Reinforcement learning-based enhanced fixed-time guidance law with no-fly zone constraints

This paper proposes a reinforcement learning-based enhanced fixed-time guidance law for attacking maneuvering targets in three-dimensional scenarios under no-fly zone constraints. Firstly, an enhanced nonsingular fixed-time guidance law is proposed to guarantee faster convergence of the line-of-sight angles and angular rates without considering no-fly zones, wherein a carefully designed sliding surface featuring an adjustable power term is explored to derive a smaller upper bound on the settling time. Secondly, to enforce safety-critical attacking under no-fly zone constraints, a reinforcement learning-based enhanced fixed-time guidance law is tailored by incorporating the deep deterministic policy gradient to generate safety guaranteed accelerations, in which the overload consumption, guidance accuracy, and no-fly zone avoidance constraints are interpreted as reward functions. The key innovation lies in formulating an optimal intelligent guidance law that synergistically balances no-fly zone avoidance and guidance performance. Finally, extensive simulation experiments are conducted to verify the feasibility and superiority of the proposed algorithm.