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

Learn More →

Hierarchical Vehicle Scheduling Research on Tide Bicycle-Sharing Traffic of Autonomous Transportation Systems

Hierarchical Vehicle Scheduling Research on Tide Bicycle-Sharing Traffic of Autonomous... Hindawi Journal of Advanced Transportation Volume 2023, Article ID 5725009, 9 pages https://doi.org/10.1155/2023/5725009 Research Article Hierarchical Vehicle Scheduling Research on Tide Bicycle-Sharing Traffic of Autonomous Transportation Systems 1,2 1,2 1,2 1,2 Mai Hao , Ming Cai , Minghui Fang , and Shuxin Jin School of Intelligent Systems Engineering, Sun Yat-Sen University, Shenzhen 518107, China Guangdong Key Laboratory of Intelligent Transport Systems, Guangzhou 510006, China Correspondence should be addressed to Shuxin Jin; jinshx3@mail.sysu.edu.cn Received 4 July 2022; Revised 7 September 2022; Accepted 7 February 2023; Published 18 February 2023 Academic Editor: Alessandro Severino Copyright © 2023 Mai Hao et al. Tis 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. To facilitate the intelligent and automated provision of mobility services by autonomous transportation systems, bike-sharing can be a supplement to public transport for addressing their point-to-point issue, namely, “last mile” service. However, according to the diferent nature of land use, the uneven spatio-temporal distribution of travel demand can directly lead to difcult access to bikes with high travel costs for users and operating costs for operators. Based on this, this paper analyzes the user behavior patterns within diferent areas by using GeoHash encoding and proposes a hierarchical autonomous vehicle scheduling model based on tide bicycle-sharing trafc, namely, HATB. It minimizes operating costs and maximizes user satisfaction to dynamically optimize scheduling routes and required vehicles within each layered zone. As for extracting historical orders of Mobike in Beijing, an example analysis through the genetic algorithm of HATB is conducted to support efective and efcient scheduling. Compared to existing scheduling methods, HATB has faster convergence and lower time complexity, which improves bike turnaround ef- fciency and practical application ability, thus making it more convenient for the public to travel during peak hours. constant billing, users never consider the capacity of the 1. Introduction parking area, resulting in an uneven spatial distribution of Nowadays, as the modern transportation systems (TSs) bikes, namely, some areas sufer a severe accumulation of develop, problems among mobility services (MSs), e.g., bikes while others are “one bike is hard to fnd” [5]. congestion, route adjustment, user dispersion, and peak- Terefore, scientifc and reasonable scheduling strategies are time conficts have become commonplace [1–3] in terms of required to overcome the imbalance between the supply and the adjustment of urban planning and the year-on-year demand of bikes and improve resource utilization. increase in car ownership. Since the MS are fundamental in Rebalancing and optimizing bicycle-sharing distribution propelling current intelligent TS (ITS) evolving towards constitutes the vehicle routing problem (VRP), and most autonomous TS (ATS), they are also being renovated to current research is based on this theory. For instance, assist the public on a daily basis and explore the advance- Caggiani et al. [6] proposed a decision support system for the ment of ATS. Some theoretical development has revealed reallocation problem by forecasting the demand for spatio- that the creation of shared transport, called bicycle-sharing, temporal bikes. Similar research can, accordingly, be divided has efectively alleviated the “last mile” service (LMS), which into static and dynamic scheduling to optimize VRP models is the most predominant pain point in the current MS [4] in with diferent objectives. Specifcally, in static scheduling, terms of the most immediate interaction with users, and Kadri et al. [7] and Dell’Amico et al. [8] developed opti- some emerging companies, e.g., Mobike have also taken this mization models for user satisfaction and operating cost, trend to a new level. However, bikes must be parked in GPS- respectively. Yan et al. [9] investigated the deterministic and identifed areas to address ill-posed problems in the case of stochastic demand for bicycle-sharing in dynamic sched- illegal parking, vandalism, or theft. Since then, to avoid uling. In general, these methods assume that the overall 2 Journal of Advanced Transportation signifcant overall imbalance. Terefore, it is vital to take supply and demand within a scheduling station are in equilibrium without supporting the mobility of bikes be- efective scheduling strategies to rebalance and optimize the distribution of bikes, thereby addressing difculties in tween zones. Moreover, limited open literature has reported that current research focuses too much on mathematical management and operation. In this context, scheduling modeling, neglecting the analysis of actual demands. bicycle-sharing can be regarded as a heuristic algorithm- Terefore, these methods have encountered three challenges based, e.g., GA, ant colony algorithm (ACO), and vehicle in practice, namely, slow convergence, high time complexity, routing problem (VRP) [18–20], which can be classifed as and problematic application, from perspectives of fulflling static or dynamic scheduling according to diferent strategies the actual demands in real-time and adjusting scheduling as and objectives. needed. Dynamic scheduling mainly focuses on peak time and As current diversifed mobility demands tend to be relies on user demands. For instance, a mathematical model for dynamic scheduling is created by Zhang et al. [21] based managed and fulflled by more intelligent and automated systems with fewer human intervention [10], it is an urgent on the parking area’s actual capacity and users’ predicted arrival times. Shui and Szeto [22] partition peak time to need to collaborate corresponding functions to renovate the conventional MS of ITS in the context of ATS [11], i.e., optimize scheduling routes by regarding scheduling in each update the ability to sense user demands and rearrange time interval as static scheduling. Chiariotti et al. [23] system supplies [12, 13]. Hence, to promote the MS provided propose that scheduling bikes can dynamically determine by ATS, this paper proposes a hierarchical autonomous the scheduling time through historical orders. In general, vehicle scheduling model based on tide bicycle-sharing dynamic scheduling lessens operating costs’ impact on trafc, namely, HATB. Tis model uses GeoHash coding operators by prescheduling bikes to avoid a shortage occurs. However, based on the uncertain use of bikes, frequent to divide the scheduling into three layers, i.e., top, middle, and bottom, corresponding to the scheduling terminus, area, scheduling with complex constraints is necessary, which may lead to higher operating costs and slower convergence, and point. Based on the genetic algorithm (GA), the model can achieve hierarchical and dynamic scheduling of vehicles thus making it challenging to fulfll user demands in real- time. and routes to maximize user satisfaction, while minimizing operating costs. Another more common scheduling strategy is static Furthermore, in contrast to current studies on sched- scheduling during of-peak time. For example, Lang [24] uling bikes in ITS, the HATB makes three main contribu- provides a multiwarehouse model based on the Tabu search tions to optimizing convergence speed, time complexity, and algorithm to minimize scheduling distance and improve application difculties of actual scheduling. Te scheduling scheduling efciency and robustness. Bae and Moon [25] use results based on actual orders ultimately demonstrate HATB a dual time window with customer service levels to reduce total transport and labor costs. Since static scheduling only can provide a rational reference for LMS in ATS and guide the development of bicycle-sharing regulation and considers the predicted demands of stations, increasing more bikes for stations to guarantee user demands means operation. Te overall structure of this paper is divided into fve that the time complexity of the heuristic algorithm grows sections. Section 2 introduces related solutions and exponentially. Moreover, to fulfll the actual demands, the emerging challenges. Te methodology relevant to HATB is allocated bikes by these studies may exceed the station’s described in Section 3. Section 4 elaborates on the sched- capacity. uling results and superiority of the model. Finally, Section 5 Besides, such above-given studies are mainly applied to summarizes this study and sketches future research typical scenarios, as presented in Figure 1, where a single directions. scheduling station serves one zone and only the routes within the scheduling zone are considered. It is often limited in actual scheduling by the service range of the station, which 1.1. Related Works. In the transport domain, LMS refers to needs to frequently adjust the boundaries of this scheduling zone, thus leading to some research on hierarchical the direct interaction between the end of public transport and users, which often sufers from scattered users, peak- scheduling strategies. By defning scheduling priorities based time conficts, and