Low cost error monitoring for improved maintainability of IoT applications

Electronic systems with power-constrained embedded devices are used for a variety of IoT applications, such as geomonitoring, parking sensors and surveillance. Such applications may tolerate few errors. However, with the increasing occurrence of faults in-the-field, devices that exhibit systematic erroneous behaviour must be eventually identified and replaced. In this paper, we propose a novel low cost error monitoring technique to assist the maintainability planning of low power IoT applications by ranking devices based on the systematic erroneous behaviour they exhibit. Small on-chip monitors are used to collect the signal probability information at the outputs of each device which is then transmitted to the system software via the communications channel of the system to rank them accordingly. To evaluate the error monitoring capabilities of the proposed technique, we injected multiple bit-flips and stuck-at faults on a set of the EPFL and the ISCAS benchmarks. Results demonstrate an average error coverage of 84.4% and 73.1% of errors induced by bit-flips and stuck-at faults, respectively, with an average area cost of 1.52%. A maintainability planning simulation shows that the proposed technique achieves a reduction of 26x to 263x in area cost and static power, and consumes over 625x less power for communications when compared against duplication and comparison