Robust Optimization Strategies for Adaptive Filters Operating with Fixed and Transient Hardware Errors

Adaptive digital filters can be designed with hardware redundancy to achieve robust performance through a mechanism called adaptive fault tolerance (AFT). Results presented in this paper shown that due to the decorrelating properties of a transform AF structure transient errors dissipate much more quickly in transform domain AFT structures than in time domain structures. It is then shown that population based stochastic optimization algorithms (e.g.. the constrained random search algorithm) are capable of overcoming degraded performance caused by ill-conditioning in many different types of AFT filter structures