Exploiting Computational Redundancy for Efficient Recovery from Soft Errors in Sensor Nodes

Most existing techniques for the design and implementation of fault tolerance use resource redundancy. As such, due to scarcity of resources, it is difficult to directly apply them for adding fault tolerance to sensor nodes in Wireless Sensor Networks (WSNs). Thus, it is desirable to develop techniques that implement fault tolerance under the constraints of memory and processing power of sensor nodes. We present a novel method for designing recovery from transient faults that cause nondeterministic bit-flips in the task queue of the scheduler of TinyOS, which is the operating system of choice for sensor nodes. Specifically, our approach exploits computational redundancy for the design of recovery instead of using resource redundancy. The presented fault-tolerant task queue recovers from bit-flips with significantly lower space/time overhead compared with the Error Correction Codes.