To address the challenges of energy-efficient trajectory planning and experimental validation for solar-powered unmanned aerial vehicles(UAVs),this paper introduces a novel trajectory planning method leveraging deep reinforcement learning.A ground-based hardware-in-the-loop simulation platform has also been established for validation.Initially,the kinematic,dynamic,energy acquisition,and consumption models of solar-powered UAVs are established.The coupling effects between flight states,such as time,attitude angles,and energy states,are thoroughly analyzed.Subsequently,an energy-optimal trajectory planning framework is formulated,and agent training is conducted using the twin delayed deep deterministic policy gradient algorithm.Effectiveness simulations are conducted to analyze the performance.Finally,a ground-based hardware-in-the-loop testing platform has been established.After thorough performance testing of the platform,the reinforcement learning controller is deployed online for hardware-in-the-loop simulation testing.The results demonstrate that compared to traditional flight strategy,the proposed method trained strategy that increased energy acquisition by 2.9%,reduced energy consumption by 30.46%,and increased energy accumulation by 11.36%within 200 seconds of flight time.The proposed strategy can more fully utilize solar energy,reduce flight power requirements,achieve the goal of increasing energy efficiency,and provide reference for high-efficiency solar-powered UAVs flights.
关键词
太阳能无人机/高能效飞行/深度强化学习/双延迟深度确定性策略梯度/航迹规划/半实物仿真
Key words
Solar-powered UAVs/High Energy Efficiency Flight/Deep Reinforcement Learning/Twin Delayed Deep Deterministic Policy Gradient/Trajectory Planning/HIL Simulation
维普
万方数据