uneven distribution. As an efective way on demand intensity, Sakakibara et al. [26] and Ni et al. [27] to cope with the LMS problems, bicycle-sharing has become highlight the feasibility and reliability of hierarchical a non-negligible component of urban transport. For ex- scheduling. In order to illustrate the fexibility, Zhang [28] ample, Cheng et al. [14] have demonstrated that bicycle- and Ma et al. [29] set stations with similar demands in the sharing increases the proportion of green transport in cities same layer in accordance with the spatio-temporal char- and solves the low efciency at the end of the travel chain. acteristics of bikes. However, the defnition of hierarchies in In general, current research on bike-sharing mainly these methods is too subjective and not clear, making it difcult to implement in practice. focuses on its development status and travel characteristics, but few on its scheduling. Researchers like Soriguera and In summary, whilst a considerable body of research has ´ been carried out on VRP, much less fts the spatial-temporal Jimenez-Meroño [15], Gimon [16], and Lu et al. [17] concur that even while bicycle-sharing has considerable quantities, and cross-regional mobility characteristics of bicycle- the spatio-temporal diferences in user demands, no fxed sharing. In addition, it seems to be a common problem parking area, and fewer available bikes may lead to a more that existing studies focus more on mathematical modeling Journal of Advanced Transportation 3 Point ID | Num. of Bikes Tese geocodes, accordingly, have good accuracy but loss fexibility; they may not meet the actual scheduling requirements. Terefore, GeoHash encoding, proposed by Morton [33], can be used to better support efective and efcient scheduling. Its maximum length of 12 bits can represent Data Collection Map Data a geographic location with arbitrary precision. For example, Parking Point the GeoHash strings WX4ER and WX4G2 represent two regions of Beijing (China), where each character is a certain rectangular area. Moreover, the order information (Data Deploy to Sources: https://biendata.xyz/competition/mobike/) on bicycle-sharing, as extracted in Table 1, also indicates the feasibility of dividing the scheduling layer via GeoHash. Redundancy Scarcity Route Optimization Te coding defnition, as described in Table 1, illustrates that the 7-bit string matches the characteristics of actual bike stops, namely, area size, and the 5-bit string suits for vehicles to dispatch bikes in light of their loading capacity, i.e., 400 bikes. Terefore, the overall framework of the proposed hierarchical scheduling model, called HATB, can be ob- tained as presented in Figure 2. In general, this framework is Scheduling Route characterized by a number of scheduling areas in each of the three layers, namely, bike stops consist of the bottom layer of Figure 1: Te schema of existing scheduling (map data (part of scheduling, while the top and middle layers likewise have Beijing): https://www.beijing.gov.cn). demands and capacity restrictions for bikes, and hence, the scheduling within the same layers is regional scheduling for but neglect the analysis of actual demands. When scheduling seeking optimization. according to the above-given methods, three challenges in terms of the problematic application in practice, namely, the high time complexity of models and slow convergence for 2.2. Model Hypotheses. Considering the complexity of actual algorithms, are increasingly apparent. scheduling, the proposed HATB in this paper defnes the Particularly driven by diverse and emerging technologies following hypotheses and the frequency of scheduling as and demands, ITS is evolving into ATS, which illustrates that once in the morning peak and once in the evening peak, MS should be autonomously fulflled and managed by more respectively. intelligent systems with fewer human intervention [30]. (1) All scheduling vehicles own the same attributes Terefore, it is an urgent need to study and improve the monolithic strategies and rationalize the actual demands to (2) In each scheduling route, the vehicle departs from one scheduling terminus (area) and returns to this achieve hierarchical and autonomous scheduling of bicycle- sharing, thus rationally guiding the provision of the LMS place after deploying bikes to corresponding areas (points) contained by ATS. (3) Fuel consumption and vehicle loss should be considered 2. The HATB Methodology (4) Each scheduling area can only be served once Tis section uses three subsections to present the framework, (5) Te actual orders determine the scheduling demand hypotheses, and construction of the proposed HATB. (6) All scheduling tasks are required to be completed within the specifed scheduling cycle 2.1. Model Framework. As existing scheduling methods, in (7) Te scheduling areas and points have sufcient space general, require frequently adjusting boundaries and in- to accommodate the bikes deployed in or out during creasing vehicles to alleviate the diference between bikes’ a scheduling cycle supply and demand, which may inevitably increase transport time and costs; the scheduling framework can be set up as a hierarchical scheduling structure, i.e., a top-middle- 2.3. Model Construction. Based on the above-given hy- bottom hierarchy of scheduling terminus-area-point. potheses and the actual operations of bike-sharing, con- However, since current hierarchical methods have sidering only the operating costs will gradually lose a subjective defnition of their hierarchies, geocode can customers, and weighing only user satisfaction runs counter ensure the objective; e.g., what3words [31] uses fxed 3m × to the essence of business proftability. Hence combining 3m squares to divide the earth, and pluscode [32] represents these two factors, this paper constructs a regional scheduling model for bikes to minimize operating costs (F each latitude and longitude level by 2-bit code, whose length ) and range in levels 1 to 3 jumps from 110km and 5.5km to 275m. maximize user satisfaction (F ). 2 4 Journal of Advanced Transportation Table 1: Te order information of corresponding GeoHash string length. GeoHash string length 5-bit 6-bit 7-bit Average orders 436.994788 28.214119 3.689289 Area width 4.89 km 1.22 km 153 m Area height 4.89 km 0.61 km 153 m Hierarchical Level Scheduling Top (4-bit) Terminus 39.1 km × 19.5 km 1 N Middle (5-bit) Scheduling 4.89 km × 4.89 km Area 1.1 1.2 1.NN.1 N.2 N.N Bottom (7-bit) Scheduling 153 km × 153 km 1.1.N 1.1.1 1.2.1 N.1.N N.2.1 N.N.N 1.2.N 1.N.1 1.N.N N.1.1 N.2.N N.N.1 Point Figure 2: Te framework of HATB. Te actual operation costs need to consider both fxed and min: zF + μF , 1 2 (1) fexible costs, as summarized in equation (2), which is de- s.t. z + μ � 1 z, μ > 0. termined jointly by the value of scheduling vehicles [34], the unit transport cost (i.e., vehicle loss: 1 CNY/km, fuel con- Te parameters z and μ indicate that bike-sharing op- sumption: 1 CNY/km, and labor cost: 100 CNY/person), and erators need to adjust the weight coefcients of operating the scheduling distance. and penalty costs according to their own emphasis. 2.3.1. F : Te Objective of Minimizing Operating Costs. minF � min(Fx d + Flx) W T I T R R (2) t t ⎝ ⎠ ⎛ ⎞ � min 􏽘 􏽘 􏽘 x C + 􏽘 􏽘 􏽘 x d M , t ij t wj ij w t i t i j I I t t s.t. 􏽘 x � 􏽘 x ≤ 1 ∀w ∈ W, t ∈ T , (3) wi wi i i T R T R t t 􏽘 􏽘 x � 􏽘 􏽘 x � 1 ∀i, j ∈ I, t ∈ T, ji ji t t (4) j�1 j�1 j≠i j≠i R R 0 ≤ 􏽘 􏽘 x d ≤ L ∀t ∈ T, ij ij t (5) j�1 j≠i t t t 0 ≤ a ≤ a x ∀i, j ∈ R, i ≠ j, t ∈ T, (6) ij ij ij x � 0 or 1 ∀i, j ∈ R, i ≠ j, t ∈ T, (7) ij t + a ∈ Z ∀i, j ∈ R, i ≠ j, t ∈ T, (8) ij Journal of Advanced Transportation 5 Table 2: Te meaning of parameters for formulations 1–14. Parameter Meaning I Scheduling area T Scheduling vehicles W Scheduling terminus R Vehicle pools for scheduling terminus and areas a Maximum capacity of vehicle t a Loading bikes of vehicle t from area i to area j ij L Maximum operation distance C Fixed cost per vehicle M Flexible cost per vehicle V Average speed d Transport distance by vehicle from area i to area j ij GAP Number of vehicles to be deployed in/out of area i ω Importance of area i TM Time for vehicle to arrive in area i S Time cost for loading/unloading bikes TM Time for vehicle to depart from terminus [FQ , FR ] Penalty time window i i distance must not exceed the maximum scheduling distance; Table 3: Examples of user travel characteristics. equation (6) points that the number of bikes loaded by the vehicle must not exceed its maximum capacity, namely, 400; Travel information Time Orders equation (7) means x as 0-1 variable; equation (8) proves that ij 7 a.m 189578 the number of bikes deployed by the vehicle is a non-negative Average distance: 815 m 18 p.m 173654 integer. 8 a.m 171011 17 p.m 164126 Median distance: 660 m 19 p.m 125383 2.3.2. F : Te Objective of Maximizing User Satisfaction. 