PD-Bug: A Novel Motion Planning Algorithm for Nonholonomic Car-Like Robots in Unknown Environments

In this work, we propose a novel motion planning algorithm for nonholonomic car-like robots in unknown environments. The proposed algorithm termed Polynomial spiral and Dubins path Bug (PD-Bug) is developed as a variant of TangentBug. Considering the errors of range measurements and the shape of a robot, an environment model termed local target set (LTS) is established to provide an optimal local target during each planning period. For localization errors, a more robust motion mode switching mechanism is integrated into PD-Bug. A lookup table for cubic polynomial spirals is generated offline, providing a smooth and optimized trajectory with curvature constraints. The Dubins path is generated online as the sub-optimal solution and utilized to meet the orientation requirement of the target. We validate the applicability of PD-Bug in outdoor scenarios with real-time requirements. Compared with the traditional Bug algorithms, PD-Bug has a significant improvement in the robustness of the motion mode and the smoothness of the trajectory.