LEAP-CBF: A Safety Filter for Uncertain Systems with Least-Effort Adversarial Potentials
Researchers propose LEAP certificates as robust control-barrier safety filters for uncertain robotic systems.
Researchers introduce Least-Effort Adversarial Potentials (LEAP), a certificate of how much disturbance effort is required to drive a state to failure. LEAP is a control barrier function for the undisturbed system and can form a safety filter robust to disturbances with bounded cumulative effort. The authors learn LEAPs with on-policy deep reinforcement learning, evaluate them on simulated multi-agent systems, and validate them on a quadruped and quadrotors.
- LEAP measures disturbance effort required to cause failure.
- The filter is robust when cumulative disturbance effort is bounded.
- Certificates are constructed with on-policy deep reinforcement learning.
- Hardware tests include a quadruped and quadrotors.
From the original article
1 min read at arxiv.orgControl barrier functions (CBF) are a popular safety filter to ensure safety for nonlinear dynamical systems. However, when the system is subject to uncertainties and disturbances, this requires the use of robust variants of CBFs, which can be difficult to construct and can be overly conservative, especially for high-dimensional systems under input constraints.
By Oswin So · arxiv.org
A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.