Experimental data of \Robust 4D Radar Odometry with Heatmap Feature Encoding and Spatiotemporal Attention Network\

To support open research and facilitate reproducibility in radar odometry studies, we release all experimental localization data used in our benchmark evaluations. The dataset consists of raw positioning data, directly collected from the radar system during experiments, without any form of preprocessing or filtering. All data are stored in plain text (.txt) format.Each file contains timestamped pose information corresponding to the full set of sequences evaluated in our study. These include both indoor and outdoor environments and cover a diverse range of motion patterns and scene structures. Ground truth trajectories are also included for the evaluation sequences, obtained via high-precision LiDAR-IMU or motion capture systems.This release allows researchers to perform independent benchmarking, trajectory reconstruction, and error analysis based on the exact data used in our comparisons between the proposed CNN+Attention model, its CNN-only variant, and the RIO baseline (ICRA 2022). The .txt format ensures ease of use across different toolchains and platforms.