Monte Carlo localisation of a mobile robot using a Doppler–Azimuth radar

This paper investigates the moving robot localisation problem using a Doppler–Azimuth radar array. The solution is formulated in the framework of nonlinear/non-Gaussian estimation using a particle filter and a random finite set (RFS) model of measurements. The proposed approach assumes the availability of a feature-based map, radar measurements and robot odometry data. The associations between the measurements and the features of the map (landmarks) are unknown. The RFS model is adopted to deal with false and missed detections and uses Murty's algorithm to reduce computation when solving the association problem. The proposed particle filter incorporates the Kullback–Leibler Distance (KLD)-Sampling to reduce computational time. Monte-Carlo simulation results demonstrate the efficacy of the proposed algorithm.