关键词: 避障约束, 攻击角度约束, Bézier曲线, 强化学习控制, 鲁棒平行最优控制

Abstract: For high-value offensive precision strike missions in complex battlefield environments, a robust parallel optimal guidance method considering obstacle avoidance and impact angle constraints is proposed in this paper. First, to accurately characterize obstacle features, a three-dimensional cylinder–hemisphere obstacle model is established, and the LogSumExp function is employed to obtain a smooth representation of obstacle avoidance constraints. Furthermore, an optimization problem incorporating trajectory length and smoothness indices is formulated, and an optimal trajectory satisfying both obstacle avoidance and terminal impact angle constraints is generated based on Bézier curves. On this basis, the optimal trajectory is taken as the reference trajectory, and a robust parallel optimal guidance method with composite learning capability is developed by combining the idea of single-network reinforcement learning. By utilizing online stored historical data, the proposed method relaxes the persistent excitation condition and eliminates the dependence on an initial admissible control policy. The uniform ultimate boundedness of the closed-loop system is proved through Lyapunov stability theory. Simulation results show that the proposed method can obtain a safe, smooth, and feasible trajectory with the desired terminal impact angle in complex obstacle environments, while achieving high-precision tracking of the reference trajectory, thereby verifying its effectiveness in complex constrained scenarios.

Key words: obstacle avoidance constraint, impact angle constraint, Bézier curve, reinforcement learning control, robust parallel optimal control