Archive

25 June 2024, Volume 24 Issue 3 Previous Issue    Next Issue

---

For Selected:

Download Citations EndNote Ris BibTeX

Toggle Thumbnails

Select

A Train Group Control Method Based on Car Following Model Under Virtual Coupling SHUAI Bina, LUO Jianan, FENG Xinyan, HUANG Wencheng 2024, 24(3): 1-11.  DOI: 10.16097/j.cnki.1009-6744.2024.03.001 Abstract ( )   PDF (2282KB) ( )   PDF(English version) (1740KB) ( 1 )   There remains a gap between transportation capacity and demand under the high-speed railway moving block mode, prompting the exploration of new approaches such as virtual coupling to enhance transportation capacity. With the concept of virtual coupling and inspired by car-following models utilized in road traffic, we propose a novel acceleration adjustment strategy by train dynamics and multi-agent methods for tracking trains based on the speed and distance relationship between adjacent trains, with the goal of ensuring train safety and passenger comfort while enabling virtual coupling within the train group. A corresponding virtual coupling acceleration adjustment model is established for train groups, aiming to achieve equal speed and distance between all trains in the group. The proposed model is validated using the CRH380A high-speed train as a case study. Simulation results demonstrate that the proposed acceleration adjustment strategies effectively realize the virtual coupling of train groups. Compared to the moving block method, adopting virtual coupling reduces the time required for train collaboration by 9.7% and decreases the distance between trains by 10.1% , thereby improving efficiency. Furthermore, the time required to achieve virtual coupling is shorter when considering the train group as a whole compared to when the group is separated into multiple groups. References | Related Articles | Metrics

Select

Dynamic Clearance Control Method for Reusing Bus Lanes Under Vehicular Networking DONG Hongzhao, YANG Jiawei, QUAN Cheng 2024, 24(3): 12-20.  DOI: 10.16097/j.cnki.1009-6744.2024.03.002 Abstract ( )   PDF (2405KB) ( )   Traditional dynamic control methods for dedicated bus lanes can be improved to ensure both the bus priority and the lane utilization rates. To address this issue, this paper proposes a dynamic clearance control method for the reutilization of dedicated bus lanes with the support from vehicular networking, which is also referred to as Dynamic Clearance Bus Lane (DCBL). This method establishes a clearance framework model that dynamically adjusts the speed of connected buses and the lane-changing time of connected private vehicles. Additionally, it defines a lane change urgency coefficient and uses the fuzzy control theory to design a lane change probability output algorithm in consideration of drivers' lane-changing psychology to simulate the actual lane-changing process. The simulation analysis was conducted to verify the effectiveness of the DCBL control method. The results indicate that the DCBL control method expands the applicable range of traffic density to 0~71 pcu · km- 1 , an increase of 9~21 pcu · km- 1 compared to traditional BLIP(Bus Lane with Intermittent priority) and IBL(Intermittent Bus Lane) control methods. In the mid-to-high-density range of 40~70 pcu · km-1 , the DCBL control method maintains the average speed of private vehicles at 45.86 km·h-1 , an improvement of 17.9%~24.7% compared to traditional control methods. The average speed of buses is maintained at 33.68 km· h-1 , only decreasing by 6.4% compared to the expected speed of buses. The DCBL control method results in a bus travel delay of less than 25 seconds mid-to-high-density range, leading to an increase in roadway throughput by 8.0%~18.3% compared to traditional control methods. References | Related Articles | Metrics

Select

Stackelberg Game-based Control Method for Driver-automation Collaboration in Ship Remote-control LI Chen, YAN Xinping, LIU Jialun, HUANG Yamin, LI Shijie 2024, 24(3): 21-31.  DOI: 10.16097/j.cnki.1009-6744.2024.03.003 Abstract ( )   PDF (2855KB) ( )   To solve the problem of human-machine control objective non-consistency in the ship remote-control process, this paper proposes a shared steering control method within the Stackelberg framework and considers the human-dominated and machine-auxiliary operating mode of the system. The human-machine interaction in ship collision avoidance collaborative steering task is described as a non-cooperative game relationship under complete information conditions. By constructing the state space of the driver and the co-pilot controller, the differential strategy is derived for Stackelberg game, and the uniqueness and existence of Nash equilibrium solution is proved with FanGlicksberg fixed points theorem. Based on model predictive control method, the trajectory tracking controller is designed with pre-allocating driving weight for different driving style and maneuvering skills, rolling and optimization in a finite time domain through feedback correction. And the control authority will be adjusted online in combination of the safety navigation boundary, collision risk and degree of human-machine conflicts. Taking the lateral displacement and driver's operational load as evaluation indexes, the effectiveness of method is verified in inland maneuvering scenarios. Simulation results show that the proposed method could provide personalized assistance for remote operators with different driving styles and maneuvering skills, while there exists intention conflict between the driver and the co-pilot controller, it can adjust the pre-allocated weight in accordance with the navigating risk dynamically, so as to make the ship motion more compliant with driver's maneuvering intentions under the premise of ensuring navigating safety. References | Related Articles | Metrics

Select

Pseudo-range Spoofing Detection Method for Satellite-based Train Positioning Using Sample Augmentation LIU Jiang, ZHANG Chu, CAI Baigen, WANG Jian, LU Debiao 2024, 24(3): 32-42.  DOI: 10.16097/j.cnki.1009-6744.2024.03.004 Abstract ( )   PDF (2993KB) ( )   Satellite-based autonomous train positioning is an important technical direction for train control systems and other key equipment in railway applications. However, satellite-based train positioning has to face many challenges. In addition to signal visibility and multipath effect, intentional spoofing and other interference attacks from the outside would directly threaten the positioning function and the achieved performance level. This paper takes the pseudo-range spoofing, which is a typical interference mode for train positioning based on the Global Navigation Satellite System (GNSS), as the study object and proposes an active detection method based on sample augmentation. This method adopted the Wasserstein Generative Adversarial Networks (WGAN) to solve the problem of imbalanced spoofing affected samples. It trained a detection model using the expanded datasets and introduces a Self-Attention (SA) mechanism to optimize the relative positional relationship between the input features from different GNSS receivers. A complete detection scheme for the pseudo-range-mode GNSS spoofing to satellite-based train positioning was established based on generative adversarial learning. According to the results from the pseudo-range GNSS spoofing injection tests to the satellite-based train positioning, the proposed method can make full use of the Generative Adversarial Network solution to solve the typical problem caused by limited spoofing samples. The detection performance derived with the self-attention mechanism is significantly enhanced over the typical conventional detection methods. It realizes the adaptability to the features not covered by the modeling samples, with enhanced detection accuracy and robustness. The F1 score obtained in the test with multiple pseudo-range GNSS spoofing mode datasets exceeded 0.99. The advantages of the proposed method in spoofing detection performance can provide great support for many GNSS-based railway applications and enable favorable conditions for effectively protecting attacks to GNSS at the information security level. References | Related Articles | Metrics

