Event
Online Sensor Diagnosis and Repair for Black-Box Robot Policy Failures
1 reports · 1 sources
Marc Ziegler and co-authors propose a structured surrogate model that diagnoses and repairs damaged sensor inputs outside a frozen locomotion policy, separating the periodic gait from sensor-dependent correction terms. The authors report that, on a simulated Go2, offline diagnosis raised walking success from 259/640 to 429/640, online repair reached 382/640, and none of the 64 fault-free held-out episodes triggered repair. The authors also note that the evidence remains limited to a single policy and a partially screened simulated fault set, and that robot foundation models and hardware validation have not yet been carried out.
Report timeline
- When Black-Box Robot Policies Fail: Online Sensor Diagnosis and Repair with a Transparent Surrogate
The study proposes a structured surrogate model that separates periodic gait from sensor-dependent correction terms, diagnosing and repairing corrupted inputs externally around a frozen locomotion policy. On simulated Go2, offline diagnosis raises walking success from 259/640 to 429/640, while online repair reaches 382/640, with no repair triggered in 64 held-out fault-free episodes. The authors note that the evidence remains limited to a single policy and a partly screened simulated fault set, and that robot foundation model and hardware validation have not yet been conducted.
arXiv cs.RO ·