Feature-based robot navigation using a Doppler-azimuth radar

Abstract

The merits of the Doppler radar compared to other existing sensors used for robot navigation, such as the LIDAR, include a lower cost, smaller size and lower weight which could prove to be useful in economically building a swarm of mobile vehicles. This paper demonstrates that, given a feature-based map and landmark associations, a Doppler radar which outputs Doppler-shift measurements with associated azimuth readings and an Extended Kalman filter for processing measurements, a robot is able to self-localise. Additionally, the Cramer–Rao lower bound (CRLB) for the estimation error of this scenario is computed to show that it is theoretically feasible for robot self-localisation, which is later verified through Monte–Carlo simulations.