Select

Longitudinal Speed Control Method of Unmanned Vehicle Under Undulating Road Conditions LI Jianshi, XU Youchun, QI Yao, XIE Desheng, LI Hua 2024, 24(3): 43-52.  DOI: 10.16097/j.cnki.1009-6744.2024.03.005 Abstract ( )   PDF (2945KB) ( )   To improve the accuracy of the speed control of unmanned vehicles under undulating road conditions, a method for preview detection of road slopes on the target trajectory was proposed. Firstly, the response delay parameters of the vehicle speed control system were measured by experiments to determine the preview time. The lidar point cloud was stored in a circle with the diameter of the vehicle width, and the number of circles and the center coordinates of the stored point cloud were determined according to the preview time, the vehicle speed, the detection blind area of the lidar, and the length of the target trajectory. According to the Euclidean distance between the center of the circle and the center of the front axle of the vehicle, the preview time, and the vehicle speed, the circles of the stored point cloud are dynamically updated and extracted, and the slope of the road surface in the preview area relative to the vehicle attitude is calculated by the trigonometric function method. Finally, the absolute slope of the road surface in the preview area is obtained by superimposing the pitch angle of the vehicle provided by the inertial navigation system. Experimental results show that the proposed road slope preview detection method can improve the accuracy of the speed control of unmanned vehicles under undulating road conditions. According to the analysis of the minimum speed of unmanned vehicles in the process of going uphill, compared with the control method that ignores the road slope and uses the pitch angle provided by inertial navigation as the road slope, the accuracy of speed control is increased by 5.9% and 2.5% respectively when passing through a 4° ramp, and the accuracy of speed control is increased by 85% and 17.5% when passing a 16° ramp, respectively. References | Related Articles | Metrics

Select

Computer Vision-based Fire Detection and Localization Inside Urban Rail Transit Stations ZHANG Jinlei, YANG Jian, LIU Xiaobing, CHEN Yao, YANG Lixing, GAO Ziyou 2024, 24(3): 53-63.  DOI: 10.16097/j.cnki.1009-6744.2024.03.006 Abstract ( )   PDF (3076KB) ( )   To efficiently address the occurrence of in-station fire incidents in rail transit, this paper proposes a computer vision-based model for fire detection and precise fire localization within the rail stations, which is referred to as Fire-Detect. First, this study created the Fire-Rail dataset using the Unity simulation and collecting internet images, which established the dataset to train the fire detection and precise localization algorithms. Then, a fire detection algorithm was developed to integrate convolutional neural networks, residual structures, and channel attention mechanisms. This algorithm classifies each frame of surveillance video within the station as either "normal" or "suspected fire" status. In the "suspected fire" status, the model activates the precise localization algorithm. It processes the "suspected fire" image along with subsequent frames, providing real-time, detailed fire localization information to station attendants. Experimental results on the Fire-Rail dataset demonstrated a fire detection accuracy of 95.12% on the test set. Furthermore, hierarchical experiments with convolutional neural network layers balance the resource consumption and accuracy. Ablation experiments confirmed the effectiveness of individual components, and robustness experiments indicated the algorithm's ability to handle most noise. The overall model achieves an average fire localization detection accuracy (mAP) of 77.3% and is suitable for deployment in video surveillance equipment within rail transit stations. References | Related Articles | Metrics

Select

Impacts of Accessibility on Spatial Heterogeneity of Multimodal Transportation Carbon Emissions in an Urban Agglomeration MA Shuhong, CHEN Xifang, YANG Lei, ZHAO Yuzhe, ZENG Yu 2024, 24(3): 64-74.  DOI: 10.16097/j.cnki.1009-6744.2024.03.007 Abstract ( )   PDF (3003KB) ( )   Under the double stress of the significant challenge posed by carbon emission reduction in the current transportation industry and the long-term plan for transportation development of China's urban agglomerations, improving the travel efficiency and simultaneously reducing carbon emission by improving accessibility is one of the key issues that need to be addressed urgently. This study proposed a method to estimate the carbon emission of intercity multimodal passenger transportation from the perspective of residents' travel using mobile phone signal data. Then, the impact of accessibility on the spatial heterogeneity of multimodal transportation carbon emissions was investigated using the gradient boosting decision tree (GBDT) model and the multi-scale geographically weighted regression (MGWR) model. The Guanzhong Plain urban agglomeration is taken as the study area. The results show that: the carbon emissions of the intercity road passenger are higher than railway, and distribution characteristics along transportation infrastructure lines. Accessibility has a positive marginal effect on the level of carbon emissions in the whole area. The MGWR can characterize the spatial heterogeneity and scale difference in the relationship between carbon emissions and accessibility indicators. Economic potential, betweenness centrality, and closeness centrality have significant positive spatial heterogeneity impacts on the intercity carbon emissions, but the scale of the effects varies. Carbon emissions of road passenger transportation are sensitive to betweenness centrality and closeness centrality, but the spatial impact of economic potential on carbon emissions is stable. The effect of increased rail travel accessibility on the central city is lower than that on the surrounding district and counties. References | Related Articles | Metrics

Select

Optimization of Urban Emergency Vehicle Rescue Route Considering Dynamic Impact of Rainstorm Disasters HU Xiaowei, LU Hongbo, AN Shi 2024, 24(3): 75-82.  DOI: 10.16097/j.cnki.1009-6744.2024.03.008 Abstract ( )   PDF (1669KB) ( )   In recent years, the frequency and intensity of extreme weather events show an increasing trend, in which the urban inland flooding caused by heavy rains escalates the likelihood of traffic emergencies. To accelerate emergency rescue during the rainstorm disasters, this study focuses on the optimization of emergency vehicle rescue routes under such conditions. To minimize the passage time considering the dynamic impact of road water on vehicle passing speed, this paper develops an emergency rescue path optimization model. A dynamic shortest path optimization algorithm is proposed to solve the model. The northeast of Changning District of Shanghai is chosen as the study area. According to the water accumulation of the urban road surface under the condition of a rainstorm in 50 years simulated by the Storm Water Management Model (SWMM), the paper sets the emergency rescue scenario and solves the emergency rescue path. By comparing the path solved by the proposed algorithm with the traditional static shortest path algorithm, the traffic time is reduced by 25.42%. Furthermore, this paper also considers the emergency supplies reserve to allocate emergency rescue tasks, expands the application scenarios of the algorithm, and forms a reliable and efficient emergency response scheme, which provides a reference for improving the efficiency of emergency response under rainstorm disasters. References | Related Articles | Metrics

