China Correspondence should be addressed to Liwei Guan; guanlw@fjnu.edu.cn Received 30 January 2022; Revised 7 March 2022; Accepted 9 March 2022; Published 30 March 2022 Academic Editor: Shan Zhong Copyright © 2022 Liwei Guan et al. +is is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited. +e application of mobile robots and artificial intelligence technology has shown great application prospects in many fields. +e ability of intelligent obstacle avoidance is the basis for the deep application of mobile robots. However, there are often more or less uncertain factors in the actual operating environment of the robot, such as people or objects that are not updated in time or temporarily appear. +erefore, it is an important step to complete the automatic learning of obstacle avoidance for mobile robots. In a nondeterministic environment, a mobile robot intelligent obstacle avoidance algorithm based on an improved fuzzy neural network with self-learning is firstly proposed. +e mobile robot intelligent obstacle avoidance system is constructed through the reaction layer, the deliberation layer, and the supervision layer. +rough the analysis of sensor performance, model accuracy, path obstacle avoidance optimization, and obstacle avoidance simulation, the following conclusions are drawn. First, through network training, the accuracy rate of the test set is stable at 98%, and the loss of the function value has also been reduced from the original 0.79 to 0.08, which is 10 times smaller. Second, the traditional single sensor cannot meet the obstacle avoidance requirements of robots, and mobile robots must combine multipurpose technology. +ird, the algorithm in this paper encounters the following. When there are obstacles, the path is dominated by straight lines, obstacle avoidance planning is optimal, and the distance is shorter. Fourth, the larger N : M, the larger the solution space, indicating that this algorithm gradually improves the search efficiency to the greatest extent and can handle any form of medium and large scale task allocation problem. On the basis of the continuous improvement of artificial 1. Introduction intelligence, the development of robot technology is be- Robot technology is an important manifestation and symbol coming more and more intelligent, and the ability of mobile of the degree of industrial automation and the national high- robots to intelligently avoid obstacles is an important in- dicator of their intelligence [2]. +ey not only reflect the tech level. Robot technology is a high-tech formed on the basis of interdisciplinary. It is one of the hot spots of current efficiency, feasibility and energy consumption of mobile scientific research [1]. Mobile robots are widely used in robot motion but also reflect the way the robot detects various fields, especially in harsh and dangerous environ- obstacles, processes obstacle information, and avoids ob- ments such as military reconnaissance, antinuclear pollu- stacles. Today’s mobile robots have moved from structured tion, mine clearance, and material handling in civil use. work spaces to uncertain environments [3]. In an unknown +erefore, mobile robots greatly facilitate people’s pro- and uncertain environment, it is an important step for duction and life. +e research on robots has received mobile robots to learn to avoid obstacles autonomously. It widespread attention. enables mobile robots to have human-like behaviour 2 Journal of Robotics strategies and avoid obstacles, thereby realizing intelligent out that based on global path planning, navigation and navigation of mobile robots [4]. How to intelligently avoid obstacle avoidance are performed according to the planned path method [13]. Jin J. combined fuzzy logic and grid graph these obstacles unknown to the environment map in an uncertain environment is a key link of mobile robots and one in the target navigation of mobile robots, using the mini- of the research hot spots of robotics. mum risk criterion as the evaluation function to improve the At present, great progress has been made in the research effect of path planning [14]. Zafar M. proposed a learning of autonomous obstacle avoidance algorithms for mobile algorithm based on neural network error backpropagation robots in the world, and many algorithms are applied in the to adjust the membership function parameters of the fuzzy autonomous obstacle avoidance systems of mobile robots. logic system to improve the trajectory smoothness of the +ese intelligent obstacle avoidance algorithms mainly in- mobile robot [15]. Cui Min applied the improved neural clude fuzzy control, grid map, artificial potential field, and network algorithm to the path planning of mobile