Research on the local path planning of an orchard mower based on safe corridor and quadratic programming

Path planning algorithms are challenging to implement with mobile robots in orchards due to kinematic constraints and unstructured environments with narrow and irregularly distributed obstacles.

To address these challenges and ensure operational safety, a local path planning method for orchard mowers is proposed in this study. This method accounts for the structural characteristics of the mowing operation route and utilizes a path-velocity decoupling method for local planning based on following the global reference operation route, which includes two innovations. First, a depth-first search method is used to quickly construct safe corridors and determine the detour direction, providing a convex space for the optimization algorithm. Second, we introduce piecewise jerk and curvature restriction into quadratic programming to ensure high-order continuity and curvature feasibility of the path, which reduce the difficulty of tracking control. We present a simulation and real-world evaluation of the proposed method.

The results of this approach implemented in an orchard environment show that in the detouring static obstacle scenario, compared with those of the dynamic lattice method and the improved hybrid A* algorithm, the average curvature of the trajectory of the proposed method is reduced by 2.45 and 3.11 cm ^-1, respectively; the square of the jerk is reduced by 124 and 436 m ²/s ⁶, respectively; and the average lateral errors are reduced by 0.55 cm and 4.97 cm, respectively, which significantly improves the path smoothness and facilitates tracking control. To avoid dynamic obstacles while traversing the operation route, the acceleration is varied in the range of -0.21 to 0.09 m/s ². In the orchard environment, using a search range of 40 m × 5 m and a resolution of 0.1 m, the proposed method has an average computation time of 9.6 ms. This is a significant improvement over the open space planning algorithm and reduces the average time by 12.4 ms compared to that of the dynamic lattice method, which is the same as that of the structured environment planning algorithm.

The results show that the proposed method achieves a 129% improvement in algorithmic efficiency when applied to solve the path planning problem of mower operations in an orchard environment and confirm the clear advantages of the proposed method.