Current Issue

• 2024, Vol. 15 No. 5
  Published on:31 October 2024 Previous issue   
  Review, Progress and Prospects

  Future of autonomous driving: Single autonomous driving and intelligent vehicle-infrastructure collaboration systems

  LIU Yang, ZHAN Jiahao, LI Shen, LI Xiaopeng, CHEN Jun

  2024, 15(5):  611-633.  doi:10.3969/j.issn.1674-8484.2024.05.001

  Abstract ( 77 )   HTML ( 9)   PDF (2377KB) ( 53 )  

  As global traffic congestion and safety concerns become increasingly prominent, the widespread application of autonomous driving technology is considered a vital solution. Two prominent areas of research in autonomous driving are single autonomous driving (SAD) and intelligent vehicle-infrastructure collaboration systems (i-VICS). This paper explores the fundamental concepts and critical technologies of both. In terms of SAD, the focus is on perception, localization, decision-making, planning, and control execution, while i-VICS is centered on cooperative perception, collaborative localization, vehicle-to-infrastructure communication, and hierarchical cloud control. Furthermore, it reviews the progress of research in these technologies and summarizes the development paths chosen by China, the United States, Germany, and Japan. The transformative impact of these technologies on the commercial and industrial supply chains is also examined. Finally, the paper analyzes the technical challenges faced by both SAD and i-VICS, along with the social and legal challenges of autonomous driving, offering insights into future development directions, and providing a reference for the innovation and application of autonomous driving technology.

  Figures and Tables | References | Related Articles | Metrics

  Review on the integrated capacity of transportation and power networks

  WU Tong, HUANG Kai, LIU Zhiyuan, JIANG Wei

  2024, 15(5):  634-649.  doi:10.3969/j.issn.1674-8484.2024.05.002

  Abstract ( 39 )   HTML ( 4)   PDF (1700KB) ( 13 )  

  Electric vehicles, as the core of transportation electrification, play an active role in reducing greenhouse gas emissions and improving energy efficiency. The significant growth in electric vehicle ownership and market share has impacted charging infrastructure, highlighting issues such as inadequate charging facilities and fluctuations in grid load. This paper provides a comprehensive review of the fundamental concepts, calculation methods, and assessment metrics related to transportation network capacity and power network capacity. It analyses the evaluation methods for the integration of transportation and energy networks and the resilience of their convergence. The paper explores potential challenges and strategies for transportation-power integration systems. It identifies urgent research gaps and outlines future research directions, aiming to optimize the efficiency of charging infrastructure, alleviate traffic congestion, and ensure the stable operation of the power grid.

  Figures and Tables | References | Related Articles | Metrics

  Intelligent Driving and Intelligent Transportation

  Supplementary capture using unmanned ground vehicle for 3D reconstruction model improvement

  YANG Donghui, WANG Yuhao

  2024, 15(5):  650-659.  doi:10.3969/j.issn.1674-8484.2024.05.003

  Abstract ( 29 )   HTML ( 2)   PDF (5790KB) ( 19 )  

  A 3D model improving method based on supplementary capture by unmanned ground vehicle was proposed to address the issue of damage and holes in 3D reconstruction models generated from images captured solely by unmanned ground vehicle. This method combined model resolution, triangular mesh structure, and manual point selection to extract areas needing improvement, generated 3D bounding boxes and normal vector information, and utilized heuristic methods to generate supplementary viewpoints. The results show that, under this method's optimization, the low-quality areas of the rough 3D model are significantly improved, with an average reduction of 66% in model projection pixel size. Therefore, this method effectively enhances the quality of 3D model reconstruction, providing a reliable solution for large-scale, detailed outdoor 3D reconstruction.

  Figures and Tables | References | Related Articles | Metrics

  Research on CAN bus anomaly detection of intelligent networked vehicle based on improved GAN

  YANG Haoran, XIE Hui, SONG Kang, YAN Long

  2024, 15(5):  660-669.  doi:10.3969/j.issn.1674-8484.2024.05.004

  Abstract ( 39 )   HTML ( 1)   PDF (2313KB) ( 25 )  

  A novel Controller Area Network (CAN) bus anomaly detection algorithm characterized by its adaptability to low anomaly traffic and strong generalization capability was proposed to enhance the safety of Intelligent Connected Vehicles (ICVs). The algorithm aimed to address potential and hard-to-detect abnormalities that may arise in vehicles, significantly improving the detection accuracy of anomalous data. This study explored the theoretical significance of Generative Adversarial Networks (GANs) and collected four types of attack data and two types of rare alarm data from an intelligent connected bus. The anomaly degree was assessed based on the reconstruction error of the computed data to validate the algorithm's adaptability. The results show that the proposed algorithm achieves an F1 score of 98.31% and a false positive rate of 2.90% on the low-traffic dataset Data4, surpassing the baseline model and the Deep Convolutional GAN (DCGAN) algorithm. Moreover, the false positive rate for rare alarm data is reduced to 3%, indicating that the algorithm is well-suited for low-traffic anomaly detection and exhibits strong generalization capabilities.