robots, neural network [5]. +e fuzzy control algorithm has the which improved the operation efficiency of the algorithm advantages of simple algorithm, easy to understand, and [16]. Chen D. proposed a reinforcement learning mecha- nism based on fuzzy neural network, which utilizes the strong robustness, but there are also problems such as the need for experience in design, low control accuracy, and no residual in the learning algorithm and obtains a better al- learning ability; neural network algorithms can use a large gorithm convergence speed in the navigation process of the amount of data to train models, which can automatically mobile robot [17]. learn parameters to obtain the final end-to-end network In the previously mentioned research, different algo- model, but there are problems of high network complexity rithms are used in mobile robot path planning and navi- and practicability. +is paper fully considers the practica- gation and obstacle avoidance, and the performance of the bility and operability of intelligent obstacle avoidance of system has been improved to a certain extent. However, in mobile robots, combines fuzzy control algorithm with the nondeterministic environment, the previously men- tioned obstacle avoidance and planning algorithms have neural network algorithm, and takes the lead in proposing an intelligent obstacle avoidance algorithm for mobile robots some deficiencies in flexibility, real-time, humanization, and intelligent performance. +is paper focuses on the problems based on improved fuzzy neural network with self-learning. existing in the previously mentioned algorithms. In a nondeterministic environment, an intelligent obstacle 2. Related Work avoidance algorithm for mobile robots based on an im- A mobile robot is a robot system that completes certain work proved fuzzy neural network with self-learning is firstly functions. It can sense the external environment and its own proposed, which improves the flexibility, autonomy, and situation through sensors and realize its autonomous stability of mobile robots. movement in an environment with obstacles [6]. Mobile robots have become a research hot spot in robotics because 3. Implementation of Intelligent Obstacle they have shown more and more extensive application Avoidance Algorithm prospects in various aspects of agriculture, industry, aero- space, medicine, and human life [7]. +is paper fully considers the importance of obstacle +e earliest research on mobile robots was from 1966 to avoidance for mobile robots in nondeterministic environ- 1972, when Nils Nilssen and Charles Rosen of Stanford ments and proposes an intelligent obstacle avoidance al- Research Institute developed an autonomous mobile robot gorithm for mobile robots based on improved fuzzy neural [8]. In the 1980s, a number of universities such as Stanford network autonomous learning. First, in view of the high and MIT established a special scientific research team for complexity and poor real-time performance of the fuzzy ALV research [9]. In the 1990s, Japan developed working algorithm, the necessary simplification of the fuzzy neural robots that can operate in extreme environments and suc- network is carried out to perform image preprocessing and cessfully developed humanoid robots [10]. In 2003, Tsinghua rough positioning of obstacles. Second, the parameters of the University in China independently developed the robot fuzzy membership function are automatically and dynam- automatic navigation system [11]. All over the world, Japan ically adjusted by the neural network with self-learning has always been at the forefront of the world in the field of ability, so that the fuzzy control rules have stronger object humanoid robots and wheeled robots. +e ASIMO biped adaptability. Furthermore, for dynamic obstacles in au- walking robot developed by Honda represents the highest tonomous obstacle avoidance, edges are introduced. +e level in the world. ASIMO uses a variety of sensors in the detection operator implements the dynamic obstacle realization of its functions, including cameras on the head, avoidance strategy of the mobile robot. ultrasonic sensors around, and ground pressure sensors on the bottom of the feet. In terms of functional realization, ASIMO not only realizes the functions of walking, posi- 3.1. Improved Fuzzy Neural Network Algorithm. +is paper tioning, and navigation but also adds functions such as face applies the improved fuzzy neural network algorithm to the recognition and communication by voice or gesture [12]. field of image processing, analyzes the image to detect ob- In recent years, breakthroughs have been made in the stacles by counting the neurons in the visible range, and uses