Markov决策过程中多目标查询的证书与见证
Certificates and Witnesses for Multi-Objective Queries in Markov Decision Processes
AI总结:
本文研究Markov决策过程中多目标可达性、不变性与平均收益查询的可验证性,将线性规划方法改造为认证算法,并生成调度器和子系统见证。
摘要:
认证验证算法不仅返回给定属性是否成立,还提供附带的、可独立检查的证书以及相应的见证。证书可用于轻松验证结果的正确性,而见证则提供有用的诊断信息,例如用于调试。因此,证书和见证显著提高了验证过程的可信度和可理解性。在本工作中,我们研究Markov决策过程中多目标可达性-不变性和平均收益查询的证书与见证,即由可达性与不变性谓词,或平均收益谓词构成的合取或析取,并同时涵盖全称量化和存在量化。由此,我们推广了以往关于单一可达性和不变性约束的证书与见证工作。为此,我们将已知的线性规划技术转化为认证算法,并证明可以获得以调度器和子系统形式呈现的见证。作为概念验证,我们报告了认证验证算法的实现及实验结果。
英文摘要:
Certifying verification algorithms not only return whether a given property holds or not, but also provide an accompanying independently checkable certificate and a corresponding witness. The certificate can be used to easily validate the correctness of the result and the witness provides useful diagnostic information, e.g. for debugging purposes. Thus, certificates and witnesses substantially increase the trustworthiness and understandability of the verification process. In this work, we consider certificates and witnesses for multi-objective reachability-invariant and mean-payoff queries in Markov decision processes, that is conjunctions or disjunctions either of reachability and invariant or mean-payoff predicates, both universally and existentially quantified. Thereby, we generalize previous works on certificates and witnesses for single reachability and invariant constraints. To this end, we turn known linear programming techniques into certifying algorithms and show that witnesses in the form of schedulers and subsystems can be obtained. As a proof-of-concept, we report on implementations of certifying verification algorithms and experimental results.
机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。