  Figures and Tables | References | Related Articles | Metrics

  Intelligent inspection method for power transmission towers, substations, and distribution poles using fixed UAV nests

  HUANG Zheng, WANG Hongxing, DU Biao, GAO Song, GAO Feng

  2024, 15(5):  670-679.  doi:10.3969/j.issn.1674-8484.2024.05.005

  Abstract ( 28 )   HTML ( 1)   PDF (1823KB) ( 13 )  

  To achieve automated inspection of power transmission towers, substations and distribution poles, a fixed unmanned aerial vehicle (UAV) nest was proposed based strategy that accounts for varying inspection frequencies. A fixed UAV nest deployment model was established based on a set cover problem, and the task assignment model for inspections was developed by enhancing the k-means clustering algorithm. The UAV path planning problem was formulated as a Multi-trip Traveling Salesman Problem with Time Windows (MTSPTW), and solved with an Adaptive Large Neighborhood Search (ALNS) algorithm. The real-world data validation was verified by utilizing a real operational and maintenance environment as an example. The results show that a single UAV nest completes 1 838 mixed inspection tasks over one month, with 130 takeoffs and a total flight distance of over 703 km. The proposed method overcomes the limitations of single-type inspections, proving effective for large-scale scenarios.

  Figures and Tables | References | Related Articles | Metrics

  Traffic flow parameter prediction method based on dynamic graphs self-attention

  SHI Tianjing, LI Xu

  2024, 15(5):  680-688.  doi:10.3969/j.issn.1674-8484.2024.05.006

  Abstract ( 23 )   HTML ( 1)   PDF (1786KB) ( 16 )  

  In order to improve the driving efficiency and safety of intelligent vehicles in the areas with frequent abnormal events, a traffic flow parameter prediction method was designed based on dynamic node self-attention to improve the accuracy of traffic flow parameter prediction. The spatial attention was used to aggregate the features of neighborhood nodes in multiple time steps, and then the traffic parameters were predicted by the temporal attention mechanism along the time dimension. The results show that the Mean absolute error (MAE), root mean square error (RMSE) and mean absolute percentage error (MAPE) of the one-hour prediction of Dynamic Picture Self-Attention (DGSA) model decrease by 3.75%, 3.45% and 11.63%, respectively. The simulated road average collision time (TTC) is longer, reaching 2.8 s. The proposed method can effectively predict the evolution trend of traffic flow and improve the safety of vehicles under abnormal events.

  Figures and Tables | References | Related Articles | Metrics

  A multimodal trajectory prediction method of pedestrians at signalized intersections for autonomous vehicles

  QU Guangyue, YANG Lan, YUAN Meng, FANG Shan, LIU Songyan

  2024, 15(5):  689-701.  doi:10.3969/j.issn.1674-8484.2024.05.007

  Abstract ( 36 )   HTML ( 1)   PDF (3487KB) ( 18 )  

  A multi-modal trajectory prediction method of pedestrians at signalized intersections for autonomous vehicles was proposed to improve the driving safety of autonomous vehicles in the mixed traffic with pedestrian and vehicles. Firstly, considering the social attributes of the Social Generative Adversarial Network (SGAN) model, the pedestrian history trajectory was taken as the model input, the generator and discriminator were trained alternately, and the cross-entropy loss function was used to optimize the model, and then a pedestrian trajectory prediction model based on SGAN was proposed. Secondly, four binding force models based on pedestrian self-drive, pedestrian interaction, zebra crossing boundary force and traffic light force were established, and then a pedestrian trajectory prediction model based on Social Force Model (SFM) was proposed. The particle swarm optimization algorithm was used to calibrate the non-measurable parameters of SFM. Finally, based on the AdaBoost algorithm, the prediction results of SGAN and SFM were fused, and the weights of each model were iteratively trained and optimized dynamically by multiple weak learners to improve the prediction accuracy of the model. Based on the pedestrian data of an intersection in Xi'an city, the experimental analysis and verification were carried out. The results show that the average displacement error (ADE) and final displacement error (FDE) of the proposed method are increased by about 21.7% and 10.5%, respectively, compared with the single SFM model and the single SGAN model. The proposed model can realize more accurate pedestrian trajectory prediction.

