On Fault Tolerance and Scalability of Swarm Robotic Systems

Abstract

This paper challenges the common assumption that swarm robotic systems are robust and scalable by default. We present an analysis based on both reliability modelling and experimental trials of a case study swarm performing team work, in which failures are deliberately induced. Our case study has been carefully chosen to represent a swarm task in which the overall desired system behaviour is an emergent property of the interactions between robots, in order that we can assess the fault tolerance of a self-organising system. Our findings show that in the presence of worst-case partially failed robots the overall system reliability quickly falls with increasing swarm size. We conclude that future large scale swarm systems will need a new approach to achieving high levels of fault tolerance.

Run-time detection of faults in autonomous mobile robots based on the comparison of simulated and real robot behaviour

“…Bjerknes and Winfield [1] demonstrated that motor failure had the most detrimental affect on a swarm's ability to carry out its task. For this reason, we use motor failure as case study in this paper.…”

“…However, Bjerknes and Winfield [1] have recently shown that this is not always the case, particularly in the event of partial failures (such as motor failure). The reliability modelling in [1] shows that overall system reliability rapidly decreases with increasing swarm size, therefore this is a problem that cannot simply be solved by adding more robots to the swarm. Instead, future large-scale swarm systems will need an active approach to dealing with failed individuals if they are to achieve a high level of fault tolerance.…”

“…The work of Christensen et al represents the state-of-the-art in exogenous fault detection for robot swarms, but it only addresses the case of completely failed robots, the effect of which on collective behaviour has been shown to be relatively benign by Winfield and Nembrini [3]. The occurrence of partial failures is of far greater concern, as highlighted by Bjerknes and Winfield [1].…”

The distributed co-evolution of an on-board simulator and controller for swarm robot behaviours

“…From a designer's point of view, only a single robotic agent is defined and the result of complex interactions must be extrapolated outwards. Through decentralisation, selforganising robotic systems they are cited as being robust, flexible, and scalable; although this is not without caveats [1].…”

Quantification and Analysis of the Resilience of Two Swarm Intelligent Algorithms

“…In [8], the authors present an analysis of the BEECLUST algorithm and experimentally verify the robustness of the algorithm by adding agents with impaired temperature sensors. In [4], the authors perform a comprehensive analysis of the resilience of the swarmtaxis algorithm with respect to failure modes such as IR sensor failure and motor failure. For the purpose of this paper, we consider the ability of a system to perform its overall goals even with suboptimal agent behavior to be its "resilience".…”