Model Checking Agent Communication

Abstract

Abstract. Model checking is a formal and automatic technique used to verify computational systems (e.g. communication protocols) against given properties. The purpose of this chapter is to describe a model checking algorithm to verify communication protocols used by autonomous agents interacting using dialogue games, which are governed by a set of logical rules. We use a variant of Extended Computation Tree Logic CTL* for specifying these dialogue games and the properties to be checked. This logic, called ACTL*, extends CTL* by allowing formulae to constrain actions as well as states. The verification method uses an on-the-fly efficient algorithm. It is based on translating formulae into a variant of alternating tree automata called Alternating Büchi Tableau Automata (ABTA). We present a tableau-based version of this algorithm and provide the soundness, completeness, termination and complexity results. Two case studies are discussed along with their respective implementations to illustrate the proposed approach. The first one is about an agent-based negotiation protocol and the second one considers a modified version of the NetBill protocol.

Communicative commitments: Model checking and complexity analysis

“…In fact, translationbased approaches have the problem of preventing verifying the real and concert semantics of commitments and related concepts as defined in the underlying logics. These approaches only provide partial solution to the problem of model checking commitments [5,16] as they reduce commitment modalities into domain variables, which-as in predicate logics-stop distinguishing among various ''modes'' of truth such as necessarily true and true in the future. Simply put, translation-based approaches use variables that do not account for the meaning of commitments, so commitments are dealt with as tokens.…”

“…Bentahar et al [4,5] and El-Menshawy et al [16] showed how social commitments can be mapped into underlying logic-based formalisms by extending CTL ⁄ with modalities to reason about social commitments and action formulae as well as state formulae in order to formally specify multi-agent interaction protocols and some desirable protocol properties. However, the intuition behind the accessibility relation used to define the semantics of commitments and how it can be implemented were not discussed in depth.…”

“…The first class extends temporal logics with predicates to represent and reason about commitments and associated actions, such as [58,59,32,34,39]. The second class enriches temporal logics with modalities for commitments and their actions, such as [46,6,7,5,47,16,17].…”

“…These approaches particularly aim to overcome the shortcomings of ACLs semantics defined using mental approaches where the semantics is expressed in terms of agents' internal states such as beliefs, desires and intensions [57]. Commitments are employed in some of these social approaches, which successfully provide a powerful representation for modeling multi-agent interactions [46,13,14,5]. In broad terms, commitments are social, public and help represent the state of affairs in the course of multi-agent interactions.…”

“…Bentahar et al [5] presented a new logic called ACTL ⁄ (an extension of CTL ⁄ with action formulae) to specify protocols and some desirable properties. Their verification method is based on an informal translation of ACTL ⁄ formulae and protocol into Alternating Büchi Tableau Automata (ABTA) in order to directly use the CWB-NC 2 automata-based model checker where the commitment states are simply defined as variables and actions as atomic propositions using CCS (the input language of CWB-NC).…”