Detecting Temporal Conflicts in Integrated Agent Specifications

Abstract

We present a model for the analysis of temporal conflicts in sets of autonomous agents restricted in the sense that they can be described in a first-order language. The dynamic aspects of agents are modelled by means of the event concept. Various kinds of conflicts can occur in a system of agents. Agents may be in conflict with themselves or with each other, all depending on time. Conflicts between agents may be due to direct incompatibility inherent in their specifications, but it mayaiso arise from the dynamic interactions protocols that govem multi-agent behaviour. The general theory for detecting conflict suggested here includes semantics for individual conflicts with respect to a first order action logic (FAL). Also introduced is the enrichment of agent systems with correspondence assertions. These express relationships between static or dynamic properties described in the formal specifications of agents. Correspondence assertions correspond to protocols for the system and can be analysed in relation to time. Representing agents as first order schemata, we construct a framework for determining when two schematacan be meaningfully integrated. We formalise this concept using the notion of non-conflict.