Data-Driven Decision Making and Near-Optimal Path Planning for Multiagent System in Games

• Published in: IEEE journal on miniaturization for air and space systems Vol. 4; no. 3; pp. 320 - 328
• Main Authors: Wang, Xindi, Liu, Hao, Gao, Qing
• Format: Journal Article
• Language: English
• Published: Piscataway IEEE 01.09.2023; The Institute of Electrical and Electronics Engineers, Inc. (IEEE)
• Subjects: Algorithms; Collision avoidance; Decision making; Dynamic programming; Dynamical systems; Game theory; Games; Heuristic algorithms; Machine learning; multiagent system; Multiagent systems; Nonlinear dynamics; Path planning; Reinforcement learning; unmanned aerial vehicle (UAV); Unmanned aerial vehicles
• ISSN: 2576-3164; 2576-3164
• DOI: 10.1109/JMASS.2023.3292259

In this article, the optimal real-time decision making and near-optimal path planning problem for multiagent systems subject to bounded state, collision avoidance, external disturbance, and partially unknown nonlinear dynamics of the multiagent system in complex games, is addressed and applied to the unmanned aerial vehicle. A mean-field decision-making model based on the neighbor information is established to transform the decision-making problem into a Bellman equation solving problem. A data-driven dynamic programming algorithm is proposed to solve the Bellman equation and generate an optimal strategy using the data from the historical database and expert knowledge. The near-optimal path planning problem is formulated with an optimal coordination control problem, and an online integral reinforcement learning algorithm is proposed to iteratively interact with the environment to obtain a near-optimal path. Simulation results are provided to verify the effectiveness of the proposed methods.