Zero-Sum Game Optimized Control With Augmented Time-Synchronized Property

• 成果类型: Article
• 署名作者: Zhang, Yuxiang; Li, Dongyu; Ge, Shuzhi Sam; Lee, Tong Heng
• 署名单位: National University of Singapore; Beihang University
• 刊物名称: IEEE TRANSACTIONS ON AUTOMATIC CONTROL
• ISSN/ISSBN: 0018-9286
• DOI: 10.1109/TAC.2026.3676685
• 发表日期: 2026
• 关键词: systems

摘要

This article proposes and rigorously develops a reinforcement learning (RL)-based optimized control strategy with notable time-synchronized stability properties for zero-sum differential games. The proposed method addresses the challenge of approximating the time-synchronized Nash equilibrium solutions in nonlinear systems governed by the Hamilton-Jacobi-Isaacs equation. By incorporating a norm-normalized sign function into the learning framework, the system ensures all state-variables converge simultaneously, improving robustness, and energy efficiency. The RL-based optimization iteratively refines the control policies while maintaining system stability underpinned rigorously by Lyapunov-based analysis. To demonstrate the effectiveness of the proposed approach, a motion control problem for an autonomous vehicle system is simulated, comparing the results with alternative existing fixed-time sliding control and time-synchronized optimized control methods. The results illustrate that the proposed control strategy enhances convergence speed, smoothness, and disturbance rejection, making it well-suited for the requirements of real-world high-precision control applications.