An integrative modelling approach for simulation and analysis of adaptive agents

Abstract

To simulate adaptive agents with abilities matching those of their real-world counterparts, a natural approach is to incorporate adaptation mechanisms such as classical conditioning into agent-based simulation. Existing models for adaptation mechanisms are usually based on quantitative methods such as DST In contrast, agent-based simulation is usually based on qualitative, logical languages. To bridge this gap, this paper puts forward an integrative approach to simulate and analyse the conditioning process of an adaptive agent, which integrates quantitative and qualitative aspects within one temporal specification language. The approach comprises (1) simulation of adaptation mechanisms in an executable language, (2) automated analysis of dynamic properties against simulation traces, and (3) verification of representation relations for internal agent states against simulation traces. Furthermore, the approach addresses the issue of realism of intermediate states in a simulated conditioning process.