Reduced-Complexity Distributed Least-Squares Estimation Over Adaptive Networks

In wireless ad-hoc networks, nodes usually possess limited processing and electrical power resources. Therefore, when performing a decentralized task over such a network, it is desirable to minimize the required in-node computations and inter-node communications. In this paper, we propose a reduced-complexity diffusion recursive least-squares (RC-diffRLS) algorithm for distributed estimation over adaptive networks. To this end, we utilize the dichotomous coordinate-descent (DCD) algorithm to solve the normal equations of the local least-squares estimation problems at the nodes. Simulation results testify that the proposed algorithm can perform very close to a previously proposed diffusion recursive least-squares (diffRLS) algorithm while being considerably simpler in computational complexity and appreciably more resilient in numerically stability. Using the proposed algorithm, one can also establish a trade-off between complexity and performance.