Li Wang
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Li Wang has a total of 28 co-authors. Here are the top co-authors by number of shared publications:
- Fumiya Iida
Department of Engineering, University of Cambridge, Cambridge, UK
2 shared publications - Mingfang Zhang
Beijing Key Lab of Urban Intelligent Traffic Control Technology, North China University of Technology, Beijing 100144, China
2 shared publications - Xiaoping Zhang
North China University of Technology
2 shared publications - Xiaoyuan Luo
Yanshan University
2 shared publications - Yijing Wang
Tianjin University
2 shared publications - Zheng Li
Tianjin Key Laboratory of Intelligent Unmanned Swarm Technology and System, School of Electrical and Information Engineering, Tianjin University, Tianjin, China
2 shared publications
Latest Publications
- Open Access
- 19 June 2024
- Published by MDPI AG in Journal of Marine Science and Engineering
- Vol. 12 (6) , 1022
Applications of Voronoi Diagrams in Multi-Robot Coverage: A Review
In recent decades, multi-robot region coverage has played an important role in the fields of environmental sensing, target searching, etc., and it has received widespread attention worldwide. Due to the effectiveness in segmenting nearest regions, Voronoi diagrams have been extensively used in recent years for multi-robot region coverage. This paper presents a survey of recent research works on region coverage methods within the framework of the Voronoi diagram, to offer a perspective for researchers in the multi-robot cooperation domain. First, some basic knowledge of the Voronoi diagram is introduced. Then, the region coverage issue under the Voronoi diagram is categorized into sensor coverage and task execution coverage problems, respectively, considering the sensor range parameter. Furthermore, a detailed analysis of the application of Voronoi diagrams to the aforementioned two problems is provided. Finally, some conclusions and potential further research perspectives in this field are given.
- 19 June 2024
- Published by MDPI AG in Journal of Marine Science and Engineering
- Vol. 12 (6) , 1022
Open AccessApplications of Voronoi Diagrams in Multi-Robot Coverage: A Review
In recent decades, multi-robot region coverage has played an important role in the fields of environmental sensing, target searching, etc., and it has received widespread attention worldwide. Due to the effectiveness in segmenting nearest regions, Voronoi diagrams have been extensively used in recent years for multi-robot region coverage. This paper presents a survey of recent research works on region coverage methods within the framework of the Voronoi diagram, to offer a perspective for researchers in the multi-robot cooperation domain. First, some basic knowledge of the Voronoi diagram is introduced. Then, the region coverage issue under the Voronoi diagram is categorized into sensor coverage and task execution coverage problems, respectively, considering the sensor range parameter. Furthermore, a detailed analysis of the application of Voronoi diagrams to the aforementioned two problems is provided. Finally, some conclusions and potential further research perspectives in this field are given.
- 19 December 2023
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Intelligent Transportation Systems
- Vol. 25 (6) , 6118-6132
DS-UKF-Based Positioning Method for Intelligent Connected Vehicles in Urban Intersection Scenarios
In order to locate the position of intelligent connected vehicles (ICVs) robustly in urban intersection scenarios, a novel compound positioning method based on the Dempster-Shafer evidence theory and the improved multi-layer unscented Kalman filter (DS-UKF) is proposed in this paper. The Euclidean distance between the heterogeneous data is used to optimize the Dempster-Shafer (D-S) evidence theory, so that the position of ICVs and the distance between ICVs and roadside positioning platform with higher confidence can be obtained. Meanwhile, considering the high complication of urban intersections, the unscented Kalman filter (UKF) algorithm is improved based on the covariance of error sequences so that the influences caused by observation noise anomalies can be reduced. To further improve the accuracy of positional prediction, the position of roadside positioning platform is extended dimensionally into the state vector of ICVs to observe the distance between ICVs and roadside positioning platform. Then the particle filter (PF) theory is introduced to integrate the upper-layer with the lower-layer UKF to reduce the positioning errors brought by signal interferences. Besides, an adaptive factor is adopted to connect D-S theory with the improved multi-layer UKF (IUKF) to dynamically update the weights of output values from sensors. Furthermore, a communication mechanism between the ICVs platform and the roadside positioning platform is established in this project with considering the reliability and low latency of vehicle to everything (V2X) communications. Finally, the feasibility and superiority of the proposed method is validated by experiments with three ICVs at an urban intersection. The experimental results have shown that the proposed method can locate the position of ICVs rapidly and stably in urban intersections and has better positioning performance compared to the current state-of-the-art algorithms.
