A Multi-Strategy Improved Differential Evolution algorithm for UAV 3D trajectory planning in complex mountainous environments

• Published in: Engineering applications of artificial intelligence 2023-10, Vol.125, p.106672, Article 106672
• Main Authors: Zhang, Miaohan, Han, Yuhang, Chen, Shiyun, Liu, Mingxian, He, Zhaolei, Pan, Nan
• Format: Article
• Language: English
• Subjects: Differential Evolution algorithm; Mountainous areas; Multi-strategy improvement; Trajectory planning; UAV
• ISSN: 0952-1976
• DOI: 10.1016/j.engappai.2023.106672

In response to the complexity of power repair in mountainous areas and the limitations of traditional vehicles due to terrain constraints, this study focuses on the three-dimensional trajectory planning problem of UAVs (Unmanned Aerial Vehicles) in mountainous environments. Our goal is to provide effective solutions for the trajectory planning problem of UAVs in mountainous environments. Firstly, a UAV trajectory planning model is established, incorporating optimization objectives such as energy consumption, trajectory cost, obstacle avoidance cost, smoothing cost, and stability cost. The trajectory planning problem is transformed into an objective function optimization task with multiple performance constraints. To overcome the inefficiency and infeasibility of traditional algorithms in solving complex three-dimensional flight environments, we propose improvements to the Differential Evolution (DE) algorithm through three strategies: incorporating mutation crossover factor optimization strategy, an adaptive guidance mechanism, and an elite disturbance mechanism based on population classification. The Multi-Strategy Improved Differential Evolution (MSIDE) algorithm is introduced, and its time and space complexity are analyzed. Finally, the proposed method is compared with various algorithms through benchmark functions tests, Friedman test, Wilcoxon rank-sum test, simulation experiments in three-dimensional environments, and parameter sensitivity analysis experiments. The simulation results show that compared with the current state-of-the-art algorithms, the MSIDE algorithm improves the objective function value by 11.34% on average in regular terrain and 5.04% on average in complex terrain environments. The results demonstrate the convergence, multi-objective search capability, and global search ability of MSIDE, validating its effectiveness in solving the trajectory planning problem of UAVs in complex mountainous environments.