Authors

Abstract

To address the challenges of complex farmland environment, an intelligent obstacle avoidance control algorithm for agricultural unmanned aerial vehicles (UAVs) is developed. The objective is to solve the problem of efficient obstacle avoidance in farmland scenarios characterized by dense dynamic obstacles and variable terrain. In this article, a target detection algorithm based on improved YOLOv5 (You Only Look Once v5) is proposed, and an intelligent obstacle avoidance system is constructed by combining reinforcement learning path planning and adaptive motion control strategy. Ghost module is introduced to improve the lightweight of YOLOv5, and the design of CIOU (Complete Intersection Over Union) loss function is optimized, which improves the detection accuracy of the model for small targets and dynamic obstacles. Experiments show that the error of path planning is reduced to less than 2.1 meters, and the time consumption is reduced by about 35%. In addition, fuzzy logic controller is used to realize adaptive PID control, which further enhances the flight stability of UAV in complex environment. The results show that the improved algorithm has excellent performance in many typical farmland scenes. This study provides theoretical and technical support for autonomous flight of agricultural UAV in complex farmland environment.

Abstract in Chinese

面向复杂农田环境，农业无人机智能避障控制算法的研究目的是解决其在动态障碍物密集、地形多变的农田场景中实现高效避障的难题。本文提出一种基于改进YOLOv5（You Only Look Once v5）的目标检测算法，并融合强化学习路径规划与自适应运动控制策略，构建了一套智能避障系统。研究中引入Ghost模块对YOLOv5进行轻量化改进，并优化了CIoU（Complete Intersection over Union）损失函数的设计，从而提高了模型对小目标及动态障碍物的检测精度。实验表明，路径规划误差控制在2.1米以内，规划耗时减少了约35%。此外，通过采用模糊逻辑控制器实现自适应PID控制，进一步提升了无人机在复杂环境中的飞行稳定性。研究结果证实，所提出的改进算法在多种典型农田场景下均表现出优异性能。本研究为农业无人机在复杂农田环境中的自主飞行提供了有力的理论支持与技术保障。