Randomised kinodynamic motion planning for an autonomous vehicle in semi-structured agricultural areas

A randomised motion planner is presented that operates within a suitable timeframe for constrained mobile robots in agricultural environment. The core of this approach relies on splitting planning into two efficient phases to reduce its computational time. The effectiveness of sampling based planners is combined with the robustness of parametric vector-valued splines. The first phase involves relaxed two-dimensional path planning using rapidly-exploring random trees (RRT). Recent advances in sampling based planning are leveraged to improve the performance of the planner. Detailed implementation of the RRT approach and parameter selection are highlighted using comprehensive analysis and simulations. Feasible continuous paths with bounded curvature for nonholonomic robots are generated using B-spline curves. Curve segment parameters are formulated with respect to vehicle specifications. Manoeuvres satisfying maximum curvature constraints and path continuity are designed based on the segment parameters. Numerical experiments are used to validate the practicality of the proposed two-phase planner in solving kinodynamic motion queries, in real-time and replanning under limited sensing conditions.