research on the theory and algorithm of navigation control the knowledge of distance images to convert the detection of mobile robots in unknown environments. Yan Z. pointed results to finally determine the location of the obstacles. Journal of Robotics 3 3.1.1. Image Preprocessing and Coarse Positioning of Among them, D(X) represents the unit neuron of the Obstacles. Image preprocessing includes image greyscale, sample space X. Various colour components of an image can image denouncing, and effective image region selection. be used as the basic measure of the amount of information, Image greyscale is to complete the conversion from colour to and grayscale representation is also a method of expressing greyscale. In the system, the video image format conversion information. Each grayscale level is the basic element of the is completed through the acquisition of the camera. +e composition. +e probability of occurrence is used to value of the three colour components of R, G, and B in the characterize the amount of information in the area. RGB space of the image is stored in the image, and greyscale Use the knowledge of neurons to analyze the image and can be achieved by using locate obstacles. +e unit information entropy expression formula of the image is applied to a certain frame of gray � 0.5 × Red + 0.23 × Green + 0.27 × Blue. (1) greyscale image and can be written as +e size of the redundancy is related to the probability of 255 M � + 􏽘 kqi + log qi. (4) occurrence of each basic element in the information. Its i�0 expression is as follows: Among them, qi represents the probability of the oc- M(X) � E[log q(ai) × 10] � 􏽘 qi(ai)log qi(ai). (2) currence of a pixel whose gray level is i; k is the number of i�0 gray levels whose qi is not zero in all gray levels. In the actual experiment, the horizontal image analysis process is as Among them, ai is the event in the sample population follows: select a row of pixels as the data source for statistics X. +e previously mentioned formula indicates that the qi; that is, count the gray level distribution of the pixels in more average the number of occurrences of each sample this row, and obtain the corresponding pixel points in each event, the greater the number of different events in the gray level i. +e number of pixels, qi, is the ratio of the sample population X, and the greater the amount of in- number of pixels contained in the gray level to the total formation. If there are N groups of samples, X , X . . .X , the 1 2 n number of pixels in the row; then, the unit information corresponding sample numbers are k ,k . . .k , and the in- 1 2 n entropy is calculated for all gray levels whose qi is not 0. formation entropy is M ,M . . .M . +e unit information 1 2 n entropy is the average value of the information contained in the unit sample in the sample population, which is embodied 3.1.2. Accurate Positioning of Obstacles in the Image. as the ratio of the population neuron to the number of spatial Since the experimental environment is relatively simple, the samples, namely, relatively simple Roberts edge detection operator can be used for detection to obtain edge point information, which D(X) � M(X)∗ n � 􏽘 qi(ai)log qi(ai) × . (3) can effectively improve the real-time performance of the i�0 system. +e Roberts operator in an image pixel array looks like this 􏽱������������������������������������������� 2 2 (5) k(a, b) � [t(a, b) + t(a + 1, b + 1)] − [t(a, b + 1) + t(a + 1, b)] . α � arctan h × y1, (7) +e Roberts operator is a diagonal derivative computed in a 2×2 neighbourhood. In practical applications, the β � arctan h × (y1 − y2), (8) previously mentioned formula can be replaced by a sim- plified calculation form, which is expressed as follows; that λ � arctan(x1 × y1), (9) is, the absolute value of Roberts is replaced. where h, y1, y2, and x1 are data that can be measured. After k(a, b) � |t(a, b) + t(a + 1, b + 1)| + |t(a, b + 1) + t(a + 1, b)|. (6) obtaining α, β, and c, x and y can be obtained from the trigonometric function relationship, and the derivation formula is as follows: 3.1.3. Distance Image Obstacle Location. +e range image y � h × tan􏽨(45 + α) − 􏼐s − u􏼑s × (α + β)􏽩, y y stores the depth information of the ray associated with each pixel and the first focus of the scene observed by the camera. x � y × tan􏼂s + v􏼃 × s + c, (10) x x According to the idea of interpolation, using the geometric 􏽱������ 􏽱������ 2 2 2 2 principle of camera imaging, using Manlius’s theorem, es- B � x − y × x + y . tablish a three-dimensional distance function, calculate the Among them, B is the distance between the vertical distance from any position in the monitoring area to the projection points of the camera