Agent-based modelling of multi-robot systems

Multi-robot systems can be used in various industrial and non-industrial applications such as manufacturing, environmental monitoring, disaster rescue missions or agricultural foraging. In this context, different problems need to be solved: robot control, robot perception, multi-robot system coordination, global and local planning/re-planning etc. For example, choosing a good coordination mechanism is a difficult task that depends on the application and can be done via trial and error experiments and comparative studies. Some guidelines can be learned from the case studies reported in the literature. Moreover, artificial intelligence provides several techniques (e.g. rule-based systems, machine learning, artificial neural networks, swarm intelligence) that can improve the performances of multi-robot systems coordination. Regarding experimental comparative studies between different methods, a cheaper solution is given by simulations. On the other hand, knowledge sharing between robots can be provided by ontologies that facilitate the inter-robot communication as a base of multi-robot system coordination. Our research work focus on modelling a multi-robot system as an agent based system which is a proper choice for distributed systems and enables simulations. In this paper we present an agent-based model for multi-robot systems that use a common ontology and reinforcement learning as agent adaptation ability. An application for environmental quality monitoring is discussed.