Decision-Theoretic Monitoring of Cyber-Physical Systems
| Autor: | A. Prasad Sistla, Miloš Žefran, Andrey Yavolovsky |
|---|---|
| Rok vydání: | 2016 |
| Předmět: |
0209 industrial biotechnology
Theoretical computer science Computer science Cyber-physical system Partially observable Markov decision process Observable 02 engineering and technology Mobile robot system Monitoring problem Construct (python library) 020901 industrial engineering & automation Life-critical system 0202 electrical engineering electronic engineering information engineering 020201 artificial intelligence & image processing Formal verification |
| Zdroj: | Runtime Verification ISBN: 9783319469812 RV |
| Popis: | Runtime monitoring has been proposed as an alternative to formal verification for safety critical systems. This paper introduces a decision-theoretic view of runtime monitoring. We formulate the monitoring problem as a Partially Observable Markov Decision Process (POMDP). Furthermore, we adopt a Partially Observable Monte-Carlo Planning (POMCP) to compute an approximate optimal policy of the monitoring POMDP. We show how to construct the POMCP for the monitoring problem and demonstrate experimentally that it can be effectively applied even when some of the state-space variables are continuous, the case where many other POMDP solvers fail. Experimental results on a mobile robot system show the effectiveness of the proposed POMDP-monitor. |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|