coordinates; S and S are the camera, and establish a distance image about the distance x y total number of pixels in the x and y directions in the image between any point in the monitoring area and the camera: 4 Journal of Robotics platform of the ideal hybrid architecture. In this chapter, we plane; u and v represent the horizontal and vertical imaging planes, respectively. are based on the mobile robot intelligent obstacle avoidance architecture, follow the general design principles of intel- +e position of the target point in ABCD can be obtained by formulas (7)–(9), that is, the coordinates of the target ligent system structure, and add supervision management point in the world coordinate system. By analyzing the image and learning functions on the basis of the original hybrid through image neurons and edge information in the image, structure. +rough the improved fuzzy neural network al- the characteristic edge points of obstacles can be obtained. gorithm, the robot can learn behavior autonomously in the +rough the knowledge in this section, these correspon- dynamic unknown environment so as to adapt to the dences with the world coordinate system are established to changes of the environment, as shown in Figure 2. obtain the position of the obstacle in the actual situation. Reactive behaviour control layer contains reflective behaviour and reactive behaviour, as well as a structure for coordinating reactive behaviour. Reflective behaviour is 3.2. Intelligent Obstacle Avoidance System. A very important used to react to emergencies in the environment, or it can be task area for mobile robots is navigation, which is to find a a single control rule. For the navigation tasks of mobile route so that it can move safely without collisions around all robots, reactive behaviors are designed for obstacle avoid- obstacles. +is planning technology that can autonomously ance, goal orientation, and roaming. +ese behaviors can be avoid obstacles and complete tasks is the frontier technology pare-designed by expert experience or acquired through for intelligent mobile robots to achieve autonomous be- learning and evolution. All behavior can directly respond to haviour control. +erefore, it is the most basic problem to the information felt from the outside world and can also apply the improved fuzzy neural network algorithm to the receive control signals from the deliberation layer and the intelligent obstacle avoidance system of mobile robots. +e supervision layer to perform actions. mobile robot intelligent obstacle avoidance system designed Regarding the deliberate behaviour control layer, the in this paper integrates different functions in the system, as task for the navigation class includes five basic modules, and shown in Figure 1. other functions can be added on this basis. +e modules are Much of the structure of the intelligent obstacle environment model knowledge base, task planning module, avoidance system involves planning, using a reactive planner positioning module, navigation module, and sequencer. +e called PRS-L. PRS-L can accept human natural language environmental model knowledge base can directly receive instructions, and then, start to run navigation tasks and the environmental model transmitted from the outside perceptual recognition routines [18]. Both planning and world and store it in the knowledge base through human- execution rely on a locally aware spatial-centrist model of the computer interaction and can also generate a map based on environment. +e reactive components of an intelligent the collected sensor information, maintain the map infor- architecture consist of behavior. +ese behavior extract mation in time, and provide orientation data to other virtual sensor input and output fuzzy rules from the local modules. +e task planning module receives the instructions perception space of the central environment model and then input by the user and transmits it to the robot for task synthesize control instructions through fuzzy logic. planning. +e positioning module uses the shaft angle en- From the intelligent obstacle avoidance system, it can be coder combined with the feature quantity extracted from the seen that a hybrid architecture should have the following external sensor data to determine the position of the robot in modules and objects. the environment at each moment. Navigation accepts the Sequencer agent is used to generate the set of tasks task from the task planning module and calculates the path required to complete the subtasks and to determine all and decomposes it into subtasks. +e sequencer is used to timing and activation conditions. Timing is usually repre- generate the task set required to complete the subtasks and