Select

Container Slot Allocation and Perceived Slot Chartering Price ZHENG Jianfeng, XIAO Yiling, ZHAO Zhihao, CHU Yanling 2024, 24(3): 83-93.  DOI: 10.16097/j.cnki.1009-6744.2024.03.009 Abstract ( )   PDF (1818KB) ( )   Container slot services mainly include: (i) serving cargo owners through slot allocation; and (ii) serving liner shipping companies through slot chartering. In order to increase the revenue of liner shipping companies while improving container slot utilization, this paper studies container slot allocation and perceived slot chartering price, by proposing a two-stage optimization method. In the first stage, the fleet deployment and slot allocation problem is studied, and a mixed integer linear programming model is constructed with the objective of maximizing the total revenue of the liner shipping company. In the second stage, the inverse optimization technique is used to determine the perceived slot chartering price, based on the solution of the first stage model. Finally, numerical experiments and sensitivity analysis are carried out based on the Asia-Europe-Oceania container transportation network, which contains 11 ship routes and 46 ports. The results show that the fleet deployment results are sensitive to the changes in container demands, while not sensitive to small fluctuations of container market freight rates. Perceived slot chartering prices are sensitive to both the changes in container demands and container market freight rate fluctuations. Liner shipping companies should always be aware of the container market dynamics, in order to determine suitable strategies for fleet deployment, slot allocation, and slot chartering. References | Related Articles | Metrics

Select

Delivery Logistics Network Design for Mountainous Rural Areas with Parcel Transportation by Bus and Simultaneous Home Delivery and Customer Self-pickup SUN Wenjie, ZHANG Jin, LIU Jiao , LI Guoqi 2024, 24(3): 94-102.  DOI: 10.16097/j.cnki.1009-6744.2024.03.010 Abstract ( )   PDF (1525KB) ( )   Efficient and smooth "first-last mile" service is the key to the sustainable operation of rural delivery logistics system. Aiming at the characteristics of small and scattered delivery demand in mountainous rural areas with wide coverage of bus service but high vacancy rate, this paper proposes an optimization model to minimize the total cost of the delivery logistics service. The model integrates parcel transportation by bus with simultaneous home delivery and customer self-pickup into the design of delivery logistics network in mountainous rural areas. The model includes an optimization to the locations of transfer points and service stations and the delivery vehicle routes. An improved Benders decomposition (BD) algorithm is designed and accelerated by some valid inequalities and a warm-start strategy. The effectiveness of the model and algorithm is verified by an actual example in a mountainous township in Qingchuan, Sichuan. The results show that the suggested network design mode has more cost advantages and can reduce the cost by at least 4.91%. The acceleration strategy enhances the computational efficiency remarkably, and the solution time of the enhanced BD algorithm is 66.06% lower than the traditional BD algorithm. The sensitivity analysis found that the total cost shows a decreasing trend with the increase of bus capacity to transport parcels. As the coverage radius decreases, the increase rate of total cost tends to decrease and then increase. References | Related Articles | Metrics

Select

An Optimization Model of Arterial Multi-path Green-band Considering Path Relationships WU Changjian, CAO Qi, REN Gang 2024, 24(3): 103-113.  DOI: 10.16097/j.cnki.1009-6744.2024.03.011 Abstract ( )   PDF (2807KB) ( )   In response to the issue of reduced green bandwidth after an increase in coordinated path numbers, this paper proposes a multi-path coordinated control optimization model considering path relationships. The relatedness of paths is analyzed, and models for partition of outbound and inbound and decomposing paths are constructed. Subgroup division parameters are introduced to improve the classical multi-path model, and intra-subgroup coordination constraints are established. The connection characteristics of paths between adjacent subgroups are analyzed, and inter-subgroup connectivity constraint conditions are defined. The weight coefficients considering the length and traffic of sub- paths are also incorporated in the model. The optimization objective is to maximize the weighted sum of green bandwidths of each subgroup. A case study is conducted on a typical main road in Nanjing City to validate the effectiveness of the model. The results show that the proposed model effectively increases the green band width, with a weighted green band width improvement of 49.44% compared to the benchmark. The application effect of the scheme is verified using VISSIM software, and the simulation results demonstrate that the proposed model scheme achieve better performance than the traditional methods. The arterial average delay is reduced by 20% and average number of stops is reduced by 27% on critical paths, while the average vehicle speed increased by 17% . The proposed model provides a theoretical basis for the coordination control of urban arterial. References | Related Articles | Metrics

Select

Coordinated Sequential Optimization for Network-wide Traffic Signal Control Based on Heterogeneous Multi-agent Transformer CHEN Xiqun, ZHU Yizhang, XIE Ningke, GENG Maosi, LV Chaofeng 2024, 24(3): 114-126.  DOI: 10.16097/j.cnki.1009-6744.2024.03.012 Abstract ( )   PDF (4128KB) ( )   Focusing on the complex traffic signal control task in an urban network, this study proposes a coordinated sequential optimization method based on a Heterogeneous Multi-Agent Transformer (HMATLight) to optimize network-wide traffic signals and improve the performance of signal control policy at intersections within the urban network. Specifically, considering the spatial correlation of multi-intersection traffic flow, a value encoder based on a self-attention mechanism is first designed to learn traffic observation representations and realize network-level communication. Secondly, in response to the non-stationary environment for multi-agent policy updates, a policy decoder based on the multi-agent advantage decomposition is constructed, which can sequentially output the optimal responsive action on the basis of the joint actions of preceding agents. Besides, an action-masking mechanism based on effective driving vehicles, adapting the decision frequency within the time-adequate interval, and a spatio-temporal pressure reward function considering the waiting fairness are constructed, which further enhance policy performance and practicality. A series of experiments are carried out on Hangzhou network datasets to validate the effectiveness of the proposed method. Experimental results show that the proposed HMATLight outperforms all baselines on two datasets with five metrics. Compared with the best-performed baseline, HMATLight decreases the average travel time by 10.89%, the average queue length by 18.84% and the average waiting time by 22.21%. Furthermore, HMATLight is dramatically higher in generalization and significantly reduces instances of long vehicle waiting times. References | Related Articles | Metrics

