Pengcheng Wang
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Pengcheng Wang has a total of 28 co-authors. Here are the top co-authors by number of shared publications:
Simiao Gao
Beihang University
2 shared publicationsZhonghao Li
Beihang University
2 shared publications- Chuan Ding
School of Transportation Science and Engineering, Beijing Key Laboratory for Cooperative Vehicle Infrastructure System and Safety Control, Beihang University, Beijing, China
1 shared publication Fu Xiufeng
China Aerospace Science and Industry Corporation 706 Research Institute
1 shared publication- Guangnian Xiao
School of Economics and Management, Shanghai Maritime University
1 shared publication - Haojie Ji
Hefei Innovation Research Institute, Beihang University, Hefei, China
1 shared publication
Latest Publications
- Open Access
Intelligent Sensing for Transportation Safety: Technologies, Innovations, and Challenges
Transportation safety remains a critical global concern, with road traffic crashes causing approximately 1
Open AccessIntelligent Sensing for Transportation Safety: Technologies, Innovations, and Challenges
Transportation safety remains a critical global concern, with road traffic crashes causing approximately 1
- 1 January 2026
- Published by Tech Science Press in Computers, Materials & Continua
- p. 1-10
Hybrid Vehicular Cluster Control Model Based on Transformer-Deep Reinforcement Learning
Connected and automated vehicles (CAVs) are capable of substantially enhancing roadway safety and traffic throughput in hybrid driving environments. To address the challenges of cluster control and validation testing for intelligent connected vehicles, a vehicular cluster control method and a multi-domain controller-in-the-loop simulation testing approach for hybrid driving environments are proposed in this paper. First, a unified modeling framework for the vehicular cluster control system is established by constructing a multiple-predecessor multiple-leader following topology, a variable time headway strategy, and a dynamics model of hybrid vehicular clusters. Second, a multi-agent reinforcement learning-based vehicular cluster control model is developed using a hierarchical Transformer framework to dynamically characterize multi-layer cluster states. Hybrid vehicular cluster control is achieved through a shared-parameter multi-head Double Deep Q-Network (DDQN) algorithm. Finally, the proposed Transformer-Double Deep Q-Network (Trans-DDQN) control model is deployed on domain controllers via model pruning and quantization techniques. A multi-domain controller-in-the-loop simulation testing system is constructed based on PreScan/MATLAB, incorporating a driving simulator to replicate hybrid driving scenarios. The final experimental validation demonstrates that, compared with the traditional Model Predictive Control (MPC) and Layered Conventional Double Deep Q-Network (LC-DDQN) model, the proposed method achieves markedly faster acceleration convergence than LC-DDQN, reduces the maximum inter-vehicle spacing deviation by 3.1% and 7.3% relative to LC-DDQN and MPC, and raises the minimum time to collision (TTC) from 1.9 to 5.0 s. These results indicate that the Trans-DDQN effectively enhances the operational stability of vehicular clusters and ensures consistent speed, acceleration, and spacing control.
- 1 January 2026
- Published by Tech Science Press in Computers, Materials & Continua
- p. 1-10
Hybrid Vehicular Cluster Control Model Based on Transformer-Deep Reinforcement Learning
Connected and automated vehicles (CAVs) are capable of substantially enhancing roadway safety and traffic throughput in hybrid driving environments. To address the challenges of cluster control and validation testing for intelligent connected vehicles, a vehicular cluster control method and a multi-domain controller-in-the-loop simulation testing approach for hybrid driving environments are proposed in this paper. First, a unified modeling framework for the vehicular cluster control system is established by constructing a multiple-predecessor multiple-leader following topology, a variable time headway strategy, and a dynamics model of hybrid vehicular clusters. Second, a multi-agent reinforcement learning-based vehicular cluster control model is developed using a hierarchical Transformer framework to dynamically characterize multi-layer cluster states. Hybrid vehicular cluster control is achieved through a shared-parameter multi-head Double Deep Q-Network (DDQN) algorithm. Finally, the proposed Transformer-Double Deep Q-Network (Trans-DDQN) control model is deployed on domain controllers via model pruning and quantization techniques. A multi-domain controller-in-the-loop simulation testing system is constructed based on PreScan/MATLAB, incorporating a driving simulator to replicate hybrid driving scenarios. The final experimental validation demonstrates that, compared with the traditional Model Predictive Control (MPC) and Layered Conventional Double Deep Q-Network (LC-DDQN) model, the proposed method achieves markedly faster acceleration convergence than LC-DDQN, reduces the maximum inter-vehicle spacing deviation by 3.1% and 7.3% relative to LC-DDQN and MPC, and raises the minimum time to collision (TTC) from 1.9 to 5.0 s. These results indicate that the Trans-DDQN effectively enhances the operational stability of vehicular clusters and ensures consistent speed, acceleration, and spacing control.
- 9 May 2025
- Published by Elsevier in Accident Analysis & Prevention
- Vol. 218, 108094
Ensuring SOTIF: Enhanced object detection techniques for autonomous driving
- 9 May 2025
- Published by Elsevier in Accident Analysis & Prevention
- Vol. 218, 108094
Ensuring SOTIF: Enhanced object detection techniques for autonomous driving
- 5 March 2025
- Published by Springer Nature
- p. 58-67
Research on Security Authentication Methods for Unmanned Aerial Vehicle Systems in Mosaic Warfare Scenarios
- 5 March 2025
- Published by Springer Nature
- p. 58-67
Research on Security Authentication Methods for Unmanned Aerial Vehicle Systems in Mosaic Warfare Scenarios
- 5 March 2025
- Published by Springer Nature
- p. 435-446
Research on Fuzzing of Autonomous Driving Systems
- 15 June 2024
- Published by Springer Nature
- p. 455-464
Fuel Consumption and Emissions Analysis of a Connected Automated Vehicle Platoon in Unstable Traffic
- 15 June 2024
- Published by Springer Nature
- p. 455-464
Fuel Consumption and Emissions Analysis of a Connected Automated Vehicle Platoon in Unstable Traffic
- 9 May 2023
- Published by Springer Nature in Journal of Intelligent & Robotic Systems
- Vol. 108 (1) , 1-17
An Enhanced Adaptive Monte Carlo Localization for Service Robots in Dynamic and Featureless Environments
- 9 May 2023
- Published by Springer Nature in Journal of Intelligent & Robotic Systems
- Vol. 108 (1) , 1-17
An Enhanced Adaptive Monte Carlo Localization for Service Robots in Dynamic and Featureless Environments
- 19 October 2021
- Published by Springer Nature in Personal and Ubiquitous Computing
- Vol. 28 (1) , 227-244
A secure and efficient authentication scheme in the Internet of Vehicle communication
- 19 October 2021
- Published by Springer Nature in Personal and Ubiquitous Computing
- Vol. 28 (1) , 227-244
A secure and efficient authentication scheme in the Internet of Vehicle communication