  Figures and Tables | References | Related Articles | Metrics

  Vehicle longitudinal speed planning based on deep reinforcement learning CLPER-DDPG

  LIU Peng, ZHAO Kegang, LIANG Zhihao, YE Jie

  2024, 15(5):  702-710.  doi:10.3969/j.issn.1674-8484.2024.05.008

  Abstract ( 30 )   HTML ( 1)   PDF (1793KB) ( 13 )  

  To solve the problems of planner convergence difficulty in vehicle longitudinal speed planning and stability issues during scenario transitions, a planner was designed using a multilayer perceptron, incorporating the Deep Deterministic Policy Gradient (DDPG) algorithm with Prioritized Experience Replay (PER) and Curriculum Learning (CL). The simulation scenarios were designed for model training and testing, as well as comparative experiments among the three algorithms of DDPG, DDPG with Prioritized Experience Replay (PER-DDPG), and DDPG with both Prioritized Experience Replay and Curriculum Learning (CLPER-DDPG). Real-vehicle experiments were also carried out on actual roads within the Park. The results show that the CLPER-DDPG algorithm, comparing with the DDPG algorithm, the convergence speed of the planner is improved by 56.45%, the mean distance error is reduced by 16.61%, the mean speed error is decreased by 15.25%, and the mean jerk is lowered by 18.96%. Furthermore, when the parameters of environmental conditions and sensor hardware in the experimental scenarios are changed, the model could ensure that the longitudinal speed planning task will be completed safely.

  Figures and Tables | References | Related Articles | Metrics

  Multi-objective bi-level programming model for optimization design of bus-HOV lanes

  YAO Ronghan, XU Wentao, LIN Zijing, WANG Libing

  2024, 15(5):  711-722.  doi:10.3969/j.issn.1674-8484.2024.05.009

  Abstract ( 20 )   HTML ( 2)   PDF (1838KB) ( 12 )  

  A multi-objective bi-level programming model for optimization design of bus-HOV (bus and high occupancy vehicle) lanes was formulated to enhance the utilization of exclusive bus lanes. The upper level of the model aimed to minimize the total travel impedance, transit operating costs and vehicle emissions; its lower level realized the user equilibrium of a multi-modal transportation network considering non-HOV, HOV and bus. The non-dominated sorting genetic algorithm Ⅱ and the methods of successive average and successive weight average were used to solve the model and lower level, respectively. The model and its solution algorithms was validated with the Nguyen-Dupuis network. The results show that the given algorithm can effectively obtain the Pareto optimal solution set of the setting scenario of the bus-HOV lanes; the reasonable setting scenario of the bus-HOV lanes can effectively promote carpooling; compared with not setting the bus-HOV lanes, permitting all and a part of high occupancy vehicles to enter the bus-HOV lanes makes the total travel impedance decrease by -0.007%~1.088% and 0.038%~4.493%, makes the transit operating costs ascend 1.057%~3.864% and 4.011%~5.611%, and makes the vehicle emissions reduce -7.598%~-1.111% and -0.677%~3.526%, respectively; when the passenger volume of original destination (OD) pairs is not lower than its critical value, setting the bus-HOV lane is helpful to increase the bus passenger volume.

  Figures and Tables | References | Related Articles | Metrics

  Research on carpooling demand prediction study based on machine learning

  WANG Di, LI Ying, HU Yujiao, SUN Haocheng

  2024, 15(5):  723-731.  doi:10.3969/j.issn.1674-8484.2024.05.010

  Abstract ( 33 )   HTML ( 1)   PDF (1308KB) ( 18 )  

  In order to improve the accuracy of carpooling demand prediction, thereby enhancing the efficiency of ride hailing services and effectively alleviating traffic congestion, a regional carpooling probability prediction model was proposed by optimizing the traditional decision tree machine learning model using time feature extraction and Kepler optimization algorithm. An experiment was conducted to predict carpooling demand based on the Chicago ride hailing probability dataset, and the model was compared with traditional decision tree models. The experimental results show that the optimized model outperforms traditional decision tree models in terms of prediction accuracy, with a decrease of 0.044 in mean absolute error (MAE) and 0.054 in root mean squared error (RMSE). The optimized model has higher accuracy in predicting carpooling demand compared to traditional decision tree models.