Select

Eco-driving Under Mixed Autonomy at Signalized Intersection: A Deep Reinforcement Learning Model XIN Qi, WANG Jiaqi, YANG Wenke , XU Meng, YUAN Wei 2024, 24(3): 127-139.  DOI: 10.16097/j.cnki.1009-6744.2024.03.013 Abstract ( )   PDF (3882KB) ( )   Dynamic programming model with eco-through constraint and safety constraint often causes computational inefficiency and even unfeasible solutions in mixed autonomy and heavy traffic conditions. This paper proposes an eco-driving-oriented and deep reinforcement learning based trajectory optimization model for Connected and Autonomous Vehicles (CAVs) in mixed autonomy. The model uses a compound reward reshaping and a twin delayed deep deterministic policy gradient algorithm to optimize CAV trajectories at the upstream of signalized intersection in mixed autonomy. The vehicular gap, speed difference, speed, distance to intersection, queue length, signal phasing and timing are selected as agent state to describe safety and driving mobility. The queue length is augmented in state representation to mitigate CAV halting possibility caused by queue of human driving vehicles. A multi-objective reward function is established based on agent state and anticipated arrival time at the intersection to optimize the CAV driving mobility, energy efficiency, comfortability, and safety. The proposed model performs better than the dynamic programing model in terms of decoupling the strong correlation between model constraints and computational complexity. The training and testing of the proposed model with simulation demonstrate that the vehicle delay at intersections significantly decreases with the increase of CAV penetration rate. Besides, the energy consumption relatively decreases by 5.47%, 4.42%, and 2.91%, compared to uncontrolled scenarios, dynamic programming-based trajectory optimization model, and deep deterministic policy gradient-based trajectory optimization model. In addition, the proposed model can ensure the CAV to cross the signalized intersection without stopping, and also show robustness against traffic demand and signal cycle. References | Related Articles | Metrics

Select

Constrained Adaptive Finite-iteration Learning Fault-tolerant Control for High-speed Train YU Qiongxia , HOU Yiteng , SUN Junjie , HOU Zhongsheng 2024, 24(3): 140-150.  DOI: 10.16097/j.cnki.1009-6744.2024.03.014 Abstract ( )   PDF (2063KB) ( )   This paper focuses on the speed control of high-speed train (HST) automatic operation system under actuator fault and speed limitation, and proposes a Finite-Iteration Constrained Adaptive Iterative Learning Fault-Tolerant Control (FI-CAILFTC) method. Based on the Barrier Composite Energy Function (BCEF), this paper defines the convergence conditions for a finite number of operations along the iterative domain, and calculates the required number of operations using the desired arbitrary tracking accuracy. The method then improves the controller parameter selection to ensure finite number of operation convergence. The iterative learning control algorithms are designed with adaptive fault tolerance for adaptive estimation and compensation of unknown time-varying and iteration-varying actuator faults. To address the overspeed issue of the HST operation, this study added an overspeed protection mechanism to ensure that the actual operation speed of the HST always meets the speed constraints and to ensure the safe operation of the train. The China Railway High-speed (CRH)-3 high-speed locomotive train is used as an example for the simulation analysis. The results show that the HST speed tracking error under the FI-CAILFTC method reaches the desired control accuracy of 0.2 after the pre-calculated 17th iteration, compared with the comparison algorithms, the control accuracy was improved by 90.70% and 90.22%, respectively. The FI-CAILFTC has faster convergence and better adaptive fault tolerance. The actual operation speed of the HST is always active to satisfy the speed constraints. Related Articles | Metrics

Select

Comprehensive Optimization of Line Planning, Ticket Pricing and Seat Allocation of High-speed Railway ZHOU Wenliang, JIANG Zhigang, CHAI Naijie, XU Guangming 2024, 24(3): 151-163.  DOI: 10.16097/j.cnki.1009-6744.2024.03.015 Abstract ( )   PDF (2317KB) ( )   In order to improve the organization and profitability of high-speed railway, this paper proposed a comprehensive optimization method of line planning, ticket pricing and seat allocation. First, the elastic demand function of time-dependent passenger demand and ticket pricing was constructed, and the broad travel cost of passengers was analyzed, including departure time deviation, travel time consumption and ticket pricing, so that a polynomial Logit model was built to describe the train choice behavior of time-dependent elastic passenger flow. Then, a comprehensive optimization model was constructed with the goal of maximizing the difference between the total ticket revenue and the operating cost. Second, the search strategy of ticket pricing was constructed by applying the partial derivation of OD revenue with respect to ticket pricing to make the ticket pricing neighborhood solution match the line planning neighborhood solution. In addition, the Cplex was used to solve the optimal seat allocation scheme, and the simulation annealing algorithm was designed to solve the model. Finally, the Zhengxi high-speed railway was used in a numerical experiment. The results show that under 7 different elastic coefficients, the passenger success travel rate and the average passenger load factor of the comprehensive optimization are above 90%, and the line planning, ticket pricing and seat allocation of the optimized solution are highly matched. In numerical experiments under 5 different scales, compared with the joint optimization of line planning and seat allocation under fixed ticket pricing and the joint optimization of ticket pricing and seat allocation under fixed line planning, the optimal net revenue of the three-factor comprehensive optimization increases by 4.11%~15.25% and 3.17%~13.42%, respectively, while the per capita unit mileage travel cost decreases by 1.69%~4.96% and 0.97%~4.35%, respectively. The results indicate that comprehensive optimization will better improve the operating revenue and passenger service level of high-speed railway. References | Related Articles | Metrics

Select

Energy-efficient Timetable Optimization for Urban Rail Transit Considering Difference of Peak and Off-peak Hours ZHANG Bonan, YAO Xiangming, ZHAO Peng, YANG Zhongping, YANG Jianguo 2024, 24(3): 164-171.  DOI: 10.16097/j.cnki.1009-6744.2024.03.016 Abstract ( )   PDF (2530KB) ( )   Reducing energy consumption is crucial for developing green, low-carbon, and sustainable rail transit. Based on the principle of "running fast during peak hours for transport capacity assurance, running slowly during off- peak hours for energy consumption reduction", this paper proposes an energy-efficient timetable optimization method based on different time criteria during peak and off-peak hours. First, an optimization model considering the characteristics of passenger service demand was developed based on the negative correlation between train running time and traction energy consumption. The objective was to reduce train traction energy consumption. Then, to address the uneven train service issue resulting from different time criteria during the transit period between peak and off-peak hours, this study developed an optimization model for train service to minimize the variability in train arrival intervals. The empirical analysis was conducted for Fuzhou Metro Line 1. The results demonstrate that the energy-efficient timetable based on different time criteria achieved a 12.56% reduction in traction energy consumption, while maintaining the transport capacity and the number of rolling stock unchanged. The proposed method is practical and valuable, providing operational managers with methods for energy-efficient timetabling. References | Related Articles | Metrics

