Quadratic optimization-based path planning for autonomous agricultural vehicles headland turning in orchard environments

Automated headland turning is a critical field operation for autonomous agricultural vehicles in agricultural production. Although traditional model-based and graph search-based path planning methods work well in arable farming, they struggle to generate a smooth, and collision free turning trajectory in complex and dynamically changing orchard headlands in real-time. To address this challenge, this paper proposed a quadratic optimization-based path planning approach for headland turning of four-wheeled agricultural vehicles in orchards. Firstly, when the vehicle arrives at the headland, an initial reference path is generated using an improved Reeds Shepp curve based on the vehicle's position and heading angle, and the span width between initial and terminal points. Secondly, projection discretization, point sampling, and real-time dynamic planning are performed to obtain a preliminary obstacle-free path. Finally, a quadratic programming solver is designed that uses the constructed convex space to obtain the final path. Simulations and field experiments (in a digital peach orchard) were conducted to evaluate the performance of the proposed method by comparing it with an improved artificial potential field (APF) method in various types of headlands. Simulation and experimental results show that, when obstacles are present at the initial, middle, and terminal sections of the headland, the path generated by APF results in collisions. However, the proposed method generates smooth, collision-free paths, implying that the proposed method demonstrates a higher success rate in path planning. In addition, compared with APF method, the proposed method achieved a 6.69% reduction in path length, a 5.87% decrease in the time duration, and a 35.29% decrease in path curvature in field experiments. In conclusion, the proposed method can plan smooth, collision-free headland paths, significantly improving the operation safety and efficiency of unmanned agricultural unmanned vehicles during headland turns.