Context-Based Collaborative Self-Test for Autonomous Wireless Sensor Networks

Reliability is a major concern in autonomous wireless sensor networks. Current approaches to maintaining high overall system availability concentrate on pseudo-random test scheduling and test vector generation based on a probabilistic approach to failure prediction. In the case of wireless sensor networks though, most of device failures can be directly associated with specific events. Furthermore, these events can often be identified using the sensors already present on the nodes and used to trigger self test of the affected devices with test vectors specifically crafted to match the possible failures. In this paper, we discuss an approach to wireless sensor node self-testing using sensor data gathered by the device itself and by the neighboring nodes. We analyze possible impact of this approach on the Mean Time To Detect (MTTD) and the overall system availability. Also, the proposed approach can help decrease energy consumption of the system through avoiding unnecessary data communication and extensive hardware testing. We also discuss advantages arising from installation of additional dedicated sensors on the nodes that help to more accurately detect and classify an event and thus the possible failure and its severity. Finally, we present a test system that implements the proposed approach.