Adaptive Risk-Certified Event-Triggered Replanning for Dynamic Navigation

Key points

• Safe navigation in dynamic environments requires robots to plan under obstacle predictions whose errors are uncertain, non-stationary, and can induce rare but safety-critical failures.1
• Existing control-barrier-function safety filters can reject immediately unsafe controls, but they provide little guidance on when the current finite-horizon planning mode itself is becoming unsafe as prediction uncertainty evolves.1
• CERT-Replan calibrates horizon-indexed obstacle-prediction residuals online and uses the resulting uncertainty radii to evaluate dynamic-obstacle safety margins.1
• In a non-stationary benchmark, CERT-Replan achieves an \(83.3%\) collision reduction relative to the safety-filter-only baseline \textcolor{black}{and a \(77.8%\) reduction relative to simple replanning triggers}, while reducing average safety-filter intervention by \(32.4%\). \textcolor{black}{With Trajectron++, CERT-Replan achieves \(96%\) collision-free operation.1

Sources (1)

• [1]Adaptive Risk-Certified Event-Triggered Replanning for Dynamic Navigation
  arXiv (AI, ML, NLP, CV, robotics, multi-agent) · Oct 7, 02:02 AM

  We propose Conformal Event-Triggered Risk-Certified Replanning (CERT-Replan), a framework that uses calibrated barrier risk as an early-warning signal for replanning.

  Safe navigation in dynamic environments requires robots to plan under obstacle predictions whose errors are uncertain, non-stationary, and can induce rare but safety-critical failures.

Extractive summary: sentences quoted from the sources.