Select

Optimization of Operation Scheme and Timetable for Interconnected Lines in Urban Rail Transit Considering Express-local Train WEN Fang, BAI Yun, CHEN Yao 2024, 24(3): 172-183.  DOI: 10.16097/j.cnki.1009-6744.2024.03.017 Abstract ( )   PDF (2405KB) ( )   The crossing-line operation has become an important trend in the development of urban rail transit, and express-local train is an effective way to reduce passenger travel time. Considering both crossing-line and express-local train services, this paper constructs an event-activity network to accurately depict the train movement and passenger behavior in urban rail transit. A train operation scheme and timetable collaboration model is developed to minimize passenger travel time and operation cost. In consideration of safety interval and train capacity constraints, the model determines train routes, stopping plan, frequencies, and departure and arrival times of all trains at each station. A customized variable neighborhood search algorithm is designed to efficiently solve the model. A numerical case study is then conducted to verify the proposed integrated optimization model and algorithm. The results indicate that when combining the crossing-line and express-local services, the operation cost is increased by 3%, while the number of transfer passengers is reduced by 82.5% and the average travel time of all passengers and crossing-line passenger respectively decreased by 1.3 and 5.3 minutes. There is a slight increase of the travel time for passengers who do not need to transfer, but the travel time of cross-line passengers has been significantly reduced. As the transfer walking time increases, the frequency of cross-line trains also increases, which results in a more significant reduction in passenger travel time. References | Related Articles | Metrics

Select

Joint Optimization of Train Timetabling and Rolling Stock Circulation Planning in Urban Rail Transit Line with Multiple Train Compositions RAN Xinchen, CHEN Jian, CHEN Shaokuan, LIU Gehui, ZOU Qingru 2024, 24(3): 184-193.  DOI: 10.16097/j.cnki.1009-6744.2024.03.018 Abstract ( )   PDF (2739KB) ( )   PDF(English version) (894KB) ( 2 )   To address the issues of peak-hour congestions and off-peak underutilization of transportation capacity on an urban rail line, a joint optimization method of train timetabling and rolling stock circulation planning with multiple train compositions is proposed. Based on dynamically changing OD passenger demand and multiple types of line resource, a two-objective optimization model is constructed to minimize the total passenger waiting time and the train operating cost. The total number of operating trains, the timetable, the train types, the entry and exit of trains from depots, and the train succession relationship are taken as decision variables. Timetable-related constraints, rolling stock circulation- related constraints, fleet size constraints, turnaround constraints, and train capacity constraints are considered in this model. Since the total number of trains is not determined, a NSGA-II (Non-dominated Sorting Genetic Algorithm-II) with variable-length chromosomes is designed to solve for the Pareto optimal solution of the twoobjective optimization model. A case study conducted on a subway line demonstrates the effectiveness of this modelling and solution approach. The results show that the optimized multi-train composition strategy simultaneously reduces the total passenger waiting time by 26.16% and the train operating costs by 25.75%. Moreover, the optimized average load factor of trains is increased by 1.3% ~9.6% , further improving the matching between transportation capacity and passenger flow demand. References | Related Articles | Metrics

Select

Metro Passenger Flow Prediction Model Using Adaptive Multi-view Fusion Graph Neural Network LU Wenbo, ZHANG Yong, LI Peikun, WANG Ting, CONG Yarong 2024, 24(3): 194-203.  DOI: 10.16097/j.cnki.1009-6744.2024.03.019 Abstract ( )   PDF (3046KB) ( )   To solve the problem of insufficient modeling of multi-view spatial interaction in metro stations by traditional methods, this study proposes an Adaptive Multi-view fusion Graph Neural Network Model (AMFGNN) to conduct spatial interaction modeling in metro stations short-term passenger flow prediction. In the spatial dimension, the model includes multiple partial views such as physical topology graph, line accessibility graph, spatial distance graph, etc., and uses the graph attention networks (GAT) to learn the dynamic spatial interaction within a single view. Taking the single-view station as the central node, combined with the station in other views as neighbor nodes, this paper constructs a fused view is and uses the GAT is to learn the dynamic interaction between multiple views. In the time dimension, the gated recurrent unit neural network is used to learn the time-varying characteristics of station passenger flow. The experiments were conducted in the Chongqing metro network, and the prediction results of the outbound passenger flow of the entire network show that compared with the physical virtual combined graph network model (PVCGN) in the baseline, the AMFGNN can reduce the average absolute error and root mean square error of the network's outbound passenger flow respectively by 3.06% and 2.49% . The visualization results of attention scores between nodes in multi-views graph show that the multi-view modeling based on the GAT can adaptively and effectively integrate station spatial information extracted from different views graph. In addition, the analysis of the impact factors of AMFGNN model performance show that using structurally stable views graph such as physical topology and line accessibility as central nodes to build a fusion view graph can obtain a more accurate and stable prediction model. References | Related Articles | Metrics

Select

Electric Vehicle Centralized Charging Station Siting and Capacity Modeling YANG Yazao, BIN Tao 2024, 24(3): 204-212.  DOI: 10.16097/j.cnki.1009-6744.2024.03.020 Abstract ( )   PDF (1902KB) ( )   The existing electric vehicle charging station have some problems such as blind construction of single mode for charging and exchanging, low resource utilization rate, and part of the charging demand cannot be met due to the lack of centralized planning. This paper proposes a centralized charging station siting and capacity selection method which considers multiple charging modes such as charging, exchanging and mobile charging. First, the charging behavior characteristics of different types of electric vehicles are analyzed, and the charging demand distribution in the planning area is simulated. Based on the principle of nearest distance, the queuing theory method is used to calculate the charging loss cost of electric vehicles, optimize the self-charging moments of mobile charging equipment, and obtain the self-charging cost of electric vehicles. Then, a site selection and capacity model is developed with the objective function of minimizing the total sum of maintenance cost, user loss cost and equipment self-charging cost. A genetic algorithm is used to solve the model to determine the number of centralized charging stations in the planning area, their locations and the number of different equipment configurations, combining with part of the actual road network of a city as the study area. The results show that: the total cost of constructing eight centralized charging stations in the planning area are the lowest. The number of vehicles with charging demand and the change of charging power have a large impact on the cost of centralized charging stations. The optimization of self-charging scheduling management of mobile charging equipment can reduce the grid load of centralized charging stations in the peak period by 32.62%, which improves the stability of the power grid. References | Related Articles | Metrics

