Trajectory planning of load transportation with multi-quadrotors based on reinforcement learning algorithm

Abstract

Unmanned aerial vehicles (UAVs) are playing more and more vital roles in transportation. In this work, a trajectory planning method based on a value-function approximation algorithm is proposed to address the problem of cable-suspended load transportation using three quadrotors. The purpose of trajectory planning is to reach the target position as soon as possible without the load swinging. In the proposed algorithm, the value function is approximated by parameter vector and problem-specific feature vector. The parameter vector is updated by batch method, and then the greedy strategy is used to generate the desired trajectory. Finally, the desired trajectory is passed to the quad-rotor controller for trajectory tracking. The simulation results show that the trained parameters can fit the value function. The suspended load can be transported smoothly from starting location to the different target locations. The value function in the greedy strategy finally converges to the maximum value, which proves the effectiveness and convergence of the proposed algorithm.