Multi Target Localization With Block Orthogonal Least Squares for Multistatic MIMO Radars

Recently, there has been a growing interest in multistatic radar configurations to improve the localization of multiple targets. Theoretically, the maximum likelihood (ML) approach enables to fuse the information provided by each radar pair to localize the different targets. However, it involves a multi-dimensional search process whose complexity exponentially grows with the number of targets. Consequently, heuristic methods, notably including the block orthogonal matching pursuit (BOMP), have been used in the multistatic radar context to approach the ML estimation greedily. Interestingly, the more accurate block orthogonal least squares (BOLS) method has not been studied in this context because the performance improvement is usually low in regard to its computational complexity. In this work, we investigate the application of BOLS to an angle-based localization of multiple targets using a multistatic multiple-input and multiple-output (MIMO) radar. First, an efficient implementation of BOLS is presented reducing its computational complexity. Then, using Monte Carlo simulations, we show evidence of the significant advantage of this efficient implementation of BOLS over BOMP in this scenario featuring highly correlated signals. The impact of radar parameters on the localization root mean square error and on the computational complexity of both algorithms is studied.