Output details

The Distributed Co-Evolution of an Embodied Simulator and Controller for Swarm Robot Behaviours (2011)

This co-authored (with Winfield and Studley), published international conference paper presents the findings and analysis of a series of experiments to test an innovative method developed for assessing the effectiveness of robotic controllers in swarm robotics. The distributed co-evolutionary method devised and tested here adapts the environmental model of an on board simulator within the context of swarm robotics. The experiments used the foraging problem, where robots randomly searched for food items and deposited them at a designated nest site to assess the perceptual problem of adaptive environmental modelling by a metric of physical task performance across the swarm of robots. A Vicon 3D Motion Capture suite was used to capture the telemetry of mobile robots with high precision. All experiments were conducted on real mobile robots that operated the co-evolutionary algorithm on board.

The paper presents the first working demonstration of the co-evolution of both a robots embodied simulator and robots behavioural controller. It also provides a further analysis of the concept of ‘the reality gap’ between assessment in simulated and real environments. By identifying and investigating the unaddressed area in evolutionary robotics of an adaptive environment model, it makes embodied simulators as a tool in robotics more feasible in general. The research also supported the claim of swarm robotics that intelligent outcomes can emerge from many simple interactions. The paper has wider implications in understanding the transfer of virtual models to physically realised outcomes, and in enabling machines to autonomously detect and correct for discrepancies in operational behaviour.

The paper was delivered at the IEEE/RSJ Intelligent Robots and Systems Conference (IROS) (25-30 September 2011), San Francisco (http://www.iros2011.org) and published as part of the conference proceedings, IEEE/RSJ International Conference on Intelligent Robots and Systems, IROS (2011).