Summary
Robots sometimes remember past mistakes but struggle to tell if those memories still apply or if they clash with what they see now. The authors present CoreSense, a way to combine past evidence with current observations to decide whether to continue, recheck, or stop and ask for help. Their system checks details like timing and contradictions to make safer decisions and records all steps for auditing. Tests using robot data and simulations showed CoreSense stopped unsafe actions completely while reducing unnecessary stopping compared to other methods. The system focuses on supporting auditability rather than guaranteeing perfect safety or recovery.
What this means in practice
- •For robot system integrators: Integrate a module that audits robot decisions by recalling past failures while avoiding unsafe or contradictory actions.
- •For industrial automation teams: Implement conflict-aware decision gating to reduce unsafe robot operations by combining live data with historic failure memories.
Abstract
Robots can recall prior failures without knowing whether recalled evidence remains valid, conflicts with current observations, or is sufficient to guide a decision. We present CoreSense, a robot-system integration architecture that combines traceable episodic evidence with a conflict-aware belief gate and bounded, auditable recommendations. The gate checks scope, provenance, time, contradiction, and support before it permits PROCEED, requests re-observation, abstains, or escalates. Evaluation follows three complementary layers without commanding a physical robot: offline public real-robot data, a frozen signal-level simulation, and a live cloud deployment path. On CableTrace-120 and BotFails-200, belief gating reduces protocol-defined unsafe proceeds from 20% and 40% to 0%. A disjointly calibrated raw-video policy also reaches 0% unsafe proceed, but overblocks every nominal episode. On public data, a ViFailback-BotFails visual detector reaches 0.778 AUROC yet remains all-blocking, whereas cycle-disjoint UR3 telemetry for protective stops yields 0% unsafe proceed, 36.1% overblocking, and 61.9% coverage; grip-loss transfer remains a negative result. Controlled physical corroboration yields 3.3%, 0%, and 42.0%, while conflict-aware fusion yields 4.7%, 0%, and 42.8%. Finally, 20/20 cloud recalls validate a CockroachDB Cloud-Amazon Bedrock deployment path. The evidence supports an auditable integration pattern, not autonomous recovery or certified safety.