Date of Award
3-3-2025
Publication Type
Master Thesis
Degree Name
M.A.Sc.
Department
Electrical and Computer Engineering
Supervisor
Ahmed H. Sakr
Rights
info:eu-repo/semantics/embargoedAccess
Creative Commons License
This work is licensed under a Creative Commons Attribution-NonCommercial-No Derivative Works 4.0 International License.
Abstract
This research introduces an advanced multi-agent reinforcement learning (MARL) framework designed to enhance autonomous vehicle platooning in both homogeneous and mixed traffic conditions. By integrating intent-sharing mechanisms, this framework enables vehicles to exchange real-time states and future predictions of speed and acceleration. This shared intent fosters more coordinated and informed decision-making, reducing inter-vehicle gap errors, and stability, mitigating the propagation of disturbances along the platoon and ensuring smoother vehicle dynamics, improved passenger comfort, and heightened safety. The framework incorporates dynamic hyperparameter optimization and employs the Multi-Agent Twin Delayed Deep Deterministic Policy Gradient (MATD3) algorithm to strike a robust balance between individual vehicle control and collective platoon dynamics. A novel reward structure is designed to address key factors such as safety, jerk minimization, gap keeping and driven behavior and predict intents. Extensive simulations validate the efficacy of this framework, demonstrating its ability to significantly reduce gap errors, speed inconsistencies, and abrupt acceleration changes. The results further underscore the framework’s potential to strengthen string stability, particularly in complex mixed traffic scenarios. This research highlights the revolutionary potential of combining reinforcement learning, intent sharing and systematic reward engineering to advance the capabilities of autonomous vehicle platooning systems, paving the way for safer, more stable, and comfortable autonomous driving experiences in real-world environments.
Recommended Citation
Tandra, Tahmina Khanom, "Multi-Agent Cooperative Adaptive Cruise Control Based on Intent Sharing Messages" (2025). Electronic Theses and Dissertations. 9696.
https://scholar.uwindsor.ca/etd/9696