sented as a correlation network or finite state machine, and determine all the timing and activation conditions. When sequencers can generate these structures and modify them the purposeful planning is required, the reaction layer ob- dynamically [19]. tains the ordered actions through the sequencer. Resource manager is used to allocate resources for be- Supervise and manage the behaviour control layer: havior including selection of schema libraries. used to monitor the execution of the deliberation layer Cartographer is used to generate, store, and maintain and the behaviour layer, so that the robot can notice map or spatial information and provide a means of accessing whether it is making progress. At present, five modules are data. set up, namely, a cross module for coordinating and Mission planner interacts with the user, passes in- reflecting deliberation behavior, a fault supervision structions to the robot, and generates mission plans. module, a supervision module for the execution of be- Performance supervision and problem solving agents are havior at the reaction layer, a supervision module for used to let the robot notice if it is making progress. behaviour planning at the deliberation layer, and a learning evolution unit. +e learning evolution unit has 3.3. Intelligent System Hierarchy. Because the mobile robot two functions. One is to learn the local optimal path planning behaviour in a static environment of delibera- obstacle avoidance system adopts the modular design method based on mufti-agent, the functions are independent tion behaviour, and obtain the design of reactive be- of each other. +erefore, it is very easy to realize the function haviour autonomously. +e second is to use the collected expansion and can be used as the design and experiment sample data for training and learning in a dynamic Supervisory layer Deliberate layer Reaction control layer Actuator Navigation task Topology planner Environmental Model Environmental classifier Knowledge Journal of Robotics 5 Resource manager Mission planner Cartographer Smart logic Software agent Performance Detection Agent People tracking Planning agent (PRS-L) Object recognition Local perception space Surface build Reactive behavior Sensor Orientation map maintenance Sequencer virtual sensor Figure 1: Intelligent obstacle avoidance system diagram. Prior knowledge base HMI Task set Learning/Evolution React layer Deliberate layer Fault monitoring Unit monitoring monitoring Mission planning Task sequencer Map maintenance Map main i i i ten Process monitoring Navigation Posi Pi Position i iti tion i i Path sequencer module Rout Rt Rt R RRt R R R R Route plan t t t teplan Avoidance Behavioral Tracking fusion Pass through Reactive behavior Along the wall ON/OFF Towards the goal Task fusion Rotation Reflective behavior Kinetics/ Emergency stop Behavior Kinematics controller Other behavior Figure 2: Mobile robot hierarchy diagram. environment to establish a prediction model for dynamic 3.4. Obstacle Avoidance Control Process. In robot route obstacle avoidance. Learning the functions of evolution planning, various information obtained from the robot itself could enable robots to autonomously adapt to environ- and its environment is synthesized, enabling the robot to mental changes and enhance their intelligence. understand its environment and make decisions through 6 Journal of Robotics controller processing. So as to avoid obstacles, find the gradient learning. +is is a supervised learning process. optimal path and move autonomously [20]. Figure 3 shows Before adding the 1∗ 1 constitutional layer, after multiple the flowchart of the mobile robot path planning. straining, the final test set accuracy rate was stable at about +e mobile robot is equipped with a positioning system 95%. After adding the 1∗ 1 constitutional layer, the network that can detect the global position and heading of the mobile is retrained, the test set accuracy rate is stable at 98%, and the robot. Six ultrasonic sensors are used to detect local obstacle loss function value is also reduced from the original 0.79 to information. +e mobile robot performs detection every 1s, the current 0.08, which is 10 times smaller. Since the added and the acquired sensor data is fused as the input of the 1∗ 1 constitutional layer contains nonlinear units, the controller. By verifying the effectiveness of the previously nonlinear expression ability of the model is improved. After mentioned improved fuzzy neural network algorithm, a adding 1∗ 1 to two different curves, the fluctuation is ob- system simulation model is established in Mat lab with viously reduced compared with the previous curve, and it is Simulation, the control rules are