Service-oriented approach to fault tolerance in CPSs

Cyber-physical systems (CPSs) are open and interconnected embedded systems that control or interact with physical processes. Failures in CASs can lead to loss of production time, damage to the equipment and environment, or loss of life, meaning that dependability and resilience are key properties for their design. However, existing fault tolerance and safety approaches are inadequate for complex, networked and dynamic CPSs. Service-orientation, on the other hand, is generally considered to be a robust architectural style, but there is a limited amount of research on fault tolerance of service-oriented architecture (SOA), especially on distributed real-time systems. We propose an approach that utilizes the loosely coupled nature of services to implement fault tolerance using a middleware-based real-time SOA (RTSOA) for CPSs. The approach, based on the concepts of fault isolation and recovery at the service level, is empirically evaluated using a demanding bilateral teleoperation (remote handling) application. The empirical evaluation demonstrates that RTSOA supports real-time fault detection and recovery, use of services as a unit of fault isolation, and it provides capability to implement fault tolerance patterns flexibly and without significant overhead. (C) 2015 Elsevier Inc. All rights reserved.