Motion planners often rely on straight-line steering because it is computationally cheap, but it wastes samples by rejecting many colliding extensions. Control-barrier-function (CBF) steering is more sample-efficient but requires expensive sequential evaluations. The new method, FLASH-CBF-RRT, computes a time horizon over which a CBF-filtered velocity can be applied constantly, using closed-form kinematic bounds on collision clearance.
This allows the planner to take adaptive step sizes: shorter near obstacles and longer in open space, which sharply reduces the number of evaluations needed. The authors implemented the approach as an extension of RRT-Connect in OMPL and evaluated it on 689 MotionBenchMaker problems.
FLASH-CBF-RRT matched RRT-Connect's 100% success rate while reducing median planning time by 3.82x, and was 3.88x faster than fixed-step CBF-RRT. In hardware experiments on a 17-DoF mobile manipulator, it delivered 2.2x faster planning and 4.2x fewer evaluations. The paper is a single-source report, so no independent comparisons or conflicting results are noted.