Sequential DoA estimation using Recursive DML

Deterministic Maximum Likelihood (DML) Direction-of-Arrival (DoA) estimation for FMCW radar is a robust and high performance DoA estimation technique. The drawback of the method is that it is a computational intensive method. For a K-target DoA estimation application, DML is basically the search for the maximum value of the Maximum-Likelihood (ML) cost function in a K-dimensional (DoA angle) space. Its high complexity is due to a large search space and due to the complexity of the evaluation of the ML cost function. Evaluating the ML cost function for a point in the K-dimensional search space requires the inversion of a K × K matrix. Moreover, in a sequential search for K=1, 2, 3, etc. targets, hardly any computations of the K-target search can be re-used in the (K + 1)-target search. In this paper the DML problem is formulated in a recursive way such that K × K matrix inversions are not needed anymore and that computations carried out for the K-target search can be re-used for a (K + 1)-target search. This enables the DML search to: 1) be carried out as a tree search, 2) have integrated number -of target estimation and 3) having a reduced search space.