Event

SCET-CBF: Speed-Calibrated Soft-Guidance Single-Step Dynamic Navigation

1 reports · 1 sources

Xiaoyu Han of University College London and Dongkun Han of the Chinese University of Hong Kong submitted a preprint proposing SCET-CBF, a speed-calibrated soft-guidance law for single-step dynamic navigation. The method targets a steering-constrained bicycle model: although braking can reduce approach speed, it also slows the lateral motion needed before an obstacle is reached; SCET-CBF derives the permissible approach speed via a short lateral escape maneuver and lets a quadratic program weigh these two effects of braking when choosing acceleration and steering. In tests with 2,500 paired tasks and a 60-second time limit, SCET achieved a 72.8% completion rate, CBF/velocity-obstacle control achieved 49.8%, and an augmented 1-second CBF-MPC achieved 69.8%. The paper also evaluates lateral acceleration limits and the applicable range under bounded motion and actuation errors.