Risk mitigation for interval signal temporal logic monitoring and synthesis
L. Baird, A. Schoer, M. Cleaveland, S. Coogan, K. Leahy
IEEE Control Systems Letters, 2026
Abstract
This letter presents a method to mitigate the risk of violation of a temporal logic specification. Given a time-varying signal with interval-valued uncertainty at each time step, we propose an optimization approach to identify time instances for which a tighter uncertainty bound is required to satisfy the specification, knowledge which can be used to, e.g., focus sensing resources to strategically reduce uncertainty. We demonstrate our method on a simulated unmanned underwater vehicle where GPS calibration is informed by the identified time instances. In contrast to existing methods that solve an uncertainty-aware optimal control program with temporal logic mixed-integer constraints, our proposed interval-tightening approach is several orders of magnitude faster to compute. Additionally, we survey methods to produce interval-valued uncertainty, specifically how probabilistic bounds may translate to confidence intervals about an entire signal.