simulated, and the fuzzy relatively stable. logic toolbox software is used to simulate the neural network algorithm. In this simulation system, the steps of path 4.2. Sensor Property Index Analysis. +e sensor control planning simulation are as follows. system and the information processing system constitute the (1) Establish environmental information: environmen- robot’s obstacle avoidance system. Among them, the sensor tal information is to establish the coordinates, di- control system is composed of sensors and microprocessors mensions of obstacles, and the starting point and that obtain environmental information. It mainly collects target point of the robot. information from unknown environments, which is the only way for robots to understand environmental information. (2) Establish a simulated robot, including some pa- +erefore, the reasonable configuration of the sensor de- rameters such as robot size and moving speed. termines the accuracy of the system’s acquisition of the According to the kinematic model of the robot in the external environment. actual system, it is assumed that the travelling speed ° As can be seen from Figure 5, the radar has the best of the mobile robot is 0.6 m/s, and it can realize 360 performance index and high cost. Ultrasonic and infrared in situ. have the lowest cost. In view of the need for obstacle (3) Establish a simulated sensor: used to perceive the avoidance, combined with the cost of the sensor, the simulated environment information, that is, to ob- performance index of the collected information, the tain the value of the obstacle distance information d hardware implementation circuit, volume, and other and the target direction angle ?. comprehensive factors, the sensor control system in this (4) +e robot controller is designed by an improved paper mainly uses ultrasonic and infrared sensors to collect fuzzy neural network algorithm, and the driving the obstacle distance of the external environment and instructions of the robot are obtained by analyzing electronic compass to obtain the target object. Considering the obstacle and target data obtained by the sensor. the unknowns and complexity of the external environment, (5) Transmission of control instructions: the driving the multipurpose technology is applied to the mobile robot obstacle avoidance system, and the hardware and algo- instructions are transmitted to the simulated robot, the mobile robot moves according to the instruc- rithms are processed accordingly to enhance the intelli- gence of the robot. tions, and then the map coordinates of the robot after executing the driving instructions are calculated; repeat steps (3) to (5) until the robot reaches the 4.3. Path Obstacle Avoidance Optimization Analysis. In a predetermined target point. complex environment with obstacles, the ultimate goal of path planning is to solve an optimal route, so that the robot can move smoothly and avoid all obstacles without collision 4. Intelligent Obstacle Avoidance Algorithm [21]. In practical applications, the environment is unknown Results and Analysis to the mobile robot. +e unknown environment, the de- 4.1. Algorithm Accuracy Analysis. In this experiment, the tection accuracy of the obstacle detector and the difference of two network models before and after the improvement were the path planning algorithm have a great influence on used for training, respectively, and the training results were whether the path planning can be successfully implemented. compared after 300 times of effective training. Before +is paper compares and analyzes the experimental paths of training, you need to import the saved network model data the BUG algorithm and the improved fuzzy neural network in advance, and use the training data set to continue training algorithm. the network model before the improvement. At the same It can be seen from Figure 6 that when the BUG algo- time, the loss value and the accuracy curve are obtained, and rithm is used, the robot moves along the edge contour of the the accuracy curve of the test set is obtained, as shown in obstacle. When it can move to the target point, it will directly Figure 4. leave the edge of the obstacle and move beyond the target It can be seen from Figure 4 that since the training, the point. It is not limited to the distance judgment between the loss value has been continuously reduced and the accuracy obstacle and the target point. H1 and H2 are the arrival has been continuously improved in the process of reverse points, and L1 and L2 are the separation points. When using Journal of Robotics 7 Accept ambient mode Start Determine the minimum distance Initialization Obstacle Type Robot state selection Capture the location of e