Runtime-Safety-Guided Policy Repair
This work addresses safety and performance issues in learning-based control for critical applications, representing an incremental improvement over existing safety architectures.
The paper tackles the problem of undesirable behaviors and reduced performance in safety-critical control systems due to frequent switching between a learning-based policy and a safety controller, by proposing a policy repair method that uses runtime data to minimize deviations while ensuring safety, with experimental results showing effectiveness even with approximated system models.
We study the problem of policy repair for learning-based control policies in safety-critical settings. We consider an architecture where a high-performance learning-based control policy (e.g. one trained as a neural network) is paired with a model-based safety controller. The safety controller is endowed with the abilities to predict whether the trained policy will lead the system to an unsafe state, and take over control when necessary. While this architecture can provide added safety assurances, intermittent and frequent switching between the trained policy and the safety controller can result in undesirable behaviors and reduced performance. We propose to reduce or even eliminate control switching by `repairing' the trained policy based on runtime data produced by the safety controller in a way that deviates minimally from the original policy. The key idea behind our approach is the formulation of a trajectory optimization problem that allows the joint reasoning of policy update and safety constraints. Experimental results demonstrate that our approach is effective even when the system model in the safety controller is unknown and only approximated.