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

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 multi-objective 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 inspections, the proposed algorithm successfully planned a collision-free path that traversed all inspection points while avoiding obstacles in the forest. Compared with PSO, ACO, and RRT algorithms, the proposed algorithm shows significant advantages in path quality, path cost, and search time.