Radar has the advantage of being robust in sensing even in adverse weather like rain and snow because it uses long-wavelength radio waves. Unlike cameras and lidar, radar can continuously collect high-quality data in a variety of driving environments, so it plays a crucial role in autonomous driving systems, which is essential for safe driving in all weather conditions. In terms of deep-learning-based odometry estimation studies, despite the benefits of these radar sensors, the radar sensor does not receive much attention because it is difficult to utilize radar data in comparison to other sensors (i.e., camera and lidar). In this paper, we propose a deep learning network for estimating odometry using 3D radar data as input and Real-Time Kinematic (RTK) information as the label. |
