To err is robotic, to tolerate immunological: fault detection in multirobot systems

Abstract

Abstract. Fault detection and fault tolerance represent two of the most important and largely unsolved issues in the field of multirobot systems. Efficient, long-term operation requires an accurate, timely detection, and accommodation of abnormally behaving robots. Most existing approaches to fault-tolerance prescribe a characterisation of normal robot behaviours, and train a model to recognise these behaviours. Behaviours unrecognised by the model are consequently labelled abnormal or faulty. Multirobot systems employing these models do not transition well to scenarios involving temporal variations in behaviour (e.g., online learning of new behaviours, or in response to environment perturbations). The vertebrate immune system is a complex distributed system capable of learning to tolerate the organism's tissues even when they change during puberty or metamorphosis, and to mount specific responses to invading pathogens, all without the need of a genetically hardwired characterisation of normality. We present a generic abnormality detection approach based on a model of the adaptive immune system, and evaluate the approach in a swarm of robots. Our results reveal the robust detection of abnormal robots simulating common electro-mechanical and software faults, irrespective of temporal changes in swarm behaviour. Abnormality detection Fault detection in multirobot system2 is shown to be scalable in terms of the number of robots in the swarm, and in terms of the size of the behaviour classification space.

Evolution of Collective Behaviors for a Real Swarm of Aquatic Surface Robots

“…Although our experiments confirmed that swarm behaviors inherently have some degree of tolerance for individual faults, in real-world applications it may be necessary to guarantee that faulty individuals will not compromise the whole swarm. Approaches for fault detection [80], fault tolerance [81], and online adaptation [79] must therefore be implemented in the system.…”

Towards Fault Diagnosis in Robot Swarms: An Online Behaviour Characterisation Approach

“…There has been previous mention of work by several researchers toward fault detection in swarm robotics systems [12], [7], [10], [16]. In each piece of work, the fundamental mechanism that enables a fault to be detected can be essentially reduced to the comparison of an individual's observed behaviour to that which the swarm or observer expects it to exhibit.…”

“…Following this process leads to the following BFV that is proposed to be sufficiently representative of the aforementioned behaviours. Following Tarapore et al [16], the BFV described here is a concatenation of five individual binary features, where a returned value of 1 or 0 indicates the presence or absence of the feature, respectively:…”

“…This paper therefore proposes a fault diagnosis approach based on behavioural feature vectors (BFV), which encode a robot's behaviour as a series of binary features. Such feature vectors have been previously used successfully by Tarapore et al [16] for the purpose of fault detection in swarm robotics systems and, as is shown in this paper, can similarly be used to perform fault diagnosis via the characterisation of robot behaviour.…”

“…There are a number of examples where fault detection in swarm systems, by various methods, has been successful in simulation (see the work by Millard et al [11], Khadidos et al [10] and Tarapore et al [16]), and in hardware (see the work by Christensen et al [7]). In contrast, there has been limited research on fault diagnosis in robots, examples include the immune-inspired work by Bi [1] and Bi et al [2].…”