ResearchResearch paperRobotics & Embodied AI · Safety & Alignment · Reinforcement Learning1 source · Oct 14, 2026

BubbleCBF: Constructive Safety for Second-Order Systems and Application to Autonomous Robot Navigation in Unknown Environments

We develop a constructive approach for designing feasible safety constraints for second-order control-affine systems.

Key points

  • Satisfying safety constraints for systems with second-order dynamics is challenging because a position constraint may require infeasible deceleration.
  • Starting from a signed distance function (SDF), we construct collision-free spheres, i.e., bubbles, in position space, and use their quadratic geometry and overlap to construct a velocity-dependent safety constraint.
  • Under mild assumptions on the dynamics, bubble overlap, and design parameters, we prove that the quadratic-program-based controller is well defined, continuous in the system state, and renders the set forward invariant.
  • We demonstrate that our BubbleCBF, applied to a differential-drive robot, achieves safe autonomous navigation in unknown environments.

Sources (1)

Extractive summary: sentences quoted from the sources.

Related