Distributed Diffusion FxLMS Algorithm for Multi-channel AVC System

This paper considers the multi-channel active vibration control problem. A distributed diffusion filter-x least mean square algorithm is designed to reduce the global error signal. First, a regularization term is additionally introduced in the cost function to smooth different intermediate estimate control filters, which actually results in a multi-task problem, more complicated than a single-task one. Subsequently, a method of selecting channel neighbors is proposed in order to achieve a uniform vibration amplitude through the minimization of the maximum error signal. Finally, the condition for the convergence of the proposed algorithm is deduced and the experiment results show that if the step size satisfies the convergence condition, the algorithm can achieve vibration reduction performance close to that of the centralized method but with lower computational complexity. It is also worth mentioning that our method can be applied to large-scale systems where the optimal control filters of each channel are quite different.