Use this URL to cite or link to this record in EThOS: http://ethos.bl.uk/OrderDetails.do?uin=uk.bl.ethos.686287
Title: Model checking and compositional reasoning for multi-agent systems
Author: Jones, Andrew Vaughan
Awarding Body: Imperial College London
Current Institution: Imperial College London
Date of Award: 2014
Availability of Full Text:
Access from EThOS:
Full text unavailable from EThOS. Please try the link below.
Access from Institution:
Abstract:
Multi-agent systems are distributed systems containing interacting autonomous agents designed to achieve shared and private goals. For safety-critical systems where we wish to replace a human role with an autonomous entity, we need to make assurances about the correctness of the autonomous delegate. Specialised techniques have been proposed recently for the verification of agents against mentalistic logics. Problematically, these approaches treat the system in a monolithic way. When verifying a property against a single agent, the approaches examine all behaviours of every component in the system. This is both inefficient and can lead to intractability: the so-called state-space explosion problem. In this thesis, we consider techniques to support the verification of agents in isolation. We avoid the state-space explosion problem by verifying an individual agent in the context of a specification of the rest of the system, rather than the system itself. We show that it is possible to verify an agent against its desired properties without needing to consider the behaviours of the remaining components. We first introduce a novel approach for verifying a system as a whole against specifications expressed in a logic of time and knowledge. The technique, based on automata over trees, supports an efficient procedure to verify systems in an automata-theoretic way using language containment. We show how the automata-theoretic approach can be used as an underpinning for assume-guarantee reasoning for multi-agent systems. We use a temporal logic of actions to specify the expected behaviour of the other components in the system. When performing modular verification, this specification is used to exclude behaviours that are inconsistent with the concrete system. We implement both approaches within the open-source model checker MCMAS and show that, for the relevant properties, the assume-guarantee approach can significantly increase the tractability of individual agent verification.
Supervisor: Lomuscio, Alessio Sponsor: Not available
Qualification Name: Thesis (Ph.D.) Qualification Level: Doctoral
EThOS ID: uk.bl.ethos.686287  DOI: Not available
Share: