Multi-objective DOA estimation in automotive radar using a Sailfish optimizer with Latin hypercube sampling

Direction of arrival (DOA) estimation is a challenging problem in automotive radar for target detection and finds potential applications in the design of driver assistance system (DAS). Accurate DOA estimation in automotive radar becomes more difficult in the presence of noise. To address this issue, in this paper, the DOA estimation problem in automotive radar is formulated as a multi-objective optimization problem by simultaneous minimization of two operators: (1) reciprocal of Maximum Likelihood function; and (2) noise variance. Accurate angle estimation can be carried out in the absence of noise, and vice-versa, thus making these two functions contradictory in nature. In this article, a multi-objective Sailfish optimizer algorithm is proposed with Latin hypercube sampling and double archive mechanism termed as MOSFO-LH. The Latin hypercube generates evenly distributed samples and thus provides potential diversification of the search agents in the objective space. Further, a double archive method is incorporated to keep the good solutions that are responsible for maintaining diversity in the Pareto Front. Superior performance of the proposed MOSFO-LH algorithm is reported over five more comparative algorithms, on ten benchmark multi-objective test functions selected from the ZDT and DTLZ test suite. Case studies are reported for DOA estimation in automotive radar under six noise conditions. Effective performance is demonstrated with estimated DOA angles, box plots of root mean square error values, Pareto optimal fronts, heat map of inverse generalized distance, maximum spread and hyper volume metric.