- 19 December 2023
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Intelligent Transportation Systems
- Vol. 25 (6) , 6118-6132
DS-UKF-Based Positioning Method for Intelligent Connected Vehicles in Urban Intersection Scenarios
In order to locate the position of intelligent connected vehicles (ICVs) robustly in urban intersection scenarios, a novel compound positioning method based on the Dempster-Shafer evidence theory and the improved multi-layer unscented Kalman filter (DS-UKF) is proposed in this paper. The Euclidean distance between the heterogeneous data is used to optimize the Dempster-Shafer (D-S) evidence theory, so that the position of ICVs and the distance between ICVs and roadside positioning platform with higher confidence can be obtained. Meanwhile, considering the high complication of urban intersections, the unscented Kalman filter (UKF) algorithm is improved based on the covariance of error sequences so that the influences caused by observation noise anomalies can be reduced. To further improve the accuracy of positional prediction, the position of roadside positioning platform is extended dimensionally into the state vector of ICVs to observe the distance between ICVs and roadside positioning platform. Then the particle filter (PF) theory is introduced to integrate the upper-layer with the lower-layer UKF to reduce the positioning errors brought by signal interferences. Besides, an adaptive factor is adopted to connect D-S theory with the improved multi-layer UKF (IUKF) to dynamically update the weights of output values from sensors. Furthermore, a communication mechanism between the ICVs platform and the roadside positioning platform is established in this project with considering the reliability and low latency of vehicle to everything (V2X) communications. Finally, the feasibility and superiority of the proposed method is validated by experiments with three ICVs at an urban intersection. The experimental results have shown that the proposed method can locate the position of ICVs rapidly and stably in urban intersections and has better positioning performance compared to the current state-of-the-art algorithms.
- Open Access
- 1 November 2023
- Published by MDPI AG in Applied Sciences
- Vol. 13 (21) , 11951
Multi-Agent Collaborative Target Search Based on the Multi-Agent Deep Deterministic Policy Gradient with Emotional Intrinsic Motivation
Multi-agent collaborative target search is one of the main challenges in the multi-agent field, and deep reinforcement learning (DRL) is a good way to learn such a task. However, DRL always faces the problem of sparse reward, which to some extent reduces its efficiency in task learning. Introducing intrinsic motivation has proved to be a useful way to make the sparse reward in DRL. So, based on the multi-agent deep deterministic policy gradient (MADDPG) structure, a new MADDPG algorithm with the emotional intrinsic motivation name MADDPG-E is proposed in this paper for the multi-agent collaborative target search. In MADDPG-E, a new emotional intrinsic motivation module with three emotions, joy, sadness, and fear, is designed. The three emotions are defined by corresponding psychological knowledge to the multi-agent embodied situations in an environment. An emotional steady-state variable function H is then designed to help judge the goodness of the emotions. Based on H, an emotion-based intrinsic reward function is finally proposed. With the designed emotional intrinsic motivation module, the multi-agent system always tries to make itself joy, which means it always learns to search the target. To show the effectiveness of the proposed MADDPG-E algorithm, two kinds of simulation experiments with a determined initial position and random initial position, respectively, are carried out, and comparisons are performed with MADDPG as well as MADDPG-ICM (MADDPG with an intrinsic curiosity module). The results show that with the designed emotional intrinsic motivation module, MADDPG-E has a higher learning speed and better learning stability, and the advantage is more obvious when facing complex situations.
- 1 November 2023
- Published by MDPI AG in Applied Sciences
- Vol. 13 (21) , 11951
Open AccessMulti-Agent Collaborative Target Search Based on the Multi-Agent Deep Deterministic Policy Gradient with Emotional Intrinsic Motivation
Multi-agent collaborative target search is one of the main challenges in the multi-agent field, and deep reinforcement learning (DRL) is a good way to learn such a task. However, DRL always faces the problem of sparse reward, which to some extent reduces its efficiency in task learning. Introducing intrinsic motivation has proved to be a useful way to make the sparse reward in DRL. So, based on the multi-agent deep deterministic policy gradient (MADDPG) structure, a new MADDPG algorithm with the emotional intrinsic motivation name MADDPG-E is proposed in this paper for the multi-agent collaborative target search. In MADDPG-E, a new emotional intrinsic motivation module with three emotions, joy, sadness, and fear, is designed. The three emotions are defined by corresponding psychological knowledge to the multi-agent embodied situations in an environment. An emotional steady-state variable function H is then designed to help judge the goodness of the emotions. Based on H, an emotion-based intrinsic reward function is finally proposed. With the designed emotional intrinsic motivation module, the multi-agent system always tries to make itself joy, which means it always learns to search the target. To show the effectiveness of the proposed MADDPG-E algorithm, two kinds of simulation experiments with a determined initial position and random initial position, respectively, are carried out, and comparisons are performed with MADDPG as well as MADDPG-ICM (MADDPG with an intrinsic curiosity module). The results show that with the designed emotional intrinsic motivation module, MADDPG-E has a higher learning speed and better learning stability, and the advantage is more obvious when facing complex situations.
