Interval POMDP Shielding for Imperfect-Perception Agents
This addresses safety concerns for autonomous systems with imperfect perception, though it is incremental as it builds on existing shielding methods.
The paper tackles the problem of unsafe decisions in autonomous systems due to misclassified sensor readings by introducing a runtime shield that blocks actions violating safety, achieving improved safety over state-of-the-art baselines in experiments on four case studies.
Autonomous systems that rely on learned perception can make unsafe decisions when sensor readings are misclassified. We study shielding for this setting: given a proposed action, a shield blocks actions that could violate safety. We consider the common case where system dynamics are known but perception uncertainty must be estimated from finite labeled data. From these data we build confidence intervals for the probabilities of perception outcomes and use them to model the system as a finite Interval Partially Observable Markov Decision Process with discrete states and actions. We then propose an algorithm to compute a conservative set of beliefs over the underlying state that is consistent with the observations seen so far. This enables us to construct a runtime shield that comes with a finite-horizon guarantee: with high probability over the training data, if the true perception uncertainty rates lie within the learned intervals, then every action admitted by the shield satisfies a stated lower bound on safety. Experiments on four case studies show that our shielding approach (and variants derived from it) improves the safety of the system over state-of-the-art baselines.