12 p.m 119883 User satisfaction can be improved by adding time window constraints, as described in equation (9), which means maxi- mizing user satisfaction can equivalently transfer into mini- Equation (3) indicates that this scheduling vehicle starts and mizing the penalty cost of scheduling timeout. ends at the terminus; equation (4) suggests that each area can only be served once; equation (5) shows that the transport maxF � min 􏽘 ω ρ t􏼁 , (9) 2 i i i s.t. TM � 0 ∀t ∈ T , (10) T R TM � 􏽘 􏽘 x 􏼐TM + TM + S · GAP 􏼑, ∀j ∈ I, (11) j ij i j i TM + TM + S · GAP − K􏼐1 − x 􏼑 ≤ TM , (12) i j i j ij FQ ≤ TM ≤ FR , ∀i ∈ I. (13) i i i Equation (10) indicates that the vehicle departs from Te meaning of the parameters in the above-given the terminus at time zero; equation (11) means that the equationsare shown in Table 2. time calculation for a vehicle to arrive in an area; 3. Case Study equation (12) demonstrates that the scheduling cannot arrive in area j from area i before TM + TM + S · GAP − i j i Te highlights of HATB solving are illustrated in terms of al- K(1 − x ); equation (13) represents that vehicle needs to gorithm settings, scheduling results, and model evaluation in this ij arrive in an area within the time window. section. 6 Journal of Advanced Transportation Input: SchedulingArea SA, DistanceMatrix DM Output: SchedulingRoute SR, SchedulingVehicle SV, DeployedBikes DR Initialization: Generation Gen � 100, CrossoverRate CR � 0.8, MutationRate MR � 0.2, Population Pop ← ∅, Chromosome CH ← ∅, Fitness Fit � Null, (1) For all the SA do (2) Pop ← CH.generate (SA.Encoding)//Randomly Generation for Population (100) (3) End for (4) Fit ← Fit.Calculation//Calculate the Fitness (5) if Fit.change or Gen < 100 then (6) For all the CH do (7) CH.select//Chromosome Selection (8) CH.crossover//Chromosome Crossover (9) CH.mutate//Chromosome Mutation (10) Pop ← CH.NewGenerate//New Population (11) Fit ← Fit.NewCalculation (12) End for (13) End if (14) SR ← CH.Decoding (15) Return SR, SV, DR ALGORITHM 1: NGA for scheduling optimisation. Area ID | Area Bikes Parking Point Data Collection Routes of Points Route Optimization for Areas Deploy to Redundancy Scarcity Scheduling Terminus Route Optimization for Points Region-by-Region Scheduling Route of Areas Scheduling Areas Scheduling Points Figure 3: Te schema of HATB. 3.1. Algorithm Settings. Te experiment data comes from the diferent vehicles, e.g., a chromosome example might be 0-3- 2017 Mobike cup algorithm challenge, which involves 0-1-2-5-7-0-4-8-6-0, namely, three vehicles serving th rd 3,214,096 orders and 485,465 bikes (10 May 2017–23 May eight zones. 2017). Te characterized user travel, as presented in Table 3, Since the crossover and mutated sub chromosomes may refects that the data are consistent with the “last mile” lead to transport overload and overtime and the time defnition [16], and hence, shows its reasonable usability window is more likely to be violated, the penalty factors for according to prominent tidal characteristics. the two constraints are set to 10 and 500, respectively. Tis paper proposes a GA with natural number encoding (NGA), as defned in Algorithm 1, to optimize the sched- 3.2. Scheduling Results. A total of 220 scheduling areas in the uling. In general, the frst and the last 0 s represent the morning peak are used as HATB test cases to obtain the optimal hierarchical scheduling routes, as shown in Figure 3 scheduling terminus or area, [1, N] represents the zones that need to be scheduled, and other 0 s separate the routes of and Tables 4 and 5. Journal of Advanced Transportation 7 Table 4: Te optimal routes for scheduling areas in the morning peak. Routes Demand Number of vehicles Total distance (km) for scheduling area 0-135-29-113-112-150-205-20-44-170-130-146-0 143 1 0-40-88-138-104-65-12-191-115-52-162-83-157-0 142 1 0-7-45-93-61-129-188-199-158-217-87-53-165-85-119-36-34-210-0 140 1 0-5-6-50-47-41-75-63-86-27-101-125-200-178-81-78-89-122-187-195-203-207-0 140 1 0-18-21-95-132-148-91-71-161-192-117-0 139 1 0-181-106-102-145-135-29-113-112-150-205-20-0 138 1 0-51-177-92-79-84-26-38-214-215-166-168-183-0 133 1 0-31-143-114-37-118-99-156-163-137-149-194-0 114 1 0-54-66-74-128-94-136-201-179-190-152-35-30-33-204-218-0 114 1 719.542007 0-90-78-89-97-64-105-185-131-127-189-140-100-8-82-22-25-98-24-144-124-197-0 113 1 0-3-70-55-169-172-182-109-208-0 113 1 0-62-103-202-43-11-16-57-23-108-153-211-0 111 1 0-4-1-2-14-17-67-48-15-42-56-72-68-99-156-219-0 110 1 0-44-170-30-146- 110 1 159-173-174-69-9-46-49-19-141-186-175-171-39-213-154-51-60-0 0-180-139-142-198-147-116-212-134-167-184-126-216-0 110 1 0-76-164-10-77-73-60-111-120-196-209-0 80 1 0-28-133-110-193-123-154-32-13-107-121-155-206-0 79 1 Table 5: Examples of hierarchical autonomous scheduling results for route one. Routes Scheduling area Total distance (km) for scheduling point a0-a11-a4-a18-a17-a9-a1-a12-a15-a19-a5-a0 170 (a0) 1.443 b0-b1-b3-b11-b4-b48-b39-b18-b42-b38-b7-b19-b54-b2-b28-b25-b26-b27-b43- b31-b12-b21-b45- b52-b30-b20-b46-b47-b6-b29-b40-b37-b14-b33-b9-b34-b51- 130 (b0) 3.905 b24-b32-b15-b55-b10-b44-b22-b16-b41-b17-b5-b13-b23-b53-b8-b50-b49-b36- b35-b0 c0-c17-c23-c13-c15-c29-c32-c10-c35-c31-c2-c34-c5-c4-c14-c33-c22-c11-c9-c3-c6- 146 (c0) 2.951 c21-c20-c25-c30-c28-c24-c16-c27-c19-c8-c12-c26-c1-c18-c7-c0 Table 6: Te mapped scheduling results with examples of route one (Table 5). Routes for scheduling Scheduling area (5-bit) point (7-bit) WX54C (no. 170) WX54C with 0W, 20, E4, 06, 0C, J8, JB, LJ, NC, BW, BX, 0W WX4GW with 00, 02, 08, PL, 06, 77, 7K, P4, NF, 7M, ZH, YY, N8, P0, PF, PY, PV, WX4GW (no. 130) PS, ND, JL, P3, PB, N9, 03, JH, PZ, 7B, 78, 7F, 7E, KH, 7U, 7Q, EQ, ZN, EN, 09, Q8,G7, ZQ, GJ, ZK, NB, P8, P9, P5, 01, 04, 2N, U4, FZ, ZV, 0D, 3K, 86, DK, 00 WX5H2 with 5X, 7C, QX, DM, GL, 6M, Y6, F9, 4N, 4J, DS, DT, VC, TY, DC, D3, WX5H2 (no. 146) TZ, ER, G0, G8, 74, UE, UK, GE, 1E, 12, QK, 7D, 75, 77, W2, EP, G9, ZG, 79, 48, 5X As described in the previous section, regional scheduling is scheduling terminus and returning to the terminus after applied for each layer. Note that, each scheduling area has completing regional scheduling sequentially and autonomously a positive or negative raw demand that refects the redundancy in accordance with the area-point (middle-bottom) hierarchy. In addition, the routes’ information in reality for Table 5 is mapped or scarcity of bikes. Furthermore, scheduling prioritizes self- satisfaction in the route, namely, redundancy supports scarcity, to Table 6 via GeoHash. and hence, route demand indicates the self-satisfaction gap for corresponding regional scheduling. 3.3. Model Evaluation. To further verify the reliability of this Terefore, the experimental results can be summarized as model, Figure 4 shows the comparison for iteration between the total scheduling distance for optimal routes in the entire the HATB and other models proposed in the literature with Beijing is around 719.5 km. As for satisfying the demands, 17 similar objectives. Specifcally, based on the GA, Gao et al. vehicles are required to participate in the scheduling to deploy [35] provided a promising perspective on improving op- bikes. Moreover, using route one in Table 4 as an example, the eration efciency by reducing operating costs and service scheduling can be summarized as a vehicle departing from the quality during peak times to minimize the total operating 8 Journal of Advanced Transportation reduce the time complexity of global optimization. Next, 1.05E+6 a GA for regional scheduling is built by combining the tidal characteristics of bicycle-sharing to minimize operating costs and maximize user satisfaction, which signifcantly 9.00E+5 accelerates the algorithm’s convergence. Last, the use of actual orders considerably enhances the ability in the practical application of instant response to any regional scheduling demand. 7.50E+5 Tis work was carried out as a preliminary to obtain the present results. However, there are still problems such as the inability to adapt scheduling throughout 24 hours or the lack 6.00E+5 of comprehensive constraints. As the closure of this study, one recommendation for further research is to use a form of “GA + Tabu” algorithm to exploit its global search capability 0 20 40 60 80 100 and thus improve the big data processing capability. Another Generation research direction is adding weather and road characteristics The Proposed HATB Gao L et al., 2019 to optimize the model reliability further. Zhao et al., 2022 Panagiotis et al., 2014 Figure 4: Te comparison between diferent scheduling models. Data Availability costs. Angeloudis et al. [36] achieved user appeal increase by Te data used to support the fndings of this study have been deposited in the 2017 Mobike Cup Algorithm Challenge ofering a new method of planning bike routes and distri- butions. Moreover, Zhao et al. [37] optimized the total repository (https://biendata.xyz/competition/mobike/). scheduling distance to accommodate large-scale scheduling via an ACO. Conflicts of Interest th Te proposed HATB converged at the 64 generation, and the total time cost is 148.9 s, with an average running time of 15 s Te authors declare that there are no conficts of interest regarding the publication of this paper. per generation. Due to the diferent objectives, only the con- vergence speed of the above models is compared. It can be seen from Figure 4 that the HATB signifcantly outperforms the Acknowledgments model proposed in the existing literature. Such a result indicates that this model is more proper for practical scheduling appli- Tis research was supported by National Key Research and Development Program of China (Grant no. cations since it optimizes with higher convergence speed and lower time complexity. 2020YFB1600400) and Collaborative Innovation Center for Transportation of Guangzhou (Grant no. 202206010056). 