Monitoring a robot swarm using a data-driven fault detection approach

Belkacem Khaldi, Fouzi Harrou, Foudil Cherif, Ying Sun

Research output: Contribution to journalArticlepeer-review

31 Scopus citations

Abstract

Using swarm robotics system, with one or more faulty robots, to accomplish specific tasks may lead to degradation in performances complying with the target requirements. In such circumstances, robot swarms require continuous monitoring to detect abnormal events and to sustain normal operations. In this paper, an innovative exogenous fault detection method for monitoring robots swarm is presented. The method merges the flexibility of principal component analysis (PCA) models and the greater sensitivity of the exponentially-weighted moving average (EWMA) and cumulative sum (CUSUM) control charts to insidious changes. The method is tested and evaluated on a swarm of simulated foot-bot robots performing a circle formation task, via the viscoelastic control model. We illustrate through simulated data collected from the ARGoS simulator that a significant improvement in fault detection can be obtained by using the proposed method where compared to the conventional PCA-based methods (i.e., T2 and Q).
Original languageEnglish (US)
Pages (from-to)193-203
Number of pages11
JournalRobotics and Autonomous Systems
Volume97
DOIs
StatePublished - Jun 30 2017

Bibliographical note

KAUST Repository Item: Exported on 2020-10-01
Acknowledged KAUST grant number(s): OSR-2015-CRG4-2582
Acknowledgements: This publication is based upon work supported by the King Abdullah University of Science and Technology (KAUST) Office of Sponsored Research (OSR) under Award No: OSR-2015-CRG4-2582. The work is done in collaboration with the LESIA Laboratory, Department of Computer Science, University of Mohamed Khider, Biskra, Algeria. We would like to thank the reviewers of this article for their insightful comments, which helped us to greatly improve its quality.

Fingerprint

Dive into the research topics of 'Monitoring a robot swarm using a data-driven fault detection approach'. Together they form a unique fingerprint.

Cite this