TY - JOUR
T1 - Cooperative capture strategy based on convergence set and MAPPO
T2 - Cooperative capture strategy based on convergence set
AU - ZHAO, Dahai
AU - ZHU, Jialong
AU - FU, Wenxing
N1 - Publisher Copyright:
© 2026 The Author(s)
PY - 2026/11
Y1 - 2026/11
N2 - Addressing a coplanar pursuit-evasion game in the atmosphere involving three pursuers against a more maneuverable and faster evader presents significant challenges. This paper proposes a cooperative capture strategy that combines convergence set coverage with multi-agent reinforcement learning algorithms. First, assuming pursuers and the evader are on an initial collision course and the control system can be neglected, an analytical solution for the convergence set is derived. This solution considers the coverage of the evader's acceleration capability range by the pursuer using biased Proportional Navigation Guidance (BPN) with specific bias terms. Subsequently, an online cooperative capture strategy is proposed, which adaptively adjusts a limited set of bias terms to achieve comprehensive coverage of the evader's acceleration capabilities and ensure successful capture in general three-pursuer, single-evader scenarios. To implement this strategy, the framework incorporates a two-phase learning approach: a pre-training phase based on Augmented Proportional Navigation guidance (APN) followed by a fine-tuning phase using Multi-Agent Proximal Policy Optimization (MAPPO). Numerical simulations demonstrate that the proposed strategy offers higher success rates and broader coverage of the evader's acceleration capabilities compared to traditional APN-based methods, ensuring effective capture against various evasion maneuvers.
AB - Addressing a coplanar pursuit-evasion game in the atmosphere involving three pursuers against a more maneuverable and faster evader presents significant challenges. This paper proposes a cooperative capture strategy that combines convergence set coverage with multi-agent reinforcement learning algorithms. First, assuming pursuers and the evader are on an initial collision course and the control system can be neglected, an analytical solution for the convergence set is derived. This solution considers the coverage of the evader's acceleration capability range by the pursuer using biased Proportional Navigation Guidance (BPN) with specific bias terms. Subsequently, an online cooperative capture strategy is proposed, which adaptively adjusts a limited set of bias terms to achieve comprehensive coverage of the evader's acceleration capabilities and ensure successful capture in general three-pursuer, single-evader scenarios. To implement this strategy, the framework incorporates a two-phase learning approach: a pre-training phase based on Augmented Proportional Navigation guidance (APN) followed by a fine-tuning phase using Multi-Agent Proximal Policy Optimization (MAPPO). Numerical simulations demonstrate that the proposed strategy offers higher success rates and broader coverage of the evader's acceleration capabilities compared to traditional APN-based methods, ensuring effective capture against various evasion maneuvers.
KW - Convergence set
KW - Cooperative capture
KW - Coverage-based guidance law
KW - Deep reinforcement learning
KW - Multi agent systems
KW - Three-pursuers and single-evader game
UR - https://www.scopus.com/pages/publications/105052638403
U2 - 10.1016/j.cja.2026.104151
DO - 10.1016/j.cja.2026.104151
M3 - 文章
AN - SCOPUS:105052638403
SN - 1000-9361
VL - 39
JO - Chinese Journal of Aeronautics
JF - Chinese Journal of Aeronautics
IS - 11
M1 - 104151
ER -