Distributed Nash Equilibrium Solution for Asymmetric Multiagent Pursuit–Evasion Differential Games via Reinforcement Learning
This article addresses an asymmetric multiagent pursuit-evasion (MPE) differential gaming scenario, where pursuers, subject to the presence of external disturbances and unknown dynamics, aim to collaboratively intercept escaping evaders. The core technical challenge lies in designing a distributed optimal gaming protocol for such incomplete-information MPE games in an adversarial environment while guaranteeing the attainment of the Nash equilibrium. To address this problem, a model-based differential gaming framework is presented to eliminate the coupled terms in the associated Hamilton-Jacobi-Isaacs (HJI) equations and accordingly derive a distributed Nash equilibrium solution for multiagent systems (MASs). To dispense with prior knowledge of the pursuers' dynamics, a data-driven off-policy reinforcement learning (RL) algorithm is devised to learn the pursuers' optimal strategies online. Finally, simulation results are elaborated to validate the effectiveness of the proposed gaming strategy.
Authors
- Ran Shi (ORCID: https://orcid.org/0009-0006-5338-4724)
- Jun Wang (ORCID: https://orcid.org/0000-0002-1305-5735)
- Hai‐Tao Zhang (ORCID: https://orcid.org/0000-0002-8819-8829)
Institutions
- City University of Hong Kong (HK)
- Huazhong University of Science and Technology (CN)
Publication Details
- Journal
- IEEE Transactions on Cybernetics
- Published
- 2026-11-17
- DOI
- https://doi.org/10.1109/tcyb.2026.3750963
- Primary Topic
- Adaptive Dynamic Programming Control
- Type
- article
- Field-Weighted Citation Impact
- 0.00
Funders
- National Natural Science Foundation of China
- National Key Research and Development Program of China