WAVERAD: An mmWave Radar-Centered Multimodal Dataset for Environmental Perception in Inland Waterways

WAVERAD is an open-access, multimodal dataset designed for environmental perception and navigation research in inland waterway environments, with a particular focus on the evaluation and development of mmWave radar systems for maritime and riverine applications.The dataset comprises synchronized recordings from two 77 GHz FMCW automotive radar sensors, one Velodyne VLP-32C 3D LiDAR, an inertial measurement unit (IMU), two stereo camera systems, and GNSS. It contains 65 complex scenes with a total of 48,618 frames, recorded over more than five hours along 37 km of the Berlin waterways (River Spree). The data captures diverse and challenging real-world conditions, including ship locks, underpasses, bridges, vessel encounters, container terminals, docking maneuvers, and urban waterfront environments with buildings, sidewalks, and railway infrastructure.In addition to the raw sensor data, WAVERAD provides radar and LiDAR-based environmental maps generated using two approaches: FAST-LIO2, a state-of-the-art LiDAR-based SLAM algorithm used for high-accuracy ground truth and ego-motion estimation and a filtered mapping method, a partially LiDAR-dependent radar mapping approach optimized for inland waterway perception.The dataset structure and organization are illustrated in the accompanying Dataset_Structure diagram.