An Optimization Method for Accelerating UAV Trajectory Planning

Abstract

Abstract Trajectory planning is a very important task in the flight of UAVs. Through trajectory planning, the autonomous flight of UAVs can be realised and the flight efficiency and safety can be improved. Particle swarm algorithm is a commonly used trajectory planning method, but there are some shortcomings in practical applications, such as easy to fall into local minimal values and other problems. Therefore, previous research on particle swarm trajectory planning has been carried out and some improved algorithms have been proposed. However, there are still some problems with these algorithms, such as excessive computational power and slow convergence speed. To address these problems, this paper proposes a method for deep residual learning to optimise the problem of local minima and slow convergence of traditional particle swarm trajectory planning. Specifically, a complete mathematical model for UAV path planning is established in this paper, and path planning results are obtained using the conventional improved particle swarm algorithm by means of cubic spline interpolation. Thereafter, to address the problems that path planning based on the particle swarm optimisation algorithm tends to reach local optimality at a later stage and the slow convergence speed of the improved algorithm, this paper introduces a deep residual learning network to improve the convergence speed and stability of the particle swarm algorithm. Finally, the performance of the algorithm before and after the improvement is compared according to four benchmark test functions. The simulation results show that the improved particle swarm algorithm has a significant enhancement in the late search capability.