Deep Learning Based Optimization of UAV Orbital Paths

UAV path planning is an important part of UAV technology, and genetic algorithms and ant colony algorithms are used for it in UAV path planning tasks. When facing some unknown environments, the performance of classical algorithms is less satisfactory. For the path planning problem of a single UAV, this paper proposes the APF-DDQN model using the artificial potential field method combined with the multi-intelligence body algorithm DDQN. The model utilizes the dominance function as a metric for selecting action values and measuring the goodness of the tie-breaking values of all actions.

I. Introduction

As a kind of flexible, low-cost flying aircraft with excellent scalability, UAVs have demonstrated their superior performance in both military and civil fields, and are widely used in environmental monitoring, disaster rescue, film and television aerial photography, express transportation, and other practical scenarios. The extremely high application value makes the in-depth exploration and research and development of unmanned aircraft to obtain a large amount of cost investment and control research.