Diagnosis of Unwanted Behaviours in Multi-Agent Systems

Abstract

Multi-Agent Systems (MAS) must increase their reliability in order to guarantee a solid and safe response to unexpected situations in complex domains. To address such an issue, both fault tolerance and software testing techniques are applied to ensure complying services; the foundation of these is the effectiveness in identifying and classifying fault events. Despite the advances in the field, most of fault-tolerant MAS rely on diagnosis mechanisms that demand for a priori knowledge of the system. Seeking to overcome this need for knowledge, we propose an extension of the Spectrum-based Fault Localization technique to be applied to MAS as it uses minimal information about the monitored system to detect agents presenting fault behaviours. For the moment, however, the proposed extension focuses on closed MAS and static environments. We successfully instantiate our approach and demonstrate its functionalities.

Extracted Key Phrases

2 Figures and Tables

Cite this paper

@inproceedings{Passos2014DiagnosisOU, title={Diagnosis of Unwanted Behaviours in Multi-Agent Systems}, author={L{\'u}cio S. Passos and Rosaldo J. F. Rossetti and Joaquim Gabriel}, year={2014} }