Phase-Adaptive Reinforcement Learning for Self-Tuning PID Control of Cruise Missiles

Conventional fixed-gain PID controllers face inherent limitations in maintaining optimal performance across the diverse and dynamic flight phases of cruise missiles. To overcome these challenges, we propose Time-Fusion Proximal Policy Optimization (TF-PPO), a novel adaptive reinforcement learning framework designed specifically for cruise missile control. TF-PPO synergistically integrates Long Short-Term Memory (LSTM) networks for enhanced temporal state perception and phase-specific reward engineering enabling self-evolution of PID parameters. Extensive hardware-in-the-loop experiments tailored to cruise missile dynamics demonstrate that TF-PPO achieves a 36.3% improvement in control accuracy over conventional PID methods. The proposed framework provides a robust, high-precision adaptive control solution capable of enhancing the performance of cruise missile systems under varying operational.