location of the obstacles acquisition target Determine the minimum Are there any obstacles around? With wall type distance Determine the Into the dead zone minimum distance Output speed U-turn in place Target-oriented behavior Figure 3: Flowchart of intelligent obstacle avoidance path. 100 Before improvement Long distance Aer improvement 80 5 Target Dust 0 20 40 60 80 100 120 140 160 Cost 0 Exclude Before improvement Aer improvement Temperature All day 0 20 40 60 80 100 120 140 160 Night Figure 4: Model loss value and accuracy curve. Ultrasound Video Infrared Radar Laser the intelligent algorithm, the robot initially moves along the Figure 5: Sensor performance graph. line connecting the starting point and the target point. When encountering an obstacle, it moves along the tangential direction of the obstacle until no obstacle is detected and then updates the current point to the target point. It extends 4.4. Obstacle Avoidance Algorithm Simulation. In order to the updated main line and repeat this process until the target verify the performance of the algorithm, a lot of simulation point is reached. After comparison, it is found that the research has been done in this paper. Let the values of N and improved fuzzy neural network algorithm in this paper, the M be 4, 10, 30, and 60, respectively, the selection probability path distance is dominated by straight lines, the distance is is 0.1, the cross-mutation probability is 0.08, the a and ß shorter, and the distance is shortened by 5–10m. values are both 1, and the population evolution stops at 200 Loss rate Accuracy 8 Journal of Robotics 0 5 10 15 20 25 0 5 10 15 20 25 H1 H1 L1 L1 L2 H2 L2 H2 Distance (m) Distance (m) Figure 6: Comparison and analysis diagram of path obstacle avoidance optimization. 0.75 0.80 0.60 0.64 0.45 0.48 0.30 0.32 0.15 0.16 0 10 20 30 40 50 60 0 10 20 30 40 50 60 First group (time /ms) Second Group (time /ms) Figure 7: Obstacle avoidance algorithm simulation curve. or 1000 generations. Research the convergence performance 5. Conclusion of the algorithm evolution under the same parameters: the evolution algebra when the algorithm converges and the +e ability to autonomously avoid obstacles is the main execution time is consumed by the evolution calculation. indicator to measure the intelligence of mobile robots, and it +e changes of the maximum fitness value during the is also an important condition for intelligent robots to drive evolution of each group of experiments are shown in safely [22]. +is paper is the first to propose an intelligent Figure 7. obstacle avoidance algorithm for mobile robots based on It can be seen from the figure that, with the increase of improved fuzzy neural network, which achieves precise N : M, the method will gradually improve the search effi- positioning through adaptive learning. Following the general ciency. +e larger N : M is, the larger the solution space is, principles of intelligent control system design, on the basis of but the convergence speed does not change, which shows response and deliberation, a monitoring layer is added. that the search efficiency of this method is the largest as the Supervise and coordinate the implementation of deliberative value increases. When the solution space of the problem is layer behaviors to learn adaptive behaviors in unknown large, the speed of convergence is obviously slowed down, environments. By collecting sample data for training, it the algebra of convergence is improved, and the convergence learns to build predictive models to avoid dynamic obstacles. time is prolonged. +is further verifies that the method is +rough the obstacle avoidance experiment simulation, the suitable for medium and large scale problem spaces. In accuracy and real-time performance of the system are summary, the algorithm can handle any form of medium-to- further verified, and finally, a good obstacle avoidance effect large-scale task assignment problem. is achieved. Journal of Robotics 9 Journal of Electrical Engineering and Technology, vol. 12, Data Availability no. 11, pp. 918–925, 2017. [12] G. Lee and D. Chwa, “Decentralized behaviour-based for- +e data used to support the findings of this study are mation control of multiple robots considering obstacle available from the corresponding author upon request. avoidance,” Intelligent Service Robotics, vol. 11, no. 9, pp. 127–138, 2018. Conflicts of Interest [13] Z. Yan, J. Li, and G. Zhang, “A real-time reaction obstacle avoidance algorithm for autonomous underwater vehicles in +e authors declare that there are no conflicts of interest or unknown environments,” Sensors, vol. 18, no. 12, pp. 438–448, ethics in this article. 