Self-repairing Robot Swarms

TY - CONF

T1 - Self-repairing Robot Swarms

AU - Millard, Alan Gregory

AU - Timmis, Jonathan

PY - 2011/10

Y1 - 2011/10

N2 - It has long been assumed that swarm systems are robust, in the sense that the failure of individual robots will have little detrimental effect on a swarm's overall collective behaviour. However, a recent study by Bjerknes has shown that this is not always the case. The task of emergent beacon taxis was used as a case study, which requires an aggregated swarm of homogeneous autonomous robots, with limited sensing abilities, to traverse an empty arena towards an infrared beacon. The failure mode found to have the most damaging effect upon the swarm's emergent beacon taxis behaviour, was motor failure. This fault renders a robot stationary, but does not affect its other electronic systems, allowing the robot to continue to contribute to the emergent behaviour of aggregation. When the swarm is required to physically translate its position during beacon taxis, a robot experiencing motor failure will anchor the swarm, impeding (or at worst, preventing) the swarm's progress towards the beacon.In our previous work we investigated whether the anchoring issues observed by Bjerknes could be alleviated through the development of self-repair mechanisms that enable a swarm to take some form of corrective action in response to the failure of individual robots. Extending the work of Ismail et al., an equivalent failure mode was considered, whereby a fault in a robot's power unit causes a sudden loss of stored energy. The amount of remaining energy is sufficient to allow for simple signalling, but insufficient to power the robot's motors, resulting in a loss of mobility. A robot experiencing this type of failure may be 'repaired' by having its battery recharged by functional robots. The specialised hardware required to implement such self-repair mechanisms has only recently become available, in the form of the Symbricator robot - a new robotic platform designed for the SYMBRION/REPLICATOR projects - which facilitates the sharing of energy between physically connected robots.We have developed a novel decentralised self-repair algorithm for the Symbricator robotic platform, inspired by the emergent aggregation behaviour of cellular slime moulds, using software available from the Player/Stage Project. A simulated swarm of robots executing this algorithm is observed to self-organise to form a physically connected ad hoc robotic 'organism' around a faulty robot, which serves as a conduit for energy transfer. The swarm's collective energy is then evenly redistributed throughout the organism, and once the faulty robot has been repaired, the robots simply disperse to resume beacon taxis. Experimental results have shown that our algorithm enables a simulated swarm of ten robots to recover from up to five simultaneously injected faults, thus affording the swarm a high degree of fault-tolerance.

AB - It has long been assumed that swarm systems are robust, in the sense that the failure of individual robots will have little detrimental effect on a swarm's overall collective behaviour. However, a recent study by Bjerknes has shown that this is not always the case. The task of emergent beacon taxis was used as a case study, which requires an aggregated swarm of homogeneous autonomous robots, with limited sensing abilities, to traverse an empty arena towards an infrared beacon. The failure mode found to have the most damaging effect upon the swarm's emergent beacon taxis behaviour, was motor failure. This fault renders a robot stationary, but does not affect its other electronic systems, allowing the robot to continue to contribute to the emergent behaviour of aggregation. When the swarm is required to physically translate its position during beacon taxis, a robot experiencing motor failure will anchor the swarm, impeding (or at worst, preventing) the swarm's progress towards the beacon.In our previous work we investigated whether the anchoring issues observed by Bjerknes could be alleviated through the development of self-repair mechanisms that enable a swarm to take some form of corrective action in response to the failure of individual robots. Extending the work of Ismail et al., an equivalent failure mode was considered, whereby a fault in a robot's power unit causes a sudden loss of stored energy. The amount of remaining energy is sufficient to allow for simple signalling, but insufficient to power the robot's motors, resulting in a loss of mobility. A robot experiencing this type of failure may be 'repaired' by having its battery recharged by functional robots. The specialised hardware required to implement such self-repair mechanisms has only recently become available, in the form of the Symbricator robot - a new robotic platform designed for the SYMBRION/REPLICATOR projects - which facilitates the sharing of energy between physically connected robots.We have developed a novel decentralised self-repair algorithm for the Symbricator robotic platform, inspired by the emergent aggregation behaviour of cellular slime moulds, using software available from the Player/Stage Project. A simulated swarm of robots executing this algorithm is observed to self-organise to form a physically connected ad hoc robotic 'organism' around a faulty robot, which serves as a conduit for energy transfer. The swarm's collective energy is then evenly redistributed throughout the organism, and once the faulty robot has been repaired, the robots simply disperse to resume beacon taxis. Experimental results have shown that our algorithm enables a simulated swarm of ten robots to recover from up to five simultaneously injected faults, thus affording the swarm a high degree of fault-tolerance.

KW - Swarm robotics

KW - Self-repair

KW - Trophallaxis

M3 - Abstract

SP - 96

EP - 97

T2 - York Doctoral Symposium on Computer Science

Y2 - 20 October 2011 through 20 October 2011

ER -