Charlie Street, Bruno Lacerda, Michal Staniaszek, Manuel Mühlig, Nick Hawes, "A Context-Aware Simulator for Rapidly Evaluating Multi-Robot Systems ", The 3rd International Symposium on Multi-Robot and Multi-Agent Systems, 2021.

Evaluation methods analyse multi-robot behaviour under a given property. Physics-based simulators and model checking are common evaluation tools. Physics-based simulators are more representative of real robot behaviour, but are often slowed by computational demand. Model checking techniques are faster, but use simplified environment models. In this paper, we combine these method's strengths to build an abstract simulator for rapidly evaluating continuous-time multi-robot behaviour. Our simulator samples through a multi-robot Markov model that captures robot dynamics using continuous phase-type distributions over action durations. We fit these distributions using data collected from a physics-based simulator. We also model the effects of robot interactions on action durations. We validate our simulator against a Gazebo simulation.