High-Accuracy R-DML Angle Estimation for FMCW Radar

This paper proposes an angle estimation technique that has advantages over conventional algorithms in terms of angular resolution, accuracy, and computational speed. This technique, named R-DML, combines the RELAX and deterministic maximum likelihood (DML) algorithms. Its performance was verified through Monte Carlo simulations and measurements using a W-band radar system. In the measurement, the azimuth DOA estimation results for two objects were compared using fast Fourier transform. In the simulation, the root mean square error was compared with the asymptotic Cramer-Rao bound, and was confirmed to be almost at the same level for a signal with an signal-to-noise ratio (SNR) of 20 dB. However, there was an inaccuracy in the angle estimate for signals with a low SNR, and the limitations were also confirmed. In addition, the same angular accuracy as that obtained using DML was achieved with a processing time of only 16.8 % compared with DML.