Formal verification of probabilistic swarm behaviours

Abstract

Robot swarms provide a way for a number of simple robots to work together to carry out a task. While swarms have been found to be adaptable, fault-tolerant and widely applicable, designing individual robot algorithms so as to ensure effective and correct swarm behaviour is very difficult. In order to assess swarm effectiveness, either experiments with real robots or computational simulations of the swarm are usually carried out. However, neither of these involve a deep analysis of all possible behaviours. In this paper we will utilise automated formal verification techniques, involving an exhaustive mathematical analysis, in order to assess whether our swarms will indeed behave as required. © 2010 Springer-Verlag Berlin Heidelberg.