基于改进RRT与GA的多目标路径规划——以无人机林区巡检为例  

Multi-objective path planning based on improved RRT and GA:taking UAV forest inspection as an example

作　　者：张彪 康峰[1] 许舒婷 Zhang Biao;Kang Feng;Xu Shuting(School of Technology,Beijing Forestry University,Beijing 100083,China)

机构地区：[1]北京林业大学工学院,北京100083

出　　处：《北京林业大学学报》2025年第4期129-141,共13页Journal of Beijing Forestry University

基　　金：国家自然科学基金项目(12402413);中央高校基本科研业务启动基金项目(BLX202224)。

摘　　要：【目的】为解决无人机在人工林区巡检任务(如病虫害监测、火灾预防等)中的路径规划问题,即求解巡检点的最优遍历序列以及生成避障飞行轨迹,本文通过融合改进快速随机扩展树(RRT)算法和遗传算法(GA),提出一种多目标路径规划算法。【方法】首先改进传统GA,使其能够在三维空间中遍历所有巡检点并求解最优序列。其次,依据该序列进行路径搜索,改进RRT算法的随机采样原理,通过靶心和绕树策略实现避障效果,并采用连续选择父节点策略,取消因避障产生的多余转折点。最后,通过3次B样条曲线优化,生成最终路径。【结果】仿真结果表明,本算法能够在复杂林区环境中遍历所有巡检点,并在短时间内规划出高质量、无碰撞的路径。与粒子群算法(PSO)、蚁群算法(ACO)和RRT算法相比,当巡检点从3个增加到9个时,PSO、ACO、RRT算法搜索时间分别增加了221.77%、332.42%、184.78%,而本算法仅增加了102.35%。在9个巡检点的复杂环境中,本算法的路径耗散分别比PSO、ACO和RRT算法降低了14.46%、30.28%、24.76%,且路径质量显著提高,消除了路径交叉重合现象。此外,通过ROS平台,利用无人机在林区点云上进行模拟飞行并验证成功,证明本算法适用于林区巡检的多目标路径规划。【结论】针对人工林区无人机巡检任务中的飞行路线规划问题,本文通过改进RRT与GA,成功规划出一条遍历所有巡检点且避开林区障碍物的无碰撞路径。相较于PSO、ACO和RRT算法,本算法在路径质量、路径耗散和搜索时间上均表现出显著优势。[Objective]To address the path planning problem of UAVs in plantation areas for inspection tasks(such as pest and disease monitoring and fire prevention),which involves solving the optimal traversal sequence of inspection points and generating collision-free flight trajectories,this paper proposes a multiobjective path planning algorithm by integrating and improving the rapidly-exploring random tree(RRT)algorithm and genetic algorithm(GA).[Method]First,the traditional GA was improved to enable traversal of all inspection points in 3D space and solve the optimal sequence.Second,based on this sequence,the path search was conducted by improving the random sampling principle of RRT algorithm.Obstacle avoidance was achieved through target and tree-avoidance strategies,and redundant turning points generated by obstacle avoidance were eliminated by continuously selecting parent nodes.Finally,the final path was generated through three iterations of B-spline curve optimization.[Result]Simulation results showed that the proposed algorithm can traverse all inspection points and plan high-quality,collision-free paths in complex forest environments within a short time.Compared with particle swarm optimization(PSO),ant colony optimization(ACO),and RRT algorithms,when the number of inspection points increased from 3 to 9,the search times of PSO,ACO,and RRT algorithms increased by 221.77%,332.42%,and 184.78%,respectively,while the proposed algorithm only increased by 102.35%.In a complex environment with 9 inspection points,the path cost of proposed algorithm was reduced by 14.46%,30.28%,and 24.76%compared with PSO,ACO,and RRT algorithms,respectively.The path quality was significantly improved,eliminating path crossing and overlap.Additionally,the algorithm was successfully validated through simulation flights on forest point clouds using a UAV on the ROS platform,demonstrating its applicability for multi-objective path planning in forest inspections.[Conclusion]For the path planning problem of UAVs in artificial forest inspecti

关 键 词：多目标优化 路径规划 快速随机扩展树(RRT) 遗传算法(GA) 无人机 粒子群算法(PSO) 蚁群算法(ACO) 

分 类 号：TP273[自动化与计算机技术—检测技术与自动化装置] V279[自动化与计算机技术—控制科学与工程] S719[航空宇航科学与技术—飞行器设计]