Approximate Optimal Control for Nonlinear Multi-Agent Cooperative Pursuit-Evasion Games Using Single-Network ADP

Aming to the difficulty in obtaining analytical control strategy in non-linear cooperative multi-agent pursuit escape games(MPE) using the Nash equilibrium method, an approximate optimal control strategy is proposed in this paper. Instead of the traditional actor-critic dual networks structure, each pursuer agent employs a single-critic neural network(NN) framework combined with approximate dynamic programming(ADP) to achieve near-optimal control strategy, such that the computational burden can be significantly reduced. Moreover, a new distributed operator is introduced in weight tuning law such that the control strategy doesn't need initial stability. Finally, the effectiveness of the approximate optimal control scheme is validated through a numerial simulation.