Collision-Free and Curvature-Continuous Path Smoothing in Cluttered Environments
Abstract¶
We present a novel trajectory computation algorithm to smoothen jerky collision-free paths computed by samplebased motion planners. Our approach uses cubic B-splines to generate G2 or curvature continuous trajectories. The algorithm performs local spline refinement to compute collision-free trajectories in narrow passages and satisfies velocity and acceleration constraints. We also present a fast and reliable algorithm for collision checking between robot and the environment along the B-spline trajectories. We highlight the performance of our algorithm on complex benchmarks, including path computation for rigid and articulated models in tight spaces and narrow passages.
Summary¶
Turns jagged collision-free paths from sampling-based planners into curvature-continuous B-spline trajectories. The method locally refines splines to handle narrow passages, checks robot-environment collisions along curved paths efficiently, and enforces velocity and acceleration constraints, making post-processing more suitable for real robots.