ekf-multi-object-tracking

EKF-based Multi Object Tracking

Features

The ekf_multi_object_tracking package supports the following features:

• Low-Frequency Detection to High-Frequency Tracking: Converts low-frequency detection data into high-frequency tracking data, ensuring smooth and accurate object tracking.

• Multiple Localization Inputs (TODO): Supports various localization input sources, allowing flexibility in integrating different localization systems.

• Multiple Prediction Models: Offers a selection of prediction models, including Constant Velocity (CV), Constant Turn Rate and Velocity (CTRV), Constant Acceleration (CA), and Constant Turn Rate and Acceleration (CTRA), to suit different tracking scenarios.

• Individual Time Consideration for Rotating LiDAR: Takes into account the individual detection times of objects from rotating LiDAR sensors, improving the accuracy of motion compensation and tracking.

• Vehicle Heading Correction: Provides functionality to correct the vehicle's heading direction, enhancing the accuracy of the tracking system.

• Class Correction: Supports class correction features, allowing for improved classification accuracy of tracked objects.

These features make the package versatile and adaptable to a wide range of multi-object tracking applications.

I/O

Input

Description	Type
Detection info	jsk_recognition_msgs::BoundingBoxArray
Localization info (optional)	nav_msgs::Odometry (optional)

Output

Description	Type
Track Bounding Box	jsk_recognition_msgs::BoundingBoxArray
Track Info	visualization_msgs::MarkerArray

How to use

1. Install ROS messages and other libraries

• for Ubuntu 20.04 (noetic)

sudo apt install ros-noetic-jsk-rviz-plugins

2. Install Code

cd ~/catkin_ws/src
git clone https://github.com/jaeyoungjo99/ekf-multi-object-tracking.git
cd ..
catkin_make

3. Launch Code

source devel/setup.bash
roslaunch ekf_multi_