4. Conclusion and Future Works References Even though the emerging and diversifed technologies and [1] S. Moridpour, X. Qu, N. Shiwakoti, and S. Hasan, “Sustainable demands are driving TS to renovate conventional MS to be self- and resilient transport infrastructure,” Journal of Advanced actuating, the current bicycle-sharing scheduling and mainte- Transportation, vol. 2021, Article ID 1576315, 2 pages, 2021. nance rely on manual experience, which lacks scientifc guidance [2] M. L. Mouronte-Lopez, ´ “Analysing the vulnerability of public and efciency. Terefore, to achieve sustainable development, transport networks,” Journal of Advanced Transportation, operators urgently need to develop a rational scheduling strategy vol. 2021, Article ID 5513311, 22 pages, 2021. to balance the distribution conficts and supply user demands [3] N. Tsanakas, J. Ekstrom, ¨ and J. Olstam, “Estimating emissions in time. from static trafc models: problems and solutions,” Journal of Terefore, this paper proposes a hierarchical scheduling Advanced Transportation, vol. 2020, Article ID 5401792, model, called HATB, to address the unsolved issues by 17 pages, 2020. current studies in terms of slow convergence, high time [4] A. Fan, X. Chen, and T. Wan, “How have travelers changed mode choices for frst/last mile trips after the introduction of complexity, and problematic application, and hence, to bicycle-sharing systems: an empirical study in Beijing, China,” support the rational and autonomous provision of LMS. In Journal of Advanced Transportation, vol. 2019, Article ID summary, according to bicycle-sharing properties, namely, 5426080, 16 pages, 2019. spatio-temporal characteristics, cross-regional mobilities, [5] B. Hu, Y. Gao, J. Yan et al., “Understanding the operational and actual demands, HATB takes 220-morning peak areas as efciency of bicycle-sharing based on the infuencing factor tests to validate its improved validity, feasibility, and ef- analyzes: a case study in nanjing, China,” Journal of Advanced ciency for practical application. Transportation, vol. 2021, pp. 1–14, Article ID 8818548, 2021. As compared to the similarly used methods, HATB can, [6] L. Caggiani, R. Camporeale, M. Ottomanelli, and W. Y. Szeto, accordingly, obtain the following improvements. A hierar- “A modeling framework for the dynamic management of free- chical framework is frst designed through GeoHash foating bike-sharing systems,” Transportation Research Part encoding to solve the cross-regional mobility of bikes and C: Emerging Technologies, vol. 87, pp. 159–182, 2018. Fitness Journal of Advanced Transportation 9 [7] A. A. Kadri, I. Kacem, and K. Labadi, “A branch-and-bound [22] C. S. Shui and W. Szeto, “Dynamic green bike repositioning algorithm for solving the static rebalancing problem in problem–A hybrid rolling horizon artifcial bee colony al- gorithm approach,” Transportation Research Part D: Trans- bicycle-sharing systems,” Computers and Industrial Engi- port and Environment, vol. 60, pp. 119–136, 2018. neering, vol. 95, pp. 41–52, 2016. [23] F. Chiariotti, C. Pielli, A. Zanella, and M. Zorzi, “A dynamic [8] M. Dell’Amico, E. Hadjicostantinou, M. Iori, and S. Novellani, approach to rebalancing bike-sharing systems,” Sensors, “Te bike sharing rebalancing problem: mathematical for- vol. 18, no. 2, p. 512, 2018. mulations and benchmark instances,” Omega, vol. 45, [24] M. Lang, “Study on the model and algorithm for multi-depot pp. 7–19, 2014. vehicle scheduling problem,” Journal of Transportation Sys- [9] S. Yan, J. R. Lin, Y. C. Chen, and F. R. Xie, “Rental bike tems Engineering and Information Technology, vol. 24, location and allocation under stochastic demands,” Com- pp. 65–69, 2006. puters and Industrial Engineering, vol. 107, pp. 1–11, 2017. [25] H. Bae and I. Moon, “Multi-depot vehicle routing problem [10] L. You, B. Tuncer, and H. Xing, “Harnessing multi-source with time windows considering delivery and installation data about public sentiments and activities for informed vehicles,” Applied Mathematical Modelling, vol. 40, no. 13–14, design,” IEEE Transactions on Knowledge and Data Engi- pp. 6536–6549, 2016. neering, vol. 31, no. 2, pp. 343–356, 2019. [26] K. Sakakibara, M. Noishiki, S. Watanabe, I. Nishikawa, and [11] L. You, J. He, W. Wang, and M. Cai, “Autonomous trans- H. Tamaki, “2b2 hierarchical approach with informational portation systems and services enabled by the next-generation feedback for pickup and delivery problems (technical session network,” IEEE Network, vol. 36, no. 3, pp. 66–72, 2022. 2b: vehicle scheduling and communication),” Proceedings of [12] L. You, B. Tunçer, R. Zhu, H. Xing, and C. Yuen, “A synergetic the International Seaweed Symposium, vol. 2006, pp. 48–53, orchestration of objects, data, and services to enable smart cities,” IEEE Internet of Tings Journal, vol. 6, no. 6, [27] F. Ni, L. Yan, K. Wu, M. Shi, J. Zhou, and X. Chen, “Hier- pp. 10496–10507, 2019. archical optimization of electric vehicle system charging plan [13] L. You, F. Zhao, L. Cheah, K. Jeong, P. C. Zegras, and M. Ben- based on the scheduling priority,” Journal of Circuits, Systems, Akiva, “A generic future mobility sensing system for travel and Computers, vol. 28, no. 13, Article ID 1950221, 2019. data collection, management, fusion, and visualization,” IEEE [28] Y. Zhang, “Investigation and Analysis of Travel Behaviors of Transactions on Intelligent Transportation Systems, vol. 21, Sharing Bicycles in Wenzhou,” United International Journal no. 10, pp. 4149–4160, 2020. for Research & Techonology (UIJRT), vol. 02, no. 06, 2021. [14] L. Cheng, Z. Mi, D. Cofman, J. Meng, D. Liu, and D. Chang, [29] Y. Ma, X. Qin, J. Xu, and W. Wang, “A hierarchical public “Te role of bike sharing in promoting transport resilience,” bicycle dispatching policy for dynamic demand,” in Pro- Networks and Spatial Economics, vol. 22, no. 3, pp. 567–585, ceedings of the 2016 IEEE International Conference on Service Operations and Logistics, and Informatics (SOLI), pp. 162–167, [15] F. Soriguera and E. Jimenez-Meroño, ´ “A continuous ap- Beijing, China, July 2016. proximation model for the optimal design of public bike- [30] L. You, G. Motta, K. Liu, and T. Ma, “City feed: a pilot system sharing systems,” Sustainable Cities and Society, vol. 52, of citizen-sourcing for city issue management,” ACM Article ID 101826, 2020. Transactions on Intelligent Systems and Technology, vol. 7, [16] D. Gimon, “Bike commuters contribution to balance shared no. 4, pp. 1–25, 2016. bike systems during peak load,” in Proceedings of the 2018 [31] A. Kempen, “What 3 Words Easy to use, easy to fnd any IEEE International Smart Cities Conference (ISC2), pp. 1-2, location anywhere,” Servamus Community-Based Saf. Secur. Kansas City, MO, USA, September 2018. Mag.vol. 115, no. 2, p. 57, 2022. [17] Y. Lu, U. Benlic, and Q. Wu, “An efective memetic algorithm ˇ ˇ [32] M. Muron, D. Prochazka, ´ and F. Darena, “How do people for the generalized bike-sharing rebalancing problem,” En- describe a location: towards automatically generated locations description,” in Proceedings of the 25th European Scientifc gineering Applications of Artifcial Intelligence, vol. 95, Article ID 103890, 2020. Conference of Doctoral Students, p. 85, Czech, Europe, No- vember 2021. [18] G. Shang, J. Xinzi, and T. Kezong, “Hybrid algorithm com- [33] G. M. Morton, A Computer Oriented Geodetic Data Base and bining ant colony optimization algorithm with genetic al- a New Technique in File Sequencing, International Business gorithm,” in Proceedings of the 2007 Chinese Control Machines Company New York, New York, NY, USA, 1966. Conference, pp. 701–704, Zhangjiajie, China, July 2007. [34] G. Jiang, S.-K. Lam, F. Ning, P. He, and J. Xie, “Peak-hour [19] M. Montazeri, R. Kiani, and S. S. Rastkhadiv, “A new ap- vehicle routing for frst-mile transportation: problem for- proach to the restart genetic algorithm to solve zero-one mulation and algorithms,” IEEE Transactions on Intelligent knapsack problem,” in Proceedings of the 2017 IEEE 4th In- Transportation Systems, vol. 21, no. 8, pp. 3308–3321, 2020. ternational Conference on Knowledge-Based Engineering and [35] L. Gao, W. Xu, and Y. Duan, “Dynamic scheduling based on Innovation (KBEI), pp. 0050–0053, Tehran, Iran, December predicted inventory variation rate for public bicycle system,” Sustainability, vol. 11, no. 7, p. 1885, 2019. [20] Y. Wang, J. Chen, and Y. Shen, “A multi-objective optimi- [36] P. Angeloudis, J. Hu, and M. G. H. Bell, “A strategic repo- zation model for vrp and vfp based on an improved ant colony sitioning algorithm for bicycle-sharing schemes,” Trans- algorithm,” in Proceedings of the 2019 IEEE 3rd Advanced portmetrica: Transportation Science, vol. 10, no. 8, Information Management, Communicates, Electronic and pp. 759–774, 2014. Automation Control Conference (IMCEC), pp. 777–780, [37] S. Zhao, C. Li, M. Tian, H. Zhang, Z. Xiao, and R. Hu, “Large- Chongqing, China, October 2019. scale scheduling model based on improved ant colony al- [21] D. Zhang, C. Yu, J. Desai, H. Lau, and S. Srivathsan, “A time- gorithm,” Mobile Information Systems, vol. 2022, Article ID space network fow approach to dynamic repositioning in 9116121, 6 pages, 2022. bicycle sharing systems,” Transportation Research Part B: Methodological, vol. 103, pp. 188–207, 2017. http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png Journal of Advanced Transportation Hindawi Publishing Corporation