- Open Access
- 7 October 2023
- Published by MDPI AG in Applied Sciences
- Vol. 13 (19) , 11045
Adaptive Locomotion Learning for Quadruped Robots by Combining DRL with a Cosine Oscillator Based Rhythm Controller
Animals have evolved to adapt to complex and uncertain environments, acquiring locomotion skills for diverse surroundings. To endow a robot’s animal-like locomotion ability, in this paper, we propose a learning algorithm for quadruped robots based on deep reinforcement learning (DRL) and a rhythm controller that is based on a cosine oscillator. For a quadruped robot, two cosine oscillators are utilized at the hip joint and the knee joint of one leg, respectively, and, finally, eight oscillators form the controller to realize the quadruped robot’s locomotion rhythm during moving. The coupling between the cosine oscillators of the rhythm controller is realized by the phase difference, which is simpler and easier to realize when dealing with the complex coupling relationship between different joints. DRL is used to help learn the controller parameters and, in the reward function design, we address the challenge of terrain adaptation without relying on the complex camera-based vision processing but based on the proprioceptive information, where a state estimator is introduced to achieve the robot’s posture and help finally utilize the food-end coordinate. Experiments are carried out in CoppeliaSim, and all of the flat, uphill and downhill conditions are considered. The results show that the robot can successfully accomplish all the above skills and, at the same time, with the reward function designed, the robot’s pitch angle, yaw angle and roll angle are very small, which means that the robot is relatively stable during walking. Then, the robot is transplanted to a new scene; the results show that although the environment is previously unencountered, the robot can still fulfill the task, which demonstrates the effectiveness and robustness of this proposed method.
- 7 October 2023
- Published by MDPI AG in Applied Sciences
- Vol. 13 (19) , 11045
Open AccessAdaptive Locomotion Learning for Quadruped Robots by Combining DRL with a Cosine Oscillator Based Rhythm Controller
Animals have evolved to adapt to complex and uncertain environments, acquiring locomotion skills for diverse surroundings. To endow a robot’s animal-like locomotion ability, in this paper, we propose a learning algorithm for quadruped robots based on deep reinforcement learning (DRL) and a rhythm controller that is based on a cosine oscillator. For a quadruped robot, two cosine oscillators are utilized at the hip joint and the knee joint of one leg, respectively, and, finally, eight oscillators form the controller to realize the quadruped robot’s locomotion rhythm during moving. The coupling between the cosine oscillators of the rhythm controller is realized by the phase difference, which is simpler and easier to realize when dealing with the complex coupling relationship between different joints. DRL is used to help learn the controller parameters and, in the reward function design, we address the challenge of terrain adaptation without relying on the complex camera-based vision processing but based on the proprioceptive information, where a state estimator is introduced to achieve the robot’s posture and help finally utilize the food-end coordinate. Experiments are carried out in CoppeliaSim, and all of the flat, uphill and downhill conditions are considered. The results show that the robot can successfully accomplish all the above skills and, at the same time, with the reward function designed, the robot’s pitch angle, yaw angle and roll angle are very small, which means that the robot is relatively stable during walking. Then, the robot is transplanted to a new scene; the results show that although the environment is previously unencountered, the robot can still fulfill the task, which demonstrates the effectiveness and robustness of this proposed method.
- Open Access
- 20 February 2023
- Published by MDPI AG in Sustainability
- Vol. 15 (4) , 3847
An RSU Deployment Scheme for Vehicle-Infrastructure Cooperated Autonomous Driving
For autonomous driving vehicles, there are currently some issues, such as limited environmental awareness and locally optimal decision-making. To increase the capacity of autonomous cars’ environmental awareness, computation, decision-making, control, and execution, intelligent roads must be constructed, and vehicle–infrastructure cooperative technology must be used. The Roadside unit (RSU) deployment, a critical component of vehicle–infrastructure cooperative autonomous driving, has a direct impact on network performance, operation effects, and control accuracy. The current RSU deployment mostly uses the large-spacing and low-density concept because of the expensive installation and maintenance costs, which can accomplish the macroscopic and long-term communication functions but fall short of precision vehicle control. Given these challenges, this paper begins with the specific requirements to control intelligent vehicles in the cooperative vehicle–infrastructure environment. An RSU deployment scheme, based on the improved multi-objective quantum-behaved particle swarm optimization (MOQPSO) algorithm, is proposed. This RSU deployment scheme was based on the maximum coverage with time threshold problem (MCTTP), with the goal of minimizing the number of RSUs and maximizing vehicle coverage of communication and control services. Finally, utilizing the independently created open simulation platform (OSP) simulation system, the model and algorithm’s viability and effectiveness were assessed on the Nguyen–Dupuis road network. The findings demonstrate that the suggested RSU deployment scheme can enhance network performance and control the precision of vehicle–infrastructure coordination, and can serve as a general guide for the deployment of RSUs in the same application situation.
