Implementation of an Adaptive Filter using an Evolvable Hardware Strategy

This paper explores the feasibility of an adaptive scheme to tolerate degradations in a high-order low-pass filter, taken as a case study. The adaptation is achieved through an evolvable hardware strategy that employs reconfigurable hardware and a genetic algorithm (GA). For establishing the filter reconfiguration, it is considered that it is tested periodically during in-field operation using Transient Analysis Method (TRAM). If TRAM founds that the filter does not fulfill the specifications, then the AG, which runs on an external computer, searches a new set of configurable parameters that allows the filter reaching the specifications. Then, the hardware is configured with these new parameters for continuing its normal operation. For the proposed scheme, simulation results show the TRAM feasibility when certain restrictions are taken for its implementation. The AG evaluation is performed through fault simulation, using four fault models that consider deviations in circuital parameters. The results show that AG founds, in all fault conditions simulated, solutions with errors smaller than the tolerable ones, demonstrating the feasibility of the proposed scheme.