Carrier-Based Aircraft Arrested Landing Control Method Based on Improved Deep Deterministic Policy Gradient Algorithm

Objectives For the automatic guidance of carrier-based aircraft arrested landing under multi-factor coupled disturbances, a phased guidance method based on deep reinforcement learning is proposed. Methods Firstly, the aircraft landing guidance process is modeled as a phased Markov Decision Process under multi-factor coupled disturbances. Secondly, considering key factors such as aircraft attitude, dynamics, and deck motion, a multi-phase, multi-dimensional information dynamic combination reward function is designed. To address the sparse reward problem in the landing guidance process, a Deep Deterministic Policy Gradient algorithm incorporating a potential function-based reward shaping mechanism is proposed. Finally, a simulation environment for carrier-based aircraft landing is constructed, integrating functions such as aircraft dynamic modeling, environmental disturbance modeling, heterogeneous visualization, and real-time interactive control. The comprehensive performance of the proposed algorithm is verified through simulation experiments. Results The experimental results show that the landing guidance method based on the improved Deep Deterministic Policy Gradient algorithm outperforms baseline algorithms in terms of convergence speed, landing success rate, landing stability, and disturbance resistance. Conclusions The research findings can provide references for deep reinforcement learning-based guidance technologies for carrier-based aircraft arrested landing.