  Figures and Tables | References | Related Articles | Metrics

  Adaptive federated learning algorithm for differential intersection based on 3DSSD

  SHI Liying, ZHOU Guofeng, LI Zexing, CAO Liling

  2024, 15(5):  732-741.  doi:10.3969/j.issn.1674-8484.2024.05.011

  Abstract ( 23 )   HTML ( 2)   PDF (2672KB) ( 7 )  

  A point-cloud object-detection algorithm (FLA3DSSD) was proposed based on a parameter adaptive Federated Learning (FL) strategy to solve the problems of lack of roadside endpoint cloud dataset and the low generalization ability of object detection models. The point-cloud based 3-D Single-Stage ObjectDetector algorithm (3DSSD) was combined with the FL in the cases in which the data from various roadside clients are not interconnected. The client model parameter update strategy was improved by uploading local models to the server for model adaptive parameter fusion, to achieve data information sharing. The results show that the aggregation model combining classical federated learning and 3DSSD algorithm showed a 5%~40% increase in detection Average Precision (AP) compared to the locally trained model deployed directly to other clients for testing in the deployment task of the vehicle road collaborative differential intersection scene algorithm; Improved parameter adaptive federated learning FLA3DSSD achieves a 1%~7% increase in AP value based on the aggregated model. Therefore, the method improves the generalization ability and detection accuracy with protecting data privacy.

  Figures and Tables | References | Related Articles | Metrics

  Formation control of vehicle clusters in multi-lane scenarios based on Multi-Agent System consensus

  JI Pengxiao, KONG Weiwei, LUO Yugong, YU Jie, LIU Yanbin, WANG Junjie, ZHU Weizhen, LIANG Weiming

  2024, 15(5):  742-752.  doi:10.3969/j.issn.1674-8484.2024.05.012

  Abstract ( 21 )   HTML ( 1)   PDF (5893KB) ( 8 )  

  A control strategy with multi-lane vehicle platoon formation was investigated for Connected-Automated-Vehicles (CAV) based on Multi-Agent System (MAS) consensus to solve the problem of safe and efficient formation control of multi-lane, randomly distributed vehicle clusters. A platoon's desired geometric topology was established based on an interleaved structure for the vehicle cluster with multiple lanes and scattered random distribution. A three-stage formation control process was proposed, including the spacing adjustment stage, lane change control stage, and formation convergence stage. A platoon formation controller was designed based on consensus theory, and the stability of the controller is demonstrated. Formation experiment of vehicle clusters was conducted in extreme initial distribution scenarios within a three-lane environment, with two typical scenarios being selected to perform numerical simulations and vehicle dynamics simulations of the formation process. The results show that for highly uneven and irregularly distributed vehicle clusters on multiple lanes, all vehicles are effectively controlled to achieve the desired formation in a safe and efficient manner. Throughout the control process, the longitudinal distance error between vehicles and the desired spacing is within 0.5 m, and the lateral position error relative to the desired position is within 5 cm, and the velocity quickly converges to the desired value after changes. Therefore, the strategy can effectively control the vehicles of the cluster to safely and efficiently achieve the desired formation.

  Figures and Tables | References | Related Articles | Metrics

  Intelligent vehicle path planning method based on peripheral vehicle trajectory prediction

  HUANG Chen, JIA Dingpeng, SUN Xiaoqiang, XU Qing

  2024, 15(5):  753-762.  doi:10.3969/j.issn.1674-8484.2024.05.013

  Abstract ( 26 )   HTML ( 2)   PDF (1557KB) ( 14 )  

  A path planning method was investigated based on the peripheral vehicle trajectory prediction with doing digital simulations to improve the driving safety and access efficiency of intelligent vehicles in dynamic driving environments. The peripheral vehicle trajectory prediction method was proposed based on the Spatio-Temporal Graph Convolutional Network (STGCN), which encoded the historical vehicle trajectories through STGCN, extracted the spatio-temporal features of traffic maps and combined with long and short-term memory networks to achieve the trajectory prediction of peripheral vehicles. On this basis, a path planning method was proposed based on an Improved Artificial Potential Field (APF), and an APF-based driving hazard evaluation module was established, which described the driving hazard by using the Frenet coordinates, and completed the path planning through the potential distribution of target obstacles and road boundaries and the gradient descent method. The results show that the proposed algorithm improves prediction accuracy by about 3% in the short-time prediction and by 1% in the long-time prediction with a path curve of the front wheel angle not exceeding 0.12 rad, and a curvature not exceeding 0.1, ensuring comfort and high efficiency during vehicle travel while effectively avoiding collisions.

