Distributed Error Detection and Diagnosis in Wireless Embedded Systems with Customized Hardware Support

Purdue University, West Lafayette IN

Investigators

Abstract

The objective of this research is to improve the reliability and robustness of networked wireless embedded systems, thereby enabling their widespread adoption in a range of mission-critical applications. The approach is to synergistically combine fine-grained node-level error detection mechanisms with a distributed network-level monitoring service to efficiently detect and diagnose various hardware and software errors at runtime. With respect to intellectual merit, the research has three major thrusts. The first thrust investigates the design of embedded hardware platforms augmented with a dedicated reliability co-processor. Such a co-processor based approach provides a logically and physically separate foundation to monitor the behavior of hardware and software at an intra-node level. The second thrust involves the design of a distributed, network-level monitoring service that uses observed event streams from individual nodes to perform error detection and root-cause diagnosis at runtime. By correlating observations collected from different nodes, the monitoring service can detect anomalies at the network-level that node-level monitoring mechanisms cannot. The final thrust deals with the creation of customized, resource-efficient communication primitives for the reliability co-processor to interact with the network-level monitoring service to enable accurate error detection and diagnosis in a cooperative manner. With respect to broader impacts, the research has the potential to positively impact the adoption of networked embedded systems in a variety of consumer and mission-critical applications. Research results are incorporated into graduate and undergraduate courses. The systems developed will also directly impact the broader community through programs at Purdue University including the Engineering Projects In Community Service and the Vertically Integrated Projects.