Select

Traffic State Division of Urban Expressway Driven by Multi-source Data Fusion GU Yuanli, DU Heng, LU Wenqi 2024, 24(3): 213-220.  DOI: 10.16097/j.cnki.1009-6744.2024.03.021 Abstract ( )   PDF (2158KB) ( )   To enhance the effectiveness of traffic state division, this paper proposes an improved fuzzy C-means clustering model based on Negative Incentive Terms (BNIT-FCM). Building upon the original FCM model, the BNITFCM considers the impact of the weight of traffic flow sample points and traffic flow parameters on the clustering. It introduces negative membership incentives, traffic flow weight amplification incentives, and traffic flow sample point weight amplification incentives to foster high intra-class coherence and low inter-class coherence in clustering results. Furthermore, the model introduces weighted sample points and employs weighted Euclidean distance to depict sample point relationships. Iterative formulas are derived via the Lagrange multiplier method and solved iteratively. To address the issue of low dimensionality in most traffic state division methods, this paper constructs high- dimensional feature inputs using parameters such as speed, speed standard deviation, flow, density, and road capacity obtained through multi-source data fusion. The classification accuracy of the BNIT-FCM model is evaluated through numerical simulation experiments. Results demonstrate that compared to the FCM model and Improved Fuzzy Membership FCM model (IFMD-FCM), the ARI of the BNIT-FCM model improves by 4.17% and 3.56% respectively. Using traffic flow data from both bayonet and floating cars on the North Ring Road in Shenzhen, experimental findings reveal that the silhouette coefficients of the BNIT-FCM model improve by 4.12% and 4.07% respectively compared to the FCM model and IFMD-FCM model. Additionally, utilizing multi-source fusion data, the speed and standard deviation of the BNIT-FCM model exhibit increases of 29.67% and 54.13% respectively compared to using bayonet data and floating car data alone. References | Related Articles | Metrics

Select

Identification of Vehicle Interaction Risk in Short Weaving Areas of Expressways Based on Driving Risk Field HU Liwei, CHEN Chen, ZHAO Xueting, LIU Bing, HOU Zhi, ZHANG Ruijie, HE Yu 2024, 24(3): 221-231.  DOI: 10.16097/j.cnki.1009-6744.2024.03.022 Abstract ( )   PDF (3033KB) ( )   Traditional risk identification models cannot continuously identify the interaction risk between vehicles when changing lanes. In order to accurately and intuitively identify the interaction risk between vehicles driving in short weaving sections of expressways, this study first collects vehicle trajectory data in the short weaving area of expressways by using drones and Tracker and filters out pairs of following vehicles and pairs of lane changing vehicles. By considering the vehicle area, the risk difference between the front and rear of the vehicle when moving forward, the risk difference between the left and right when turning, and the lateral distance between vehicles, the existing driving safety field model DRF (Driving Safety Field) is adaptively improved, and the model parameters are calibrated using a genetic algorithm and the Polankov model. To verify the effectiveness of the model and calibrated parameters, the improved driving risk field model is compared with the reciprocal of headway (THWI) and the reciprocal of collision time (TTCI) in identifying the following interaction risk and lane changing interaction risk between vehicles, and verify the effectiveness of the model. The improved driving risk field model is compared with the reciprocal of headway (THWI), the reciprocal of collision time (TTCI), and the driver's recognition of the risk of following and lane changing interactions between vehicles. The results show that the proposed model consistent with the driver's driving psychology better compared to THWI and TTCI, and perceives changes in risk before the driver. The recognition rate of vehicle lane changing interaction risk is increased by 52.45% compared to THWI and 83.66% compared to TTCI. This model performs better in identifying lane changing interaction risk. Finally, based on the proposed model, the risks generated by the joint action of multiple vehicles in the short weaving area can be visualized, which can assist traffic management departments in identifying key areas that require refined organization, and can also serve as a visualization tool for evaluating the effectiveness of management measures for improvement. References | Related Articles | Metrics

Select

Traffic Organization Methods for Parallel Expressway Exits CHEN Yongheng, LI Shihao, YANG Suicheng, LIU Bo, ZHAN Tianshu 2024, 24(3): 232-239.  DOI: 10.16097/j.cnki.1009-6744.2024.03.023 Abstract ( )   PDF (2137KB) ( )   Aiming to address the traffic congestion issue of queuing overflow at the exit of parallel expressways, this paper proposes five traffic organization methods for such exits under conditions of no signal control and signal control. By considering the headway distribution of both the main and auxiliary traffic flows, a calculation model for the delay of these flows is derived. From a perspective of scheme selection, with the average delay of vehicles as the evaluation index, the most suitable traffic organization method is examined at the exit of parallel expressways under different traffic flow conditions. The results indicate that when the combined flow of the main road outflow and the auxiliary road traffic of the parallel expressway is below 1500 pcu · h-1 , the no-signal control method is recommended. When the traffic volume of the main road outflow is high, giving priority to the main road outflow is recommended; similarly, when the traffic volume of the auxiliary road is high, giving priority to the auxiliary road is recommended. For total flows are [1500, 2300] pcu · h- 1 , a combination of auxiliary road signaling and yield control is recommended. Specifically, when the traffic volume of the main road outflow is substantial, implementing auxiliary road signaling with main road priority is recommended. Conversely, when the traffic volume of the auxiliary road is significant, adopting auxiliary road signaling with auxiliary road priority is preferable. For total flows exceeding 2300 pcu·h-1 , full signaling for both the main and auxiliary roads is advocated. References | Related Articles | Metrics

Select

Airport Terminal Departure Aggregation Passenger Flow Prediction Considering Flight Delay Characteristics LI Mingjie, WANG Tao, HUANG Xinning, TIAN Jie, YAO Linhao 2024, 24(3): 240-254.  DOI: 10.16097/j.cnki.1009-6744.2024.03.024 Abstract ( )   PDF (4345KB) ( )   To improve the accuracy and efficiency of predicting passenger flow for resource planning and passenger management within terminal buildings during flight delays, this paper proposes a departure aggregation passenger flow prediction method considering flight delay characteristics. The method introduces the flight delay parameters to quantitatively characterize the fluctuation of the departure aggregation passenger flow in the airport terminal (DAPFT). The fluctuation pattern and distribution characteristic of departing aggregation passenger flow are analyzed under flight delay. A short-term terminal aggregation passenger flow prediction model is proposed based on adaptive noise complete ensemble empirical modal decomposition (CEEMDAN), permutation entropy algorithm (PE), and whale optimization algorithm (WOA) optimised long- short- term memory neural network (LSTM). The CEEMDAN is applied to decompose the aggregation passenger flow data series into several modal components intrinsic mode function (IMF) and a residual Res to reduce the complexity and non-stationarity of the data of the original series. To reduce the computational scale of the model and improve the prediction efficiency and accuracy at the same time, the PE algorithm is used to calculate the entropy value of each IMF component and reconstruct the components based on the entropy value. Then, the WOA-LSTM (W-L) passenger flow aggregation prediction model is established, the whale optimization algorithm is used to optimize the LSTM's hyperparameters, and the reconstructed components' predictions are superimposed to obtain the final aggregation passenger flow prediction target value. The model has been applied to a hub airport in the Yangtze River Delta for predictive performance validation. The results show that the CEEMDANPE-WOA-LSTM (C-P-W-L) prediction model has the best performance, and compared with the simple LSTM model, the root mean square error is reduced by 42.78%, the average absolute error is reduced by 44.00%, and the percentage error of the prediction of departure hall aggregation passenger flow(DHAPF), is reduced by 45.62%. The prediction efficiency is improved by 41.64% compared with the CEEMDAN-WOA-LSTM (C-W-L) model. The proposed model can effectively fit the departure hall aggregation passenger flow data with significant nonlinear and non- stationary characteristics, and has high prediction accuracy and computational efficiency. References | Related Articles | Metrics

