Requirements Planning with Event Calculus for Self-adaptive Multi-agent System

Abstract

Self-adaptation of Multi-agent cooperative systems requires dynamic decision making and planning at runtime. Modeling the contextual and executable requirements of such systems as planning actions and states, this paper proposes a requirements-driven planning approach to self-adaptation. The planning model includes the states of the system context and the actions describing the behaviors of its multiple agents; the interactions between these agents and their environment are computed through an expansion of the requirements-driven planning graph, which is then used to verify whether the agents can collaborate in order to reach the desired goal states from their current states. In addition, the requirements are represented for Event Calculus to facilitate monitoring and reasoning about the actions of agents, achieving requirements driven planning at runtime.