Advanced Optimization and Game Theory for Interaction-Aware Autonomous Driving

Abstract

With the rapid development of technologies in autonomous driving, recent research has witnessed a surge in studies of decision-making and trajectory planning techniques for autonomous vehicles. Although great advancements have been achieved, decision-making and trajectory planning methods for autonomous driving with interaction awareness re-main largely unexplored, as many existing methods decouple the prediction process and the planning process, resulting in predictive planning schemes that fail to capture the interactive nature of traffic scenarios. In this dissertation, we aim to introduce optimal decision-making and trajectory planning schemes for autonomous driving with interaction-awareness. Particularly, we hope to address the following two critical technical problems: 1) how to effectively model the interaction between traffic agents through mathematical formulations, and 2) how to solve the formulated problems with high computational efficiency to obtain optimal decision-making and trajectory planning results on a real-time basis. We approach these problems by investigating both the interaction between connected autonomous vehicles and the interaction between autonomous vehicles and human drivers. Under various applications and settings, methods based on advanced optimization theory and game theory are utilized to provide effective problem formulations for modeling interactions between traffic agents without loss of generalization ability. Together with techniques like distributed optimization, the computation efficiency and scalability are enhanced. For validation of the proposed methods, a series of simulations and/or experiments is performed for each introduced method, showing the advancement of the proposed interactive autonomous driving schemes in terms of computational efficiency, scalability, optimality, and robustness.

Date of Award	2025
Original language	English
Awarding Institution	The Hong Kong University of Science and Technology
Supervisor	Jun MA (Supervisor) & Shaojie SHEN (Supervisor)