Confidence Composition for Monitors of Verification Assumptions

Ruchkin, Ivan; Cleaveland, Matthew; Ivanov, Radoslav; Lu, Pengyuan; Carpenter, Taylor J.; Sokolsky, Oleg; Lee, Insup

doi:10.1109/iccps54341.2022.00007

Cited by 6 publications

(1 citation statement)

References 28 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Intuitively, what this definition means is that if the state estimator says that there is probability p that the system is in state s, then the system will be in state s with probability p. Calibration is an increasingly common requirement for learning-based detectors [16,27,14,31], and we validate it in our experiments. Now we are ready for our main theoretical result: assuming that M AS is conservative with respect to M OS and that the state estimator is well-calibrated, we show that the safety estimates of our monitoring are conservatively calibrated.…”

Section: Conservatism Guaranteesmentioning

confidence: 94%

Conservative Safety Monitors of Stochastic Dynamical Systems

Cleaveland¹,

Ruchkin²,

Sokolsky³

et al. 2023

Preprint

View full text Add to dashboard Cite

Generating accurate runtime safety estimates for autonomous systems is vital to ensuring their continued proliferation. However, accurately reasoning about future system behaviors is generally too complex to do at runtime. To better reason about system safety at runtime, we propose a method for leveraging design time model checking results at runtime. Specifically, we model the system as a probabilistic automaton (PA) and compute bounded time reachability probabilities over the states of the PA at design time. At runtime, we combine distributions of state estimates with the safety probabilities from design time to produce a bounded time safety estimate. We argue that our approach produces well calibrated safety probabilities, assuming the estimated state distributions are well calibrated. We evaluate our approach using a case study of simulated water tanks.

show abstract

Section: Conservatism Guaranteesmentioning

confidence: 94%