Hierarchical Vehicle Scheduling Research on Tide Bicycle-Sharing Traffic of Autonomous Transportation Systems

Download PDF
9 pages

Loading next page...
 
/lp/hindawi-publishing-corporation/hierarchical-vehicle-scheduling-research-on-tide-bicycle-sharing-0Yf3MKe0gf

References

References for this paper are not available at this time. We will be adding them shortly, thank you for your patience.

Publisher
Hindawi Publishing Corporation
ISSN
0197-6729
eISSN
2042-3195
DOI
10.1155/2023/5725009
Publisher site
See Article on Publisher Site

Abstract

Hindawi Journal of Advanced Transportation Volume 2023, Article ID 5725009, 9 pages https://doi.org/10.1155/2023/5725009 Research Article Hierarchical Vehicle Scheduling Research on Tide Bicycle-Sharing Traffic of Autonomous Transportation Systems 1,2 1,2 1,2 1,2 Mai Hao , Ming Cai , Minghui Fang , and Shuxin Jin School of Intelligent Systems Engineering, Sun Yat-Sen University, Shenzhen 518107, China Guangdong Key Laboratory of Intelligent Transport Systems, Guangzhou 510006, China Correspondence should be addressed to Shuxin Jin; jinshx3@mail.sysu.edu.cn Received 4 July 2022; Revised 7 September 2022; Accepted 7 February 2023; Published 18 February 2023 Academic Editor: Alessandro Severino Copyright © 2023 Mai Hao et al. Tis 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. To facilitate the intelligent and automated provision of mobility services by autonomous transportation systems, bike-sharing can be a supplement to public transport for addressing their point-to-point issue, namely, “last mile” service. However, according to the diferent nature of land use, the uneven spatio-temporal distribution of travel demand can directly lead to difcult access to bikes with high travel costs for users and operating costs for operators. Based on this, this paper analyzes the user behavior patterns within diferent areas by using GeoHash encoding and proposes a hierarchical autonomous vehicle scheduling model based on tide bicycle-sharing trafc, namely, HATB. It minimizes operating costs and maximizes user satisfaction to dynamically optimize scheduling routes and required vehicles within each layered zone. As for extracting historical orders of Mobike in Beijing, an example analysis through the genetic algorithm of HATB is conducted to support efective and efcient scheduling. Compared to existing scheduling methods, HATB has faster convergence and lower time complexity, which improves bike turnaround ef- fciency and practical application ability, thus making it more convenient for the public to travel during peak hours. constant billing, users never consider the capacity of the 1. Introduction parking area, resulting in an uneven spatial distribution of Nowadays, as the modern transportation systems (TSs) bikes, namely, some areas sufer a severe accumulation of develop, problems among mobility services (MSs), e.g., bikes while others are “one bike is hard to fnd” [5]. congestion, route adjustment, user dispersion, and peak- Terefore, scientifc and reasonable scheduling strategies are time conficts have become commonplace [1–3] in terms of required to overcome the imbalance between the supply and the adjustment of urban planning and the year-on-year demand of bikes and improve resource utilization. increase in car ownership. Since the MS are fundamental in Rebalancing and optimizing bicycle-sharing distribution propelling current intelligent TS (ITS) evolving towards constitutes the vehicle routing problem (VRP), and most autonomous TS (ATS), they are also being renovated to current research is based on this theory. For instance, assist the public on a daily basis and explore the advance- Caggiani et al. [6] proposed a decision support system for the ment of ATS. Some theoretical development has revealed reallocation problem by forecasting the demand for spatio- that the creation of shared transport, called bicycle-sharing, temporal bikes. Similar research can, accordingly, be divided has efectively alleviated the “last mile” service (LMS), which into static and dynamic scheduling to optimize VRP models is the most predominant pain point in the current MS [4] in with diferent objectives. Specifcally, in static scheduling, terms of the most immediate interaction with users, and Kadri et al. [7] and Dell’Amico et al. [8] developed opti- some emerging companies, e.g., Mobike have also taken this mization models for user satisfaction and operating cost, trend to a new level. However, bikes must be parked in GPS- respectively. Yan et al. [9] investigated the deterministic and identifed areas to address ill-posed problems in the case of stochastic demand for bicycle-sharing in dynamic sched- illegal parking, vandalism, or theft. Since then, to avoid uling. In general, these methods assume that the overall 2 Journal of Advanced Transportation signifcant overall imbalance. Terefore, it is vital to take supply and demand within a scheduling station are in equilibrium without supporting the mobility of bikes be- efective scheduling strategies to rebalance and optimize the distribution of bikes, thereby addressing difculties in tween zones. Moreover, limited open literature has reported that current research focuses too much on mathematical management and operation. In this context, scheduling modeling, neglecting the analysis of actual demands. bicycle-sharing can be regarded as a heuristic algorithm- Terefore, these methods have encountered three challenges based, e.g., GA, ant colony algorithm (ACO), and vehicle in practice, namely, slow convergence, high time complexity, routing problem (VRP) [18–20], which can be classifed as and problematic application, from perspectives of fulflling static or dynamic scheduling according to diferent strategies the actual demands in real-time and adjusting scheduling as and objectives. needed. Dynamic scheduling mainly focuses on peak time and As current diversifed mobility demands tend to be relies on user demands. For instance, a mathematical model for dynamic scheduling is created by Zhang et al. [21] based managed and fulflled by more intelligent and automated systems with fewer human intervention [10], it is an urgent on the parking area’s actual capacity and users’ predicted arrival times. Shui and Szeto [22] partition peak time to need to collaborate corresponding functions to renovate the conventional MS of ITS in the context of ATS [11], i.e., optimize scheduling routes by regarding scheduling in each update the ability to sense user demands and rearrange time interval as static scheduling. Chiariotti et al. [23] system supplies [12, 13]. Hence, to promote the MS provided propose that scheduling bikes can dynamically determine by ATS, this paper proposes a hierarchical autonomous the scheduling time through historical orders. In general, vehicle scheduling model based on tide bicycle-sharing dynamic scheduling lessens operating costs’ impact on trafc, namely, HATB. Tis model uses GeoHash coding operators by prescheduling bikes to avoid a shortage occurs. However, based on the uncertain use of bikes, frequent to divide the scheduling into three layers, i.e., top, middle, and bottom, corresponding to the scheduling terminus, area, scheduling with complex constraints is necessary, which may lead to higher operating costs and slower convergence, and point. Based on the genetic algorithm (GA), the model can achieve hierarchical and dynamic scheduling of vehicles thus making it challenging to fulfll user demands in real- time. and routes to maximize user satisfaction, while minimizing operating costs. Another more common scheduling strategy is static Furthermore, in contrast to current studies on sched- scheduling during of-peak time. For example, Lang [24] uling bikes in ITS, the HATB makes three main contribu- provides a multiwarehouse model based on the Tabu search tions to optimizing convergence speed, time complexity, and algorithm to minimize scheduling distance and improve application difculties of actual scheduling. Te scheduling scheduling efciency and robustness. Bae and Moon [25] use results based on actual orders ultimately demonstrate HATB a dual time window with customer service levels to reduce total transport and labor costs. Since static scheduling only can provide a rational reference for LMS in ATS and guide the development of bicycle-sharing regulation and considers the predicted demands of stations, increasing more bikes for stations to guarantee user demands means operation. Te overall structure of this paper is divided into fve that the time complexity of the heuristic algorithm grows sections. Section 2 introduces related solutions and exponentially. Moreover, to fulfll the actual demands, the emerging challenges. Te methodology relevant to HATB is allocated bikes by these studies may exceed the station’s described in Section 3. Section 4 elaborates on the sched- capacity. uling results and superiority of the model. Finally, Section 5 Besides, such above-given studies are mainly applied to summarizes this study and sketches future research typical scenarios, as presented in Figure 1, where a single directions. scheduling station serves one zone and only the routes within the scheduling zone are considered. It is often limited in actual scheduling by the service range of the station, which 1.1. Related Works. In the transport domain, LMS refers to needs to frequently adjust the boundaries of this scheduling zone, thus leading to some research on