Research on dynamic multi-constraint path planning of carrier-based aircraft based on deep reinforcement learning

Objective Most existing path planning methods for carrier-based aircraft fail to account for the practical spatial constraints encountered during their transfer process and have difficulty adapting to the highly dynamic conditions on the deck. To address these limitations, this paper proposes a dynamic path planning algorithm for carrier-based aircraft that comprehensively considers positional and kinematic constraints and desired final heading orientations.

Method Initially, the geometric shape of the carrier-based aircraft is modeled using the polygon method. A kinematic model is then formulated based on parameters such as the aircraft's movement speed and heading angle. Subsequently, the path planning problem for the carrier-based aircraft is formulated as a Markov decision process (MDP). The action and state spaces are defined based on the aircraft's motion characteristics. A reward function is designed by incorporating factors such as position, orientation, safety, and efficiency. A deep reinforcement learning-based path planning algorithm for carrier-based aircraft is then proposed. Finally, simulations are conducted to validate the effectiveness of the proposed algorithm.

Results The results demonstrate that, compared to traditional algorithms, the proposed algorithm reduces scheduling time by an average of 9.2% and decreases target heading angle error by an average of 98.7%.

Conclusion The proposed method effectively improves the transfer efficiency of carrier-based aircraft and provides valuable insights for decision-making in aircraft coordination and deck operations.