多无人机协同除草任务分配与路径优化方法研究

Research on Task Allocation and Path Optimization Methods for Multi-UAV Cooperative Weed Control Operations

摘要: 针对大田环境下多无人机协同除草的效率低、能耗高及农药浪费问题，提出一种融合农田栅格化、杂草区聚类与改进离散粒子群算法(DPSO)的任务分配与路径优化方法。首先，通过栅格化处理将农田映射为规则数值矩阵，结合密度聚类(HDBSCAN)算法对杂草区进行空间分区，平衡区域内杂草区密度与无人机负载；其次，构建考虑无人机电量与药箱容量约束的改进车辆路径问题(VRP)模型，以最小化系统总能耗为优化目标；最后，设计含排斥算子与逆转变异策略的DPSO算法，分两级完成区域间路径规划、区域内杂草区访问顺序优化。以大疆T40无人机为实验对象，在1000 m × 1000 m模拟农田中进行测试，经实验证明，本文方法能够实现“直奔杂草区域”的精准作业，最终达成节能、高效、环保的除草目标。

Abstract: To address the issues of low efficiency, high energy consumption, and pesticide waste in multi-UAV cooperative weed control in large-scale field environments, this study proposes a task allocation and path optimization method integrating farmland rasterization, weed zone clustering, and an improved Discrete Particle Swarm Optimization (DPSO) algorithm. First, the farmland is mapped into a regular numerical matrix through grid processing. Combined with the HDBSCAN algorithm, weed areas are spatially partitioned to balance weed density within regions and drone payload capacity. Second, an improved Vehicle Routing Problem (VRP) model is constructed, considering drone battery constraints and spray tank capacity, with the optimization goal of minimizing total system energy consumption. Finally, a DPSO algorithm incorporating rejection operators and inverse mutation strategies was designed to perform two-stage path planning: inter-area route optimization and intra-area weed zone access sequence optimization. Using the DJI T40 drone as the experimental subject in a 1000 m × 1000 m simulated farmland, tests demonstrated that this method achieves precise “straight-to-weed-zone” operations, ultimately fulfilling energy-saving, efficient, and environmentally friendly weed control objectives.