- 20 February 2023
- Published by MDPI AG in Sustainability
- Vol. 15 (4) , 3847
Open AccessAn RSU Deployment Scheme for Vehicle-Infrastructure Cooperated Autonomous Driving
For autonomous driving vehicles, there are currently some issues, such as limited environmental awareness and locally optimal decision-making. To increase the capacity of autonomous cars’ environmental awareness, computation, decision-making, control, and execution, intelligent roads must be constructed, and vehicle–infrastructure cooperative technology must be used. The Roadside unit (RSU) deployment, a critical component of vehicle–infrastructure cooperative autonomous driving, has a direct impact on network performance, operation effects, and control accuracy. The current RSU deployment mostly uses the large-spacing and low-density concept because of the expensive installation and maintenance costs, which can accomplish the macroscopic and long-term communication functions but fall short of precision vehicle control. Given these challenges, this paper begins with the specific requirements to control intelligent vehicles in the cooperative vehicle–infrastructure environment. An RSU deployment scheme, based on the improved multi-objective quantum-behaved particle swarm optimization (MOQPSO) algorithm, is proposed. This RSU deployment scheme was based on the maximum coverage with time threshold problem (MCTTP), with the goal of minimizing the number of RSUs and maximizing vehicle coverage of communication and control services. Finally, utilizing the independently created open simulation platform (OSP) simulation system, the model and algorithm’s viability and effectiveness were assessed on the Nguyen–Dupuis road network. The findings demonstrate that the suggested RSU deployment scheme can enhance network performance and control the precision of vehicle–infrastructure coordination, and can serve as a general guide for the deployment of RSUs in the same application situation.
- Top Cited Papers
- 18 December 2020
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Intelligent Vehicles
- Vol. 6 (3) , 513-522
MPC-Based Cooperative Control Strategy of Path Planning and Trajectory Tracking for Intelligent Vehicles
In this paper, we propose a progressive model predictive control scheme (PMPCS) by considering the cooperative control of local planning and path tracking for intelligent vehicles. An improved particle swarm optimization (IPSO) based model predictive control (MPC) method is developed to solve the planning and tracking problem. With the PMPCS, the total computational burden can be reduced sharply because of the seamless connection and mutual promotion between the optimization of two layers. Besides we also propose a novel planning algorithm, which can take traffic lights and overtaking time constraint into account. To solve these problems, we first combine model predictive control with artificial potential field (APF) to get a collision-free path by treating the timevarying safety constraints as the scope of the repulsive force and an asymmetrical lane potential field function. Furthermore, the pseudo velocity planning method is adopted to take traffic lights into account in the planning module. Simulation results show the reliability of the proposed algorithm and the advantages of the scheme compared with general hierarchical algorithm.
- 18 December 2020
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Intelligent Vehicles
- Vol. 6 (3) , 513-522
Top Cited PapersMPC-Based Cooperative Control Strategy of Path Planning and Trajectory Tracking for Intelligent Vehicles
In this paper, we propose a progressive model predictive control scheme (PMPCS) by considering the cooperative control of local planning and path tracking for intelligent vehicles. An improved particle swarm optimization (IPSO) based model predictive control (MPC) method is developed to solve the planning and tracking problem. With the PMPCS, the total computational burden can be reduced sharply because of the seamless connection and mutual promotion between the optimization of two layers. Besides we also propose a novel planning algorithm, which can take traffic lights and overtaking time constraint into account. To solve these problems, we first combine model predictive control with artificial potential field (APF) to get a collision-free path by treating the timevarying safety constraints as the scope of the repulsive force and an asymmetrical lane potential field function. Furthermore, the pseudo velocity planning method is adopted to take traffic lights into account in the planning module. Simulation results show the reliability of the proposed algorithm and the advantages of the scheme compared with general hierarchical algorithm.
