Robust Monte Carlo Localisation Using a Ground Penetrating Radar

Systems for localising mobile robots have certain disadvantages depending on the measuring principle. For example, systems with cameras or lidars reach their limits in harsh environments with dust, dirt andweather. Infrastructure-based localisation such asRFID is non-flexible and expensive to install. GNSS-based systems have problems with multipathing and indoor applications. To overcome these problems, a robust localisation system consisting of graduated frequency ground penetrating radar (GPR) andMonte Carlo localisation (MCL) is developed to remedy the situation. For this purpose, the radar scans long-term stable features in the underground which are used for the localisation. In this paper we will mention the test setup: Test robot, used stepped frequency radar and the test ground. In detail, the signal processing, the localisation approach with mapping and the calculation of the yaw angle are explained. In the exemplary field test, the orientation error is +/- 5 degrees and laterally +/- 5 cm.