Select

Traffic Equilibrium of Automated Container Terminal Road Network Based on Potential Game XU Bowei, FAN Huiyao, LI Junjun 2024, 24(3): 255-264.  DOI: 10.16097/j.cnki.1009-6744.2024.03.025 Abstract ( )   PDF (2607KB) ( )   To improve the utilization rate and traffic equilibrium of the automated container terminals road network, this paper focuses on the distributed automated guided vehicle (AGV) traffic scheduling problem and proposes a game model for the AGVs scheduling based on the factors of the congestion degree of each sub-path of the automated container terminal, the velocity of the AGV, and the selection probability of each sub-path. The balanced scheduling strategy of AGV is solved by two algorithms. The distributed route algorithm is designed to help AGV update strategies. The information update algorithm of the system is used to update road information and select the AGV for the strategy update. The numerical simulation was carried out with three actual ships. The results show that the distributed route method has better performance than the traditional route selection method. The quantity of ship tasks, the number and velocity of AGVs are the key factors that affect the efficiency of AGV transportation. When the AGV velocity is 6 m·s -1 , the total revenue of the ship is the largest, which provides decision support for the terminal manager to determine the AGV scheduling strategy References | Related Articles | Metrics

Select

Load Balancing Optimization of Logistics Hub Network Based on User's Preference Behavior LIU Xinquan , WANG Xinyu, HUANG Yingyi 2024, 24(3): 265-276.  DOI: 10.16097/j.cnki.1009-6744.2024.03.026 Abstract ( )   PDF (2322KB) ( )   Considering the effects of different size discount policies and distances on users' hub selection preferences, this paper proposes a workload balancing design method for logistics transit hub networks based on users' finite rational preference behavior to address the hub load imbalance caused by users' preference behavior in logistics hub networks. First, a polynomial Logit rule with constraints under incomplete information is used to simulate users' preference behavior under limited rationality, and a multi-objective optimization model is constructed to minimize the generalized cost and maximize the temporal utility, including hub low load utilization and congestion penalty cost, with hub location and transportation routes as decision variables. To address this problem, a hybrid algorithm framework is developed, which first divides the allocation decision space according to the Leuven algorithm to reduce the difficulty of the solution, and then adds multiple population mechanism and cooperative search strategy based on the nondominated genetic algorithm (NSGAⅡ) to improve the convergence ability of the algorithm. At last, the effectiveness of the model and algorithm is verified by taking Guangxi logistics network as an example. The results show that the load of the hub network is more balanced under the fully rational condition, and the user's preference behavior will aggravate the phenomenon of hub load imbalance, leading to the increase of the generalized cost and time consumption of the hub network. Considering the preference behavior of customers under finite rationality, the load balancing capability of the entire hub network under the active discount scheme performs better than that of the general discount scheme and close to the load balancing capability of the fully rational state of users. The average load ratio and average congestion rate are respectively 60.17% and 34.40%. With the consideration of the user preferences, the load ratio of the hub increases with the increase of discounts, which also makes the congestion rate increase. The designed hybrid evolutionary algorithm converges to a more balanced objective value, exhibits strong search and optimization performance, and can obtain an effective solution to the model. References | Related Articles | Metrics

Select

Collaborative Driving Decision-making Method of Unmanned Mining Trucks in Open-pit Mine Operation Areas NI Haoyuan, YU Guizhen, LI Han, CHEN Peng, LIU Xi, WANG Wenda 2024, 24(3): 277-289.  DOI: 10.16097/j.cnki.1009-6744.2024.03.027 Abstract ( )   PDF (3340KB) ( )   The long parking and waiting time of unmanned mining trucks in open-pit mines during transportation in the loading and unloading operation area is a bottleneck that restricts the efficiency improvement of unmanned transportation systems in open-pit mines. To improve the transportation efficiency of unmanned mining trucks, this paper combines the transportation operation process in the operation area and proposes a multi-vehicle collaborative driving decision-making method based on dynamic travelable distance. The decision-making model was formulated as a mixed integer linear programming (MILP) model to express the optimization objective and problem constraints. Considering the challenge of meeting real-time decision-making requirements in solving the MILP model, the multivehicle conflict resolution was implemented based on Monte Carlo tree search (MCTS). The core idea was to use the derivation capability of the search tree to conduct forward simulation of multi- vehicle driving, calculate the optimal driving priority of multi-vehicle, and thereby dynamically adjust the travelable distance of multi-vehicle. In addition, different MCTS node value functions were designed based on the operating characteristics of unmanned mining trucks in the operation area to achieve driving priority ranking that comprehensively considered transportation efficiency and operating characteristics. A multi- vehicle driving simulation experiment was designed in the scenario of 4, 8, and 12 parking spots in the operation area. Compared with the method based on first-come-first-served (FCFS), the throughput was increased by 22.03% to 28.00% and the average parking waiting time was shortened by 31.71% to 50.79% . In addition, a 6-parking spots operation area scenario experimental platform for miniature intelligent vehicles was built. The total multi-vehicle single-operation time was reduced by 18.84% compared to the FCFS. The results of simulation and miniature intelligent vehicles experiments indicated that the proposed method could enhance the efficiency of multi-vehicle transportation in open-pit mine operation areas. References | Related Articles | Metrics

