PATH PLANNING FOR GRAIN HARVESTERS BASED ON THE VS-IRRT ALGORITHM

To address the problems of slow path planning speed, high path cost, and visual positioning errors encountered by grain harvesters during field operations, this study proposes an improved rapidly-exploring random tree algorithm integrated with visual servoing (VS-IRRT). By employing visual servoing technology to acquire environmental information in real time, the algorithm enables accurate positioning and attitude correction of the harvester. On this basis, heuristic sampling strategies and a path optimization function are introduced to enhance node expansion efficiency and accelerate the convergence of the search tree. To further reduce path cost, a path evaluation model incorporating environmental feature costs is established, which comprehensively considers terrain complexity, crop distribution density, and the machine’s turning radius. This model dynamically adjusts the search direction and improves path smoothness. Simulation and field navigation experiment results indicate that the VS-IRRT algorithm reduces path planning time by approximately 32% compared to the traditional RRT algorithm, decreases the average yaw error by 42%, reduces the path curvature variation rate by 33%, and lowers turning frequency by 21%. The algorithm also maintains high robustness and planning accuracy under visual noise and positioning disturbances. Overall, this study provides an effective path planning approach and technical support for autonomous navigation and efficient operation of grain harvesters in complex agricultural environments.