  Figures and Tables | References | Related Articles | Metrics

  Explainable lane change intention prediction based on LSTM-multi-head mixed attention

  GAO Kai, LIU Jian, LIU Linhong, LIU Xinyu, ZHANG Jinlai, DU Ronghua

  2024, 15(5):  763-773.  doi:10.3969/j.issn.1674-8484.2024.05.014

  Abstract ( 23 )   HTML ( 2)   PDF (1758KB) ( 8 )  

  An interpretable lane change intention prediction model was proposed to enable the autonomous vehicle to accurately predict the lane change intention of the vehicles around them. This model based on the Long Short-Term Memory (LSTM) and the multi-head mixed attention, which can fully extract the spatiotemporal interaction between the target vehicle and its surrounding vehicles. A Shapley additive interpretation method (SHAP) based on maximum entropy was proposed to explain the degree of influence of each feature on the model output at a specific time step, and experiments on the HighD dataset were carried out. The results show that the comprehensive accuracy of the proposed model is 4.03%, 9.51%, and 5.16% higher than that of the LSTM, the Convolutional Neural Network (CNN), and the multi-head attention, respectively, before lane change, which fully proves the validity of the model in the long time horizon. And the wrong prediction samples can be attributed to model defects or sparse samples on the other hand, guiding users to optimize the model.

  Figures and Tables | References | Related Articles | Metrics

  Research on transient driving risk vector modeling method under strong constraints of traffic regulations

  ZHENG Xunjia, JIANG Junhao, LI Huilan, CHEN Xing, LIU Hui, WANG Jianqiang, GAO Jianjie

  2024, 15(5):  774-782.  doi:10.3969/j.issn.1674-8484.2024.05.015

  Abstract ( 24 )   HTML ( 2)   PDF (3133KB) ( 7 )  

  To mitigate or alleviate the occurrence of serious accidents where a preceding vehicle stops to yield at a traffic light intersection and was rear-ended by an out-of-control vehicle, a vector field modeling method for vehicle risk was proposed based on the fundamental model of driving risk field force established in the previous studies. An intersection scenario without traffic signals was designed and the safety simulations was conducted under six driving conditions. A dangerous scenario was developed, where a vehicle at a traffic light intersection was at risk of being rear-ended by an out-of-control following vehicle; then, four evasive maneuvers (going straight, turning left, turning right, and making a U-turn) was analyzed without considering road traffic regulations; finally, the force distribution of driving risks across twelve conditions were compared and analyzed. The results show that the proposed model can effectively identify driving risks. The evasive maneuver of the vehicle making a U-turn into the opposite lane is the most optimal, reducing overall risk by 67.41% when the speed is 3 m/s.

  Figures and Tables | References | Related Articles | Metrics

  Research on dynamic modeling of port autonomous driving truck

  XIE Zhen, ZHOU Guofeng, WU Mingyu, CAO Shouqi

  2024, 15(5):  783-794.  doi:10.3969/j.issn.1674-8484.2024.05.016

  Abstract ( 24 )   HTML ( 3)   PDF (2510KB) ( 7 )  

  A subsystem-coupled tractor-trailer articulation dynamic modeling method for autonomous driving trucks was proposed to meet the high-precision tractor-trailer articulation dynamic modeling requirements for autonomous driving simulation testing. Firstly, based on the tractor-trailer articulation relationship, an accurate kinematic description of trucks was carried out. According to the kinematic relationship between the tractor and trailer of autonomous driving trucks, the lateral, longitudinal, and yaw dynamics models of the tractor and trailer were established using Newtonian mechanics. Considering the control requirements of autonomous driving simulation testing, the drive, brake, tire, steering, and aerodynamic subsystems were described separately. Secondly, considering the variability of the loading quality of trucks, the position of the center of mass and the moment of inertia of the truck were calculated, and the vertical load of the tire was obtained based on assumptions, completing the construction of the coupled dynamics model of the entire subsystem. And numerical simulations were conducted under emergency braking and double shift line conditions. The accuracy was compared with TruckSim. Aiming at the port environment, a virtual simulation system was established based on the truck dynamics model. The results show that the RMSE is below 0.05 when the accuracy was compared with TruckSim; Under the port environment, the maximum deviation in path tracking testing was less than 0.6 m, indicating that this method can accurately describe the dynamic response of the container truck under different operating conditions.

  Figures and Tables | References | Related Articles | Metrics