Select

Impact of Non-driving Tasks on Drivers' Visual and Physiological Characteristics Under Take-over Scenario ZHANG Lei , PENG Jinshuan , CHEN Xiaoli 2024, 24(3): 290-298.  DOI: 10.16097/j.cnki.1009-6744.2024.03.028 Abstract ( )   PDF (2796KB) ( )   To investigate the changes in drivers' visual and physiological characteristics during the process of take-over vehicle control, this paper analyzes six types of take-over scenarios for urban roads and highways conditions. The driving experiments were conducted and the parameters were extracted from various dimensions, such as visual features, electro cardio graphic features, and electro dermal activity features. The two-way ANOVA was used to identify the factors affecting drivers' state parameters. The results show that the type of take-over scenario and nondriving tasks significantly affect drivers' pupil area changing rate, fixation entropy, fixation probability in areas of interest, heart rate growth, heart rate variability, and electro dermal activity growth (p < 0.05) . In urban road conditions, the drivers' pupil area changing rate (M = 26.91, SD = 10.17) is higher than on highways (M = 21.32, SD = 7.69) . The fixation entropy (M = 3.84, SD = 1.53) during cognitive distraction state is lower than in baseline state (M = 4.46 , SD = 1.87) , and there is a significant reduction in visual search range (p < 0.05) . Compared to baseline and cognitive distraction states, drivers' heart rate growth increased by 34.69% in composite distraction state, while heart rate variability is reduced by 10.55%. Additionally, in composite distraction state, drivers exhibit the highest electro dermal activity growth, surpassing the other two non-driving task by 82.43%. The research results provide important references for the evaluation and improvement of driver take-over performance and the optimization of humancomputer interaction mode of automated driving. References | Related Articles | Metrics

Select

Accurate Detection Method of Small-scale Vehicles from Perspective of Unmanned Aerial Vehicle High-altitude Aerial Photography ZHANG Heshan, TAN Xin , FAN Mengwei, PAN Cunshu, XU Jin, ZHANG Yu 2024, 24(3): 299-309.  DOI: 10.16097/j.cnki.1009-6744.2024.03.029 Abstract ( )   PDF (4019KB) ( )   The proportion of vehicle pixels in high-altitude aerial images taken from unmanned aerial vehicle (UAV) is low, and there are limited visual information of can be identified for the targets, which result in missed or false detection in the detection tasks. This paper proposes a detection method of small-scale vehicles based on improved YOLOX (You Only Look Once X) from the perspective of high-altitude aerial photography. First, to enhance the network 's ability to extract low-level features, this study added a shallow feature extraction network of 160 pixel× 160 pixel to the original YOLOX prediction head. Then, a Normalization-based Attention Module (NAM) was embedded after the backbone network to suppress redundant non-significant feature expression. To increase the relative pixel ratio of small-scale vehicles and improve the ability of the network to capture effective feature information, an image segmentation detection method was proposed based on sliding window. The experiment results show that the improved YOLOX network shows good detection performance, and the detection accuracy reaches 84.58%, which is better than the typical target detection network Faster R-CNN (79.95%), YOLOv3 (83.69%), YOLOv5 (84.31% ), and YOLOX (83.10% ). In addition, the improved YOLOX can effectively solve the problem of missed detection and false detection of small-scale vehicles in high-altitude aerial images of UAV, and the prediction box is more suitable for the actual contour of the vehicle. At the same time, it has high robustness in target detection tasks at different aerial heights. References | Related Articles | Metrics

Select

Collaborative Optimization of Gate Lane and Yard Crane Configuration Considering Dual Transaction Trucks Priority DIAO Cuijie, WANG Wenmin, CAI Jiaxin, JIN Zhihong, GUO Shujuan 2024, 24(3): 310-322.  DOI: 10.16097/j.cnki.1009-6744.2024.03.030 Abstract ( )   PDF (2834KB) ( )   Dual transaction external container trucks deliver one container and pick up another in a single trip, increasing the efficiency of terminal collection and distribution, and thus should be given higher priority. In order to investigate the effect of prioritizing dual transactions trucks on gate lane and yard crane configuration, the dual transactions truck priority channel is introduced at the gate, and a three-level priority queuing system is established in the yard. The three-level queuing system prioritizes internal truck operations, followed by dual transactions external trucks and then single transaction external trucks. The queuing model for trucks at the terminal gate and in the yard is formulated. Based on the queuing model, a bi-objective mixed-integer programming model is developed, with the objective of minimizing the number of gate lanes activated and yard cranes configured and minimizing the number of external container trucks queuing. A multi-objective evolutionary algorithm based on decomposition, with pointwise stationary fluid flow approximation, is proposed. Numerical analysis demonstrates the effectiveness of the method. For the same terminal resource configuration, the average number of dual transaction trucks queuing at the gates and at the yard is decreased by 21.82% and 18.27% after considering the priority of dual transaction trucks. References | Related Articles | Metrics

Select

A Trajectory Segmentation and Analysis Method for Non-motor Vehicle Overtaking Behavior Recognition ZHANG Rui , DUAN Yu, KONG Lingzheng, HOU Xianlei 2024, 24(3): 323-332.  DOI: 10.16097/j.cnki.1009-6744.2024.03.031 Abstract ( )   PDF (2400KB) ( )   The number of overtaking by non-motor vehicles is crucial for evaluating the level of services. With the widespread availability of roadside video equipment, it provides valuable video data for recognizing non-motor vehicle overtaking behavior. However, existing research on trajectory segmentation is difficult to adapt to the frequent changes in driving angles and lengthy overtaking trajectories exhibited by non-motor vehicles. In this paper, non-motor vehicle trajectories are segmented into fixed-length segments (trajectory segment length) based on specific interval time (segmentation interval). Using data from 640 overtaking trajectories captured from videos, various overtaking behaviors of non-motor vehicles are analyzed. Key characteristic parameters such as average longitudinal velocity, longitudinal velocity standard deviation, average lateral velocity, and lateral velocity standard deviation are selected. A classification model is then established using the K-Nearest Neighbor(KNN) algorithm to classify non-motor vehicle overtaking behaviors. Based on the average duration of overtaking behaviors, trajectory segment lengths for bicycles range from 4 to 9 seconds with segmentation intervals from 1 to 8 seconds, while for electric bicycles, trajectory segment lengths range from 4 to 8 seconds with segmentation intervals from 1 to 7 seconds. The classification model is then employed to identify overtaking behavior in non- motor vehicle trajectory segments for each parameter combination. Results indicate that for bicycles, a trajectory segment length of 8 seconds and a segmentation interval of 6 seconds yield the lowest recognition error rate and the most effective segmentation. Similarly, for electric bicycles, a trajectory segment length of 6 seconds and a segmentation interval of 4 seconds achieve optimal performance. References | Related Articles | Metrics