Skip to content

Latest commit

 

History

History
68 lines (51 loc) · 4.58 KB

PARAMETERS.md

File metadata and controls

68 lines (51 loc) · 4.58 KB

Parameter Explanations

/launch/radar4motion.launch

  • b_flag_run_vod_dataset (bool)

    • If set to true, the node will read point clouds from the View-of-Delft dataset located in str_vod_dataset_base_path.
    • If set to false, the node will switch to online mode, subscribing to a live topic (e.g., /afi910_cloud_node/cloud) of type sensor_msgs/PointCloud2.

  • str_vod_dataset_base_path (string)

    • Specifies the path where the VoD dataset is located.
    • If you have symbolically linked the dataset to a local dataset/ folder (e.g., ln -s /path/to/view_of_delft_PUBLIC dataset/view_of_delft_PUBLIC), set this to $(find radar4motion)/dataset/.

  • sequence_num (int)

    • Indicates which VoD dataset sequence to load for offline processing.
    • The actual sequence metadata (clip information) is stored in text files (.txt) under the clips/ folder.
    • Note: We gratefully utilized the .txt files provided by CMFlow). However, slight modifications were made as the sequence numbers in the provided files differ from those referenced in major papers. Please ensure the correct mapping when using the sequences.

  • i_vod_frame_rate (int)

    • Defines a virtual frame rate (in Hz) for sequentially reading and publishing VoD data.
    • Useful for controlling how fast the offline dataset is “played back” to the odometry node.

  • vod_eval_folder_path (string)

    • When this is set, the system will store evaluation results (e.g., estimated poses) in the specified directory.
    • These results can subsequently be analyzed using tools like evo for odometry evaluation.

  • s_input_points_ (string)

    • Online mode parameter: If b_flag_run_vod_dataset = false, the node will subscribe to this topic to receive live radar/lidar point clouds.
    • Default: /afi910_cloud_node/cloud.

  • str_origin_ & str_base_link_ (string)

    • Used to define the TF frames in the transformation tree.
    • str_origin_ is often set to "Map", while str_base_link_ is the vehicle or sensor reference frame (e.g., "radar").

  • str_rcs_field_name_, str_power_field_name_ (string)

    • Defines the field names for RCS or power in the incoming point cloud.
    • Adjust if your point cloud messages use different field names.

  • b_accum_whole_point_clouds_ (bool)

    • If true, the code will accumulate the entire point cloud over time for visualization or debugging purposes.

  • d_map_voxel_leaf_size_ (double)

    • Sets the voxel size for downsampling the accumulated point cloud.
    • Larger values yield more aggressive downsampling.

  • min_static_points_num (int)

    • Minimum number of “static” points required by the radar ego-motion module.
    • If fewer points are detected as static, a warning will be issued, and the algorithm may rely on a different motion model.

/config/radar_point_cloud_odometry.ini

  • m_i_submap_scan_num

    • Specifies how many scans are retained to build a local “submap.” For example, if m_i_submap_scan_num = 5, the odometry algorithm will accumulate the most recent 5 scans and use them for matching.

  • m_i_m_scan_num

    • Determines how many consecutive scans (“M-scan accumulation”) are grouped together for a single matching step. Increasing this value can improve matching robustness but may also increase computation time.

  • m_d_radar_ego_estimation_vel_x_sigma_threshold_ms

    • Defines the maximum allowed uncertainty (standard deviation) in the radar-based ego-motion estimate along the X-axis (in m/s). If the radar’s estimated velocity uncertainty is larger than this threshold, the system will revert to using an alternative prediction (e.g., KISS-ICP) for the X-axis.

  • m_d_radar_ego_estimation_vel_y_sigma_threshold_ms

    • Same concept as above, but for the Y-axis velocity uncertainty threshold.

  • m_d_radar_ego_estimation_vel_z_sigma_threshold_ms

    • Same concept as above, but for the Z-axis velocity uncertainty threshold.

  • m_i_min_input_points_num

    • Defines the minimum number of points required to perform the odometry update. If a scan contains fewer points than this value, the algorithm will only use a prediction step without matching.

  • Radar Calibration Parameters (m_d_radar_calib_x_m, m_d_radar_calib_y_m, m_d_radar_calib_z_m,

    • m_d_radar_calib_roll_deg, m_d_radar_calib_pitch_deg, m_d_radar_calib_yaw_deg)
      • Provide the translation (in meters) and rotation (in degrees) offsets to transform from the radar frame to the vehicle frame. These values are used to construct an Eigen::Affine3d representing the radar’s mounting position and orientation.