2018. [14] J. Jin and W. Chung, “Obstacle avoidance of two-wheel differential robots considering the uncertainty of robot mo- Acknowledgments tion on the basis of encoder odometer information,” Sensors, vol. 19, no. 17, pp. 289–376, 2019. +is study was supported by the Key Project of Industry [15] M. N. Zafar and J. C. Mohanta, “Methodology for path Guidance of Science and Technology Department of Fujian planning and optimization of mobile robots: a review,” Province“R and Dand application of large space automatic Procedia Computer Science, vol. 133, no. 32, pp. 141–152, 2018. tracking and positioningJet Fire Control System for Intel- [16] M. Cui, H. Liu, and W. Liu, “An adaptive unscented kalman ligent Security”(no.2020Y0021). filter-based controller for simultaneous obstacle avoidance and tracking of wheeled mobile robots with unknown slipping parameters,” Journal of Intelligent and Robotic Systems, References vol. 92, no. 46, pp. 489–504, 2018. [17] D. Chen, “Fuzzy obstacle avoidance optimization of soccer [1] S. M. H. Rostami, A. K. Sangaiah, and J. Wang, “Obstacle robot based on an improved genetic algorithm,” Journal of avoidance of mobile robots using modified artificial potential Ambient Intelligence and Humanized Computing, vol. 11, field algorithm,” EURASIP Journal on Wireless Communi- no. 12, pp. 6187–6198, 2020. cations and Networking, vol. 22, no. 4, pp. 1–19, 2019. [18] Z. Yi, G. Li, and S. Chen, “A navigation method for mobile [2] B. K. Patle, A. Pandey, A. Jagadeesh, and D. R. Parhi, “Path robots using interval type-2 fuzzy neural network fitting planning in uncertain environment by using firefly algo- Q-learning in unknown environments,” Journal of Intelligent rithm,” Defence technology, vol. 14, no. 6, pp. 691–701, 2018. and Fuzzy Systems, vol. 37, no. 35, pp. 1113–1121, 2019. [3] M. Radmanesh, M. Kumar, and P. H. Guentert, “Overview of [19] W. Zheng, H. B. Wang, and Z. M. Zhang, “Multi-layer feed- path-planning and obstacle avoidance algorithms for UAVs: a forward neural network deep learning control with hybrid comparative study,” Unmanned Systems, vol. 6, no. 12, position and virtual-force algorithm for mobile robot obstacle pp. 95–118, 2018. avoidance[J],” International Journal of Control, Automation [4] A. Nasrinahar and J. H. Chuah, “Intelligent motion planning and Systems, vol. 17, no. 8, pp. 1007–1018, 2019. of a mobile robot with dynamic obstacle avoidance,” Journal [20] D. Q. Bao and I. Zelinka, “Obstacle avoidance for swarm robot on Vehicle Routing Algorithms, vol. 1, no. 2, pp. 89–104, 2018. based on self-organizing migrating algorithm,” Procedia [5] A. Adib and B. Masoumi, “Mobile robots navigation in un- Computer Science, vol. 150, no. 22, pp. 425–432, 2019. known environments by using fuzzy logic and learning [21] A. K. Rath, D. R. Parhi, H. C. Das, M. K. Muni, and automata,” Artificial Intelligence and Robotics, vol. 56, no. 15, P. B. Kumar, “Analysis and use of fuzzy intelligent technique pp. 111–123, 2017. for navigation of humanoid robot in obstacle prone zone,” [6] Q. Wu, H. Lin, Y. Jin, Z. Chen, S. Li, and D. Chen, “A new Defence Technology, vol. 14, no. 6, pp. 677–682, 2018. fallback beetle antennae search algorithm for path planning of [22] D. C. Rao, M. R. Kabat, and P. K. Das, “Hybrid IWD-DE: a mobile robots with collision-free capability,” Soft Computing, novel approach to model cooperative navigation planning for vol. 24, no. 3, pp. 2369–2380, 2020. multi-robot in unknown dynamic environment,” Journal of [7] L. Zhang, J. Wang, and Z. Lin, “Distributed cooperative Bionics Engineering, vol. 16, no. 10, pp. 235–252, 2019. obstacle avoidance for mobile robots using independent virtual centre points,” Journal of Intelligent and Robotic Systems, vol. 98, no. 9, pp. 791–805, 2020. [8] J. Y. Jhang, C. J. Lin, and C. T. Lin, “Navigation control of mobile robots using an interval type-2 fuzzy controller based on dynamic-group particle swarm optimization,” Interna- tional Journal of Control, Automation and Systems, vol. 16, no. 10, pp. 2446–2457, 2018. [9] F. Shamsfakhr and B. Sadeghibigham, “A neural network approach to navigation of a mobile robot and obstacle avoidance in dynamic and unknown environments,” Turkish Journal of Electrical Engineering and Computer Sciences, vol. 25, no. 23, pp. 1629–1642, 2017. [10] S. Muthukumaran and R. Sivaramakrishnan, “Optimal path planning for an autonomous mobile robot using dragonfly algorithm,” International Journal of Simulation Modelling, vol. 18, no. 4, pp. 397–407, 2019. [11] R. Zhao and H. K. Lee, “Fuzzy-based path planning for multiple mobile robots in unknown dynamic environment,”

Journal

Journal of Robotics – Hindawi Publishing Corporation

Published: Mar 30, 2022

There are no references for this article.