Optimal motion planning and navigation for nonholonomic agricultural robots in multi-constraint and multi-task environments

Agricultural robot navigation requires not only point-to-point movement but also seamless multi-task switching and high-precision operations. The varying constraints at different stages and complex agricultural environment pose challenges for continuous motion planning and precise tracking control. To address these, we propose a comprehensive autonomous navigation framework that enables efficient multi-task transitions and ensures high-precision execution in agricultural operations. The proposed framework consists of optimal continuous motion planning and lateral-longitudinal combined control. For motion planning, a trajectory optimization problem is formulated as a convex QP using segmented Bézier curves. Considering multi-task, safety, dynamic, and waypoint constraints, the objective function is optimized to generate a safe, dynamically feasible, and energy-efficient trajectory. For tracking control, based on the error-based dynamic model of a nonholonomic agricultural robot, a decoupled analysis of lateral and longitudinal tracking error types is conducted, and a combined lateral-longitudinal controller is designed to ensure continuous tracking of position, velocity, and heading angle. We evaluated the proposed method through motion trajectory generation and tracking control experiments. Results show maximum navigation errors of 0.0400 m (lateral), 0.0596 m (longitudinal), 0.0760 m/s (velocity), and 0.0867 rad (heading angle), with RMSEs of 0.0111 m, 0.0213 m, 0.0245 m/s, and 0.0010 rad, respectively. The method ensures seamless multi-task transitions without stopping for adjustments, achieving precise control across all operational phases.