- Open Access
An Approach to Segment and Track-Based Pedestrian Detection from Four-Layer Laser Scanner Data
Pedestrian detection is a critical perception task for autonomous driving and intelligent vehicle, and it is challenging due to the potential variation of appearance and pose of human beings as well as the partial occlusion. In this paper, we present a novel pedestrian detection method via four-layer laser scanner. The proposed approach deals with the occlusion problem by fusing the segment classification results with past knowledge integration from tracking process. First, raw point cloud is segmented into the clusters of independent objects. Then, three types of features are proposed to capture the comprehensive cues, and 18 effective features are extracted with the combination of the univariate feature selection algorithm and feature correlation analysis process. Next, based on the segment classification at individual frame, the track classification is conducted further for consecutive frames using particle filter and probability data association filter. Experimental results demonstrate that both back-propagation neural network and Adaboost classifiers based on 18 selected features have their own advantages at the segment classification stage in terms of pedestrian detection performance and computation time, and the track classification procedure can improve the detection performance particularly for partially occluded pedestrians in comparison with the single segment classification procedure.
Open AccessAn Approach to Segment and Track-Based Pedestrian Detection from Four-Layer Laser Scanner Data
Pedestrian detection is a critical perception task for autonomous driving and intelligent vehicle, and it is challenging due to the potential variation of appearance and pose of human beings as well as the partial occlusion. In this paper, we present a novel pedestrian detection method via four-layer laser scanner. The proposed approach deals with the occlusion problem by fusing the segment classification results with past knowledge integration from tracking process. First, raw point cloud is segmented into the clusters of independent objects. Then, three types of features are proposed to capture the comprehensive cues, and 18 effective features are extracted with the combination of the univariate feature selection algorithm and feature correlation analysis process. Next, based on the segment classification at individual frame, the track classification is conducted further for consecutive frames using particle filter and probability data association filter. Experimental results demonstrate that both back-propagation neural network and Adaboost classifiers based on 18 selected features have their own advantages at the segment classification stage in terms of pedestrian detection performance and computation time, and the track classification procedure can improve the detection performance particularly for partially occluded pedestrians in comparison with the single segment classification procedure.
- Top Cited Papers
- 25 July 2019
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Vehicular Technology
- Vol. 68 (9) , 8546-8556
Trajectory Planning and Safety Assessment of Autonomous Vehicles Based on Motion Prediction and Model Predictive Control
Security problem is a fundamental issue for autonomous vehicles. Trajectory planning is a significant component of autonomous vehicle system, which directly influences the automated traffic safety. In this paper, the motion prediction of other traffic participants is considered. We use Monte Carlo simulation to predict the probabilistic occupancy of the object, and give a map from probability statistics to actual scenarios. The non-time based reference trajectory can be obtained by using high precision map (HD map) and lane detection. Then model predictive control (MPC) is utilized to optimize the reference trajectory according to the current state of autonomous vehicle. Different prediction horizons and coordinate transformation are adopted to optimize the planning. By doing so, the constraint conditions can be easily involved and the result is more intuitive. The probabilistic occupancy of other traffic participants are computed offline and then the obtained results are used in real time application. Therefore, the real-time computational burden is reduced. The crash probability is put forward to verify the feasibility of real-time trajectory in safety assessment module. Two typical scenarios are analyzed: lane change on the straight road and turning at the intersection. The simulation results illustrate the efficiency of our method.
- 25 July 2019
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Vehicular Technology
- Vol. 68 (9) , 8546-8556
Top Cited PapersTrajectory Planning and Safety Assessment of Autonomous Vehicles Based on Motion Prediction and Model Predictive Control
Security problem is a fundamental issue for autonomous vehicles. Trajectory planning is a significant component of autonomous vehicle system, which directly influences the automated traffic safety. In this paper, the motion prediction of other traffic participants is considered. We use Monte Carlo simulation to predict the probabilistic occupancy of the object, and give a map from probability statistics to actual scenarios. The non-time based reference trajectory can be obtained by using high precision map (HD map) and lane detection. Then model predictive control (MPC) is utilized to optimize the reference trajectory according to the current state of autonomous vehicle. Different prediction horizons and coordinate transformation are adopted to optimize the planning. By doing so, the constraint conditions can be easily involved and the result is more intuitive. The probabilistic occupancy of other traffic participants are computed offline and then the obtained results are used in real time application. Therefore, the real-time computational burden is reduced. The crash probability is put forward to verify the feasibility of real-time trajectory in safety assessment module. Two typical scenarios are analyzed: lane change on the straight road and turning at the intersection. The simulation results illustrate the efficiency of our method.