Trajectory planning for drone landing, incorporating wind-sensing capabilities, operational and safety objectives, and reinforcement learning

Drone flight safety and operational efficiency are challenged in the landing phase, especially in windy conditions. While flight control should be a key focus, flight trajectory also plays a critical role in landing. Inspired by the multiple objectives, wind sensory capability, and skill learning of avian species, this study proposes a reinforcement learning-based trajectory planner for a drone to perform trajectory planning in the wind, aiming to balance multiple landing-related objectives based on onboard wind sensory capability and address the safety-operational efficiency dilemma of a landing site. Through four key experiments, this study demonstrates successful training, balanced landing performance, and strong generalization capability of the trajectory planner. The experiments highlight the importance of velocity sensory capability while indicating that wind sensory capability is less critical to the trajectory planner. The proposed framework with multiple objectives, wind sensory capability, and skill learning can benefit applications such as improving drone performance.