Low complexity metaheuristics for joint ML estimation problems

Joint maximum likelihood (ML) estimation of multiple parameters is an important problem with wide-spread relevance in many domains. The high computational complexity involved in joint ML problems has led to the search for more efficient methods. Efficient heuristic algorithms for joint ML problems can be developed by exploiting the characteristics of the objective functions used in the estimation problem. This paper proposes a novel reformulation of existing heuristic algorithms, which considerably reduces their computational complexity with significant improvement in performance. The method is applicable for joint maximum likelihood estimation problems, with cost functions that exhibit asymptotic separability with increase in observation vector size. The proposed method is adopted to five recently discovered heuristic algorithms and consequently applied to a relevant recent signal processing problem in wireless communication. It is found that the reformulated algorithms deliver both reduced computational complexity as well as better mean square error (MSE) performance. The significant features of the proposed method are substantiated through extensive computer simulation studies.