hierarchical the direct interaction between the end of public transport and users, which often sufers from scattered users, peak- scheduling strategies. By defning scheduling priorities based time conficts, and uneven distribution. As an efective way on demand intensity, Sakakibara et al. [26] and Ni et al. [27] to cope with the LMS problems, bicycle-sharing has become highlight the feasibility and reliability of hierarchical a non-negligible component of urban transport. For ex- scheduling. In order to illustrate the fexibility, Zhang [28] ample, Cheng et al. [14] have demonstrated that bicycle- and Ma et al. [29] set stations with similar demands in the sharing increases the proportion of green transport in cities same layer in accordance with the spatio-temporal char- and solves the low efciency at the end of the travel chain. acteristics of bikes. However, the defnition of hierarchies in In general, current research on bike-sharing mainly these methods is too subjective and not clear, making it difcult to implement in practice. focuses on its development status and travel characteristics, but few on its scheduling. Researchers like Soriguera and In summary, whilst a considerable body of research has ´ been carried out on VRP, much less fts the spatial-temporal Jimenez-Meroño [15], Gimon [16], and Lu et al. [17] concur that even while bicycle-sharing has considerable quantities, and cross-regional mobility characteristics of bicycle- the spatio-temporal diferences in user demands, no fxed sharing. In addition, it seems to be a common problem parking area, and fewer available bikes may lead to a more that existing studies focus more on mathematical modeling Journal of Advanced Transportation 3 Point ID | Num. of Bikes Tese geocodes, accordingly, have good accuracy but loss fexibility; they may not meet the actual scheduling requirements. Terefore, GeoHash encoding, proposed by Morton [33], can be used to better support efective and efcient scheduling. Its maximum length of 12 bits can represent Data Collection Map Data a geographic location with arbitrary precision. For example, Parking Point the GeoHash strings WX4ER and WX4G2 represent two regions of Beijing (China), where each character is a certain rectangular area. Moreover, the order information (Data Deploy to Sources: https://biendata.xyz/competition/mobike/) on bicycle-sharing, as extracted in Table 1, also indicates the feasibility of dividing the scheduling layer via GeoHash. Redundancy Scarcity Route Optimization Te coding defnition, as described in Table 1, illustrates that the 7-bit string matches the characteristics of actual bike stops, namely, area size, and the 5-bit string suits for vehicles to dispatch bikes in light of their loading capacity, i.e., 400 bikes. Terefore, the overall framework of the proposed hierarchical scheduling model, called HATB, can be ob- tained as presented in Figure 2. In general, this framework is Scheduling Route characterized by a number of scheduling areas in each of the three layers, namely, bike stops consist of the bottom layer of Figure 1: Te schema of existing scheduling (map data (part of scheduling, while the top and middle layers likewise have Beijing): https://www.beijing.gov.cn). demands and capacity restrictions for bikes, and hence, the scheduling within the same layers is regional scheduling for but neglect the analysis of actual demands. When scheduling seeking optimization. according to the above-given methods, three challenges in terms of the problematic application in practice, namely, the high time complexity of models and slow convergence for 2.2. Model Hypotheses. Considering the complexity of actual algorithms, are increasingly apparent. scheduling, the proposed HATB in this paper defnes the Particularly driven by diverse and emerging technologies following hypotheses and the frequency of scheduling as and demands, ITS is evolving into ATS, which illustrates that once in the morning peak and once in the evening peak, MS should be autonomously fulflled and managed by more respectively. intelligent systems with fewer human intervention [30]. (1) All scheduling vehicles own the same attributes Terefore, it is an urgent need to study and improve the monolithic strategies and rationalize the actual demands to (2) In each scheduling route, the vehicle departs from one scheduling terminus (area) and returns to this achieve hierarchical and autonomous scheduling of bicycle- sharing, thus rationally guiding the provision of the LMS place after deploying bikes to corresponding areas (points) contained by ATS. (3) Fuel consumption and vehicle loss should be considered 2. The HATB Methodology (4) Each scheduling area can only be served once Tis section uses three subsections to present the framework, (5) Te actual orders determine the scheduling demand hypotheses, and construction of the proposed HATB. (6) All scheduling tasks are required to be completed within the specifed scheduling cycle 2.1. Model Framework. As existing scheduling methods, in (7) Te scheduling areas and points have sufcient space general, require frequently adjusting boundaries and in- to accommodate the bikes deployed in or out during creasing vehicles to alleviate the diference between bikes’ a scheduling cycle supply and demand, which may inevitably increase transport time and costs; the scheduling framework can be set up as a hierarchical scheduling structure, i.e., a top-middle- 2.3. Model Construction. Based on the above-given hy- bottom hierarchy of scheduling terminus-area-point. potheses and the actual operations of bike-sharing, con- However, since current hierarchical methods have sidering only the operating costs will gradually lose a subjective defnition of their hierarchies, geocode can customers, and weighing only user satisfaction runs counter ensure the objective; e.g., what3words [31] uses fxed 3m × to the essence of business proftability. Hence combining 3m squares to divide the earth, and pluscode [32] represents these two factors, this paper constructs a regional scheduling model for bikes to minimize operating costs (F each latitude and longitude level by 2-bit code, whose length ) and range in levels 1 to 3 jumps from 110km and 5.5km to 275m. maximize user satisfaction (F ). 2 4 Journal of Advanced Transportation Table 1: Te order information of corresponding GeoHash string length. GeoHash string length 5-bit 6-bit 7-bit Average orders 436.994788 28.214119 3.689289 Area width 4.89 km 1.22 km 153 m Area height 4.89 km 0.61 km 153 m Hierarchical Level Scheduling Top (4-bit) Terminus 39.1 km × 19.5 km 1 N Middle (5-bit) Scheduling 4.89 km × 4.89 km Area 1.1 1.2 1.NN.1 N.2 N.N Bottom (7-bit) Scheduling 153 km × 153 km 1.1.N 1.1.1 1.2.1 N.1.N N.2.1 N.N.N 1.2.N 1.N.1 1.N.N N.1.1 N.2.N N.N.1 Point Figure 2: Te framework of HATB. Te actual operation costs need to consider both fxed and min: zF + μF , 1 2 (1) fexible costs, as summarized in equation (2), which is de- s.t. z + μ � 1 z, μ > 0. termined jointly by the value of scheduling vehicles [34], the unit transport cost (i.e., vehicle loss: 1 CNY/km, fuel con- Te parameters z and μ indicate that bike-sharing op- sumption: 1 CNY/km, and labor cost: 100 CNY/person), and erators need to adjust the weight coefcients of operating the scheduling distance. and penalty costs according to their own emphasis. 2.3.1. F : Te Objective of Minimizing Operating Costs. minF � min(Fx d + Flx) W T I T R R (2) t t ⎝ ⎠ ⎛ ⎞ � min 􏽘 􏽘 􏽘 x C + 􏽘 􏽘 􏽘 x d M , t ij t wj ij w t i t i j I I t t s.t. 􏽘 x � 􏽘 x ≤ 1 ∀w ∈ W, t ∈ T , (3) wi wi i i T R T R t t 􏽘 􏽘 x � 􏽘 􏽘 x � 1 ∀i, j ∈ I, t ∈ T, ji ji t t (4) j�1 j�1 j≠i j≠i R R 0 ≤ 􏽘 􏽘 x d ≤ L ∀t ∈ T, ij ij t (5) j�1 j≠i t t t 0 ≤ a ≤ a x ∀i, j ∈ R, i ≠ j, t ∈ T, (6) ij ij ij x � 0 or 1 ∀i, j ∈ R, i ≠ j, t ∈ T, (7) ij t + a ∈ Z ∀i, j ∈ R, i ≠ j, t ∈ T, (8) ij Journal of Advanced Transportation 5 Table 2: Te meaning of parameters for formulations 1–14. Parameter Meaning I Scheduling area T Scheduling vehicles W Scheduling terminus R Vehicle pools for scheduling terminus and areas a Maximum capacity of vehicle t a Loading bikes of vehicle t from area i to area j ij L Maximum operation distance C Fixed cost per vehicle M Flexible cost per vehicle V Average speed d Transport distance by vehicle from area i to area j ij GAP Number of vehicles to be deployed in/out of area i ω Importance of area i TM Time for vehicle to arrive in area i S Time cost for loading/unloading bikes TM Time for vehicle to depart from terminus [FQ , FR ] Penalty time window i i distance must not exceed the maximum scheduling distance; Table 3: Examples of user travel characteristics. equation (6) points that the number of bikes loaded by the vehicle must not exceed its maximum capacity, namely, 400; Travel information Time Orders equation (7) means x as 0-1 variable; equation (8) proves that ij 7 a.m 189578 the number of bikes deployed by the vehicle is a non-negative Average distance: 815 m 18 p.m 173654 integer. 8 a.m 171011 17 p.m 164126 Median distance: 660 m 19 p.m 125383 2.3.2. F : Te Objective of Maximizing User Satisfaction. 12 p.m 119883 User satisfaction can be improved by adding time window constraints, as described in equation (9), which means maxi- mizing user satisfaction can equivalently transfer into mini- Equation (3) indicates that this scheduling vehicle starts and mizing the penalty cost of scheduling timeout. ends at the terminus; equation (4) suggests that each area can only be served once; equation (5) shows that the transport maxF � min 􏽘 ω ρ t􏼁 , (9) 2 i i i s.t. TM � 0 ∀t ∈ T , (10) T R TM � 􏽘 􏽘 x 􏼐TM + TM + S · GAP 􏼑, ∀j ∈ I, (11) j ij i j i TM + TM + S · GAP − K􏼐1 − x 􏼑 ≤ TM , (12) i j i j ij FQ ≤ TM ≤ FR , ∀i ∈ I. (13) i i i Equation (10) indicates that the vehicle departs from Te meaning of the parameters in the above-given the terminus at time zero; equation (11) means that the equationsare shown in Table 2. time calculation for a vehicle to arrive in an area; 3. Case Study equation (12) demonstrates that the scheduling cannot arrive in area j from area i before TM + TM + S · GAP − i j i Te highlights of HATB solving are illustrated in terms of al- K(1 − x ); equation (13) represents that vehicle needs to gorithm settings, scheduling results, and model evaluation in this ij arrive in an area within the time window. section. 6 Journal of Advanced Transportation Input: SchedulingArea SA, DistanceMatrix DM Output: SchedulingRoute SR, SchedulingVehicle SV, DeployedBikes DR Initialization: Generation Gen � 100, CrossoverRate CR � 0.8, MutationRate MR � 0.2, Population Pop ← ∅, Chromosome CH ← ∅, Fitness Fit � Null, (1) For all the SA do (2) Pop ← CH.generate (SA.Encoding)//Randomly Generation for Population (100) (3) End for (4) Fit ← Fit.Calculation//Calculate the Fitness (5) if Fit.change or Gen < 100 then (6) For all the CH do (7) CH.select//Chromosome Selection (8) CH.crossover//Chromosome Crossover (9) CH.mutate//Chromosome Mutation (10) Pop ← CH.NewGenerate//New Population (11) Fit ← Fit.NewCalculation (12) End for (13) End if (14) SR ← CH.Decoding (15) Return SR, SV, DR ALGORITHM 1: NGA for scheduling optimisation. Area ID | Area Bikes Parking Point Data Collection Routes of Points Route Optimization for Areas Deploy to Redundancy Scarcity Scheduling Terminus Route Optimization for Points Region-by-Region Scheduling Route of Areas Scheduling Areas Scheduling Points Figure 3: Te schema of HATB. 3.1. Algorithm Settings. Te experiment data comes from the diferent vehicles, e.g., a chromosome example might be 0-3- 2017 Mobike cup algorithm challenge, which involves 0-1-2-5-7-0-4-8-6-0, namely, three vehicles serving th rd 3,214,096 orders and 485,465 bikes (10 May 2017–23 May eight zones. 2017). Te characterized user travel, as presented in Table 3, Since the crossover and mutated sub chromosomes may refects that the data are consistent with the “last mile” lead to transport overload and overtime and the time defnition [16], and hence, shows its reasonable usability window is more likely to be violated, the penalty factors for according to prominent tidal characteristics. the two constraints are set to 10 and 500, respectively. Tis paper proposes a GA with natural number encoding (NGA), as defned in Algorithm 1, to optimize the sched- 3.2. Scheduling Results. A total of 220 scheduling areas in the uling. In general, the frst and the last 0 s represent the morning peak are used as HATB test cases to obtain the optimal hierarchical scheduling routes, as shown in Figure 3 scheduling terminus or area, [1, N] represents the zones that need to be scheduled, and other 0 s separate the routes of and Tables 4 and 5. Journal of Advanced Transportation 7 Table 4: Te optimal routes for scheduling areas in the morning peak. Routes Demand Number of vehicles Total distance (km) for scheduling area 0-135-29-113-112-150-205-20-44-170-130-146-0 143 1 0-40-88-138-104-65-12-191-115-52-162-83-157-0 142 1 0-7-45-93-61-129-188-199-158-217-87-53-165-85-119-36-34-210-0 140 1 0-5-6-50-47-41-75-63-86-27-101-125-200-178-81-78-89-122-187-195-203-207-0 140 1 0-18-21-95-132-148-91-71-161-192-117-0 139 1 0-181-106-102-145-135-29-113-112-150-205-20-0 138 1 0-51-177-92-79-84-26-38-214-215-166-168-183-0 133 1 0-31-143-114-37-118-99-156-163-137-149-194-0 114 1 0-54-66-74-128-94-136-201-179-190-152-35-30-33-204-218-0 114 1 719.542007 0-90-78-89-97-64-105-185-131-127-189-140-100-8-82-22-25-98-24-144-124-197-0 113 1 0-3-70-55-169-172-182-109-208-0 113 1 0-62-103-202-43-11-16-57-23-108-153-211-0 111 1 0-4-1-2-14-17-67-48-15-42-56-72-68-99-156-219-0 110 1 0-44-170-30-146- 110 1 159-173-174-69-9-46-49-19-141-186-175-171-39-213-154-51-60-0 0-180-139-142-198-147-116-212-134-167-184-126-216-0 110 1 0-76-164-10-77-73-60-111-120-196-209-0 80 1 0-28-133-110-193-123-154-32-13-107-121-155-206-0 79 1 Table 5: Examples of hierarchical autonomous scheduling results for route one. Routes Scheduling area Total distance (km) for scheduling point a0-a11-a4-a18-a17-a9-a1-a12-a15-a19-a5-a0 170 (a0) 1.443 b0-b1-b3-b11-b4-b48-b39-b18-b42-b38-b7-b19-b54-b2-b28-b25-b26-b27-b43- b31-b12-b21-b45- b52-b30-b20-b46-b47-b6-b29-b40-b37-b14-b33-b9-b34-b51- 130 (b0) 3.905 b24-b32-b15-b55-b10-b44-b22-b16-b41-b17-b5-b13-b23-b53-b8-b50-b49-b36- b35-b0 c0-c17-c23-c13-c15-c29-c32-c10-c35-c31-c2-c34-c5-c4-c14-c33-c22-c11-c9-c3-c6- 146 (c0) 2.951 c21-c20-c25-c30-c28-c24-c16-c27-c19-c8-c12-c26-c1-c18-c7-c0 Table 6: Te mapped scheduling results with examples of route one (Table 5). Routes for scheduling Scheduling area (5-bit) point (7-bit) WX54C (no. 170) WX54C with 0W, 20, E4, 06, 0C, J8, JB, LJ, NC, BW, BX, 0W WX4GW with 00, 02, 08, PL, 06, 77, 7K, P4, NF, 7M, ZH, YY, N8, P0, PF, PY, PV, WX4GW (no. 130) PS, ND, JL, P3, PB, N9, 03, JH, PZ, 7B, 78, 7F, 7E, KH, 7U, 7Q, EQ, ZN, EN, 09, Q8,G7, ZQ, GJ, ZK, NB, P8, P9, P5, 01, 04, 2N, U4, FZ, ZV, 0D, 3K, 86, DK, 00 WX5H2 with 5X, 7C, QX, DM, GL, 6M, Y6, F9, 4N, 4J, DS, DT, VC, TY, DC, D3, WX5H2 (no. 146) TZ, ER, G0, G8, 74, UE, UK, GE, 1E, 12, QK, 7D, 75, 77, W2, EP, G9, ZG, 79, 48, 5X As described in the previous section, regional scheduling is scheduling terminus and returning to the terminus after applied for each layer. Note that, each scheduling area has completing regional scheduling sequentially and autonomously a positive or negative raw demand that refects the redundancy in accordance with the area-point (middle-bottom) hierarchy. In addition, the routes’ information in reality for Table 5 is mapped or scarcity of bikes. Furthermore, scheduling prioritizes self- satisfaction in the route, namely, redundancy supports scarcity, to Table 6 via GeoHash. and hence, route demand indicates the self-satisfaction gap for corresponding regional scheduling. 3.3. Model Evaluation. To further verify the reliability of this Terefore, the experimental results can be summarized as model, Figure 4 shows the comparison for iteration between the total scheduling distance for optimal routes in the entire the HATB and other models proposed in the literature with Beijing is around 719.5 km. As for satisfying the demands, 17 similar objectives. Specifcally, based on the GA, Gao et al. vehicles are required to participate in the scheduling to deploy [35] provided a promising perspective on improving op- bikes. Moreover, using route one in Table 4 as an example, the eration efciency by reducing operating costs and service scheduling can be summarized as a vehicle departing from the quality during peak times to minimize the total operating 8 Journal of Advanced Transportation reduce the time complexity of global optimization. Next, 1.05E+6 a GA for regional scheduling is built by combining the tidal characteristics of bicycle-sharing to minimize operating costs and maximize user satisfaction, which signifcantly 9.00E+5 accelerates the algorithm’s convergence. Last, the use of actual orders considerably enhances the ability in the practical application of instant response to any regional scheduling demand. 7.50E+5 Tis work was carried out as a preliminary to obtain the present results. However, there are still problems such as the inability to adapt scheduling throughout 24 hours or the lack 6.00E+5 of comprehensive constraints. As the closure of this study, one recommendation for further research is to use a form of “GA + Tabu” algorithm to exploit its global search capability 0 20 40 60 80 100 and thus improve the big data processing capability. Another Generation research direction is adding weather and road characteristics The Proposed HATB Gao L et al., 2019 to optimize the model reliability further. Zhao et al., 2022 Panagiotis et al., 2014 Figure 4: Te comparison between diferent scheduling models. Data Availability costs. Angeloudis et al. [36] achieved user appeal increase by Te data used to support the fndings of this study have been deposited in the 2017 Mobike Cup Algorithm Challenge ofering a new method of planning bike routes and distri- butions. Moreover, Zhao et al. [37] optimized the total repository (https://biendata.xyz/competition/mobike/). scheduling distance to accommodate large-scale scheduling via an ACO. Conflicts of Interest th Te proposed HATB converged at the 64 generation, and the total time cost is 148.9 s, with an average running time of 15 s Te authors declare that there are no conficts of interest regarding the publication of this paper. per generation. Due to the diferent objectives, only the con- vergence speed of the above models is compared. It can be seen from Figure 4 that the HATB signifcantly outperforms the Acknowledgments model proposed in the existing literature. Such a result indicates that this model is more proper for practical scheduling appli- Tis research was supported by National Key Research and Development Program of China (Grant no. cations since it optimizes with higher convergence speed and lower time complexity. 2020YFB1600400) and Collaborative Innovation Center for Transportation of Guangzhou (Grant no. 202206010056). 4. Conclusion and Future Works References Even though the emerging and diversifed technologies and [1] S. Moridpour, X. Qu, N. Shiwakoti, and S. Hasan, “Sustainable demands are driving TS to renovate conventional MS to be self- and resilient transport infrastructure,” Journal of Advanced actuating, the current bicycle-sharing scheduling and mainte- Transportation, vol. 2021, Article ID 1576315, 2 pages, 2021. nance rely on manual experience, which lacks scientifc guidance [2] M. L. Mouronte-Lopez, ´ “Analysing the vulnerability of public and efciency. Terefore, to achieve sustainable development, transport networks,” Journal of Advanced Transportation, operators urgently need to develop a rational scheduling strategy vol. 2021, Article ID 5513311, 22 pages, 2021. to balance the distribution conficts and supply user demands [3] N. Tsanakas, J. Ekstrom, ¨ and J. Olstam, “Estimating emissions in time. from static trafc models: problems and solutions,” Journal of Terefore, this paper proposes a hierarchical scheduling Advanced Transportation, vol. 2020, Article ID 5401792, model, called HATB, to address the unsolved issues by 17 pages, 2020. current studies in terms of slow convergence, high time [4] A. Fan, X. Chen, and T. Wan, “How have travelers changed mode choices for frst/last mile trips after the introduction of complexity, and problematic application, and hence, to bicycle-sharing systems: an empirical study in Beijing, China,” support the rational and autonomous provision of LMS. In Journal of Advanced Transportation, vol. 2019, Article ID summary, according to bicycle-sharing properties, namely, 5426080, 16 pages, 2019. spatio-temporal characteristics, cross-regional mobilities, [5] B. Hu, Y. Gao, J. Yan et al., “Understanding the operational and actual demands, HATB takes 220-morning peak areas as efciency of bicycle-sharing based on the infuencing factor tests to validate its improved validity, feasibility, and ef- analyzes: a case study in nanjing, China,” Journal of Advanced ciency for practical application. Transportation, vol. 2021, pp. 1–14, Article ID 8818548, 2021. As compared to the similarly used methods, HATB can, [6] L. Caggiani, R. Camporeale, M. Ottomanelli, and W. Y. Szeto, accordingly, obtain the following improvements. A hierar- “A modeling framework for the dynamic management of free- chical framework is frst designed through GeoHash foating bike-sharing systems,” Transportation Research Part encoding to solve the cross-regional mobility of bikes and C: Emerging Technologies, vol. 87, pp. 159–182, 2018. Fitness Journal of Advanced Transportation 9 [7] A. A. Kadri, I. Kacem, and K. Labadi, “A branch-and-bound [22] C. S. Shui and W. Szeto, “Dynamic green bike repositioning algorithm for solving the static rebalancing problem in problem–A hybrid rolling horizon artifcial bee colony al- gorithm approach,” Transportation Research Part D: Trans- bicycle-sharing systems,” Computers and Industrial Engi- port and Environment, vol. 60, pp. 119–136, 2018. neering, vol. 95, pp. 41–52, 2016. [23] F. Chiariotti, C. Pielli, A. Zanella, and M. Zorzi, “A dynamic [8] M. Dell’Amico, E. Hadjicostantinou, M. Iori, and S. Novellani, approach to rebalancing bike-sharing systems,” Sensors, “Te bike sharing rebalancing problem: mathematical for- vol. 18, no. 2, p. 512, 2018. mulations and benchmark instances,” Omega, vol. 45, [24] M. Lang, “Study on the model and algorithm for multi-depot pp. 7–19, 2014. vehicle scheduling problem,” Journal of Transportation Sys- [9] S. Yan, J. R. Lin, Y. C. Chen, and F. R. Xie, “Rental bike tems Engineering and Information Technology, vol. 24, location and allocation under stochastic demands,” Com- pp. 65–69, 2006. puters and Industrial Engineering, vol. 107, pp. 1–11, 2017. [25] H. Bae and I. Moon, “Multi-depot vehicle routing problem [10] L. You, B. Tuncer, and H. Xing, “Harnessing multi-source with time windows considering delivery and installation data about public sentiments and activities for informed vehicles,” Applied Mathematical Modelling, vol. 40, no. 13–14, design,” IEEE Transactions on Knowledge and Data Engi- pp. 6536–6549, 2016. neering, vol. 31, no. 2, pp. 343–356, 2019. [26] K. Sakakibara, M. Noishiki, S. Watanabe, I. Nishikawa, and [11] L. You, J. He, W. Wang, and M. Cai, “Autonomous trans- H. Tamaki, “2b2 hierarchical approach with informational portation systems and services enabled by the next-generation feedback for pickup and delivery problems (technical session network,” IEEE Network, vol. 36, no. 3, pp. 66–72, 2022. 2b: vehicle scheduling and communication),” Proceedings of [12] L. You, B. Tunçer, R. Zhu, H. Xing, and C. Yuen, “A synergetic the International Seaweed Symposium, vol. 2006, pp. 48–53, orchestration of objects, data, and services to enable smart cities,” IEEE Internet of Tings Journal, vol. 6, no. 6, [27] F. Ni, L. Yan, K. Wu, M. Shi, J. Zhou, and X. Chen, “Hier- pp. 10496–10507, 2019. archical optimization of electric vehicle system charging plan [13] L. You, F. Zhao, L. Cheah, K. Jeong, P. C. Zegras, and M. Ben- based on the scheduling priority,” Journal of Circuits, Systems, Akiva, “A generic future mobility sensing system for travel and Computers, vol. 28, no. 13, Article ID 1950221, 2019. data collection, management, fusion, and visualization,” IEEE [28] Y. Zhang, “Investigation and Analysis of Travel Behaviors of Transactions on Intelligent Transportation Systems, vol. 21, Sharing Bicycles in Wenzhou,” United International Journal no. 10, pp. 4149–4160, 2020. for Research & Techonology (UIJRT), vol. 02, no. 06, 2021. [14] L. Cheng, Z. Mi, D. Cofman, J. Meng, D. Liu, and D. Chang, [29] Y. Ma, X. Qin, J. Xu, and W. Wang, “A hierarchical public “Te role of bike sharing in promoting transport resilience,” bicycle dispatching policy for dynamic demand,” in Pro- Networks and Spatial Economics, vol. 22, no. 3, pp. 567–585, ceedings of the 2016 IEEE International Conference on Service Operations and Logistics, and Informatics (SOLI), pp. 162–167, [15] F. Soriguera and E. Jimenez-Meroño, ´ “A continuous ap- Beijing, China, July 2016. proximation model for the optimal design of public bike- [30] L. You, G. Motta, K. Liu, and T. Ma, “City feed: a pilot system sharing systems,” Sustainable Cities and Society, vol. 52, of citizen-sourcing for city issue management,” ACM Article ID 101826, 2020. Transactions on Intelligent Systems and Technology, vol. 7, [16] D. Gimon, “Bike commuters contribution to balance shared no. 4, pp. 1–25, 2016. bike systems during peak load,” in Proceedings of the 2018 [31] A. Kempen, “What 3 Words Easy to use, easy to fnd any IEEE International Smart Cities Conference (ISC2), pp. 1-2, location anywhere,” Servamus Community-Based Saf. Secur. Kansas City, MO, USA, September 2018. Mag.vol. 115, no. 2, p. 57, 2022. [17] Y. Lu, U. Benlic, and Q. Wu, “An efective memetic algorithm ˇ ˇ [32] M. Muron, D. Prochazka, ´ and F. Darena, “How do people for the generalized bike-sharing rebalancing problem,” En- describe a location: towards automatically generated locations description,” in Proceedings of the 25th European Scientifc gineering Applications of Artifcial Intelligence, vol. 95, Article ID 103890, 2020. Conference of Doctoral Students, p. 85, Czech, Europe, No- vember 2021. [18] G. Shang, J. Xinzi, and T. Kezong, “Hybrid algorithm com- [33] G. M. Morton, A Computer Oriented Geodetic Data Base and bining ant colony optimization algorithm with genetic al- a New Technique in File Sequencing, International Business gorithm,” in Proceedings of the 2007 Chinese Control Machines Company New York, New York, NY, USA, 1966. Conference, pp. 701–704, Zhangjiajie, China, July 2007. [34] G. Jiang, S.-K. Lam, F. Ning, P. He, and J. Xie, “Peak-hour [19] M. Montazeri, R. Kiani, and S. S. Rastkhadiv, “A new ap- vehicle routing for frst-mile transportation: problem for- proach to the restart genetic algorithm to solve zero-one mulation and algorithms,” IEEE Transactions on Intelligent knapsack problem,” in Proceedings of the 2017 IEEE 4th In- Transportation Systems, vol. 21, no. 8, pp. 3308–3321, 2020. ternational Conference on Knowledge-Based Engineering and [35] L. Gao, W. Xu, and Y. Duan, “Dynamic scheduling based on Innovation (KBEI), pp. 0050–0053, Tehran, Iran, December predicted inventory variation rate for public bicycle system,” Sustainability, vol. 11, no. 7, p. 1885, 2019. [20] Y. Wang, J. Chen, and Y. Shen, “A multi-objective optimi- [36] P. Angeloudis, J. Hu, and M. G. H. Bell, “A strategic repo- zation model for vrp and vfp based on an improved ant colony sitioning algorithm for bicycle-sharing schemes,” Trans- algorithm,” in Proceedings of the 2019 IEEE 3rd Advanced portmetrica: Transportation Science, vol. 10, no. 8, Information Management, Communicates, Electronic and pp. 759–774, 2014. Automation Control Conference (IMCEC), pp. 777–780, [37] S. Zhao, C. Li, M. Tian, H. Zhang, Z. Xiao, and R. Hu, “Large- Chongqing, China, October 2019. scale scheduling model based on improved ant colony al- [21] D. Zhang, C. Yu, J. Desai, H. Lau, and S. Srivathsan, “A time- gorithm,” Mobile Information Systems, vol. 2022, Article ID space network fow approach to dynamic repositioning in 9116121, 6 pages, 2022. bicycle sharing systems,” Transportation Research Part B: Methodological, vol. 103, pp. 188–207, 2017.

Journal

Journal of Advanced Transportation – Hindawi Publishing Corporation

Published: Feb 18, 2023

There are no references for this article.