Increasing Positioning Accuracy in Urban Environments Using Radar-Based Point Clouds

Accurate Positioning is critical for the safe operation of autonomous vehicles, especially when all newly manufactured vehicles have some autonomous features. However, in dense urban environments, the reliability of Global Navigation Satellite System (GNSS) positioning is compromised due to signal blockage and multipath errors. In these circumstances, relying entirely on the vehicle’s Inertial Measurement Unit (IMU) poses challenges due to the accumulation of drift errors over time. To bridge the gap, proprioceptive sensors like cameras, Lidar, and radar can enhance IMU performance, providing a more accurate positioning solution. This work focuses on radar sensors for their resilience to environmental factors like rain, fog, and low illumination and their ability to measure unique properties such as radar cross section (RCS) and Doppler velocity. Radar point clouds in urban environments are often cluttered with noise, ghost detections, and cross-talk, leading to potential registration errors, particularly parked vehicles resembling static landmarks. This paper introduces a combined velocity and geometric filtering approach to improve radar point clouds by removing dynamic objects, noise, and some parked vehicles. Refining the radar data enhances the performance of map registration techniques in urban environments. The proposed filtering method reduced the root mean square error (RMSE) of radar scan to map registration approach by 31%, highlighting its potential to enhance autonomous vehicle navigation.