Deep Reinforcement Learning-based Swing-Free Trajectories Planning Algorithm for UAV with a Suspended Load

UAV transportation has great application prospects, and can be used to transport some materials that need to be delivered quickly, such as food, medicine, and even organ transplantation, in the face of many special situations available using rope suspended load, but, when using suspension transport, the swing of the load will increase the instability of the flight. To address this situation, we propose a method based on Deep Q-Networks to generate the minimum swing trajectory of UAV suspended load. First, a training environment is established using the kinematic model of UAV suspended load to discretize the action, which will greatly accelerate the training speed; Finally, a step-by-step regional approximation exploration reward function is designed, and the final training simulation results verify the feasibility of the method, and the suspended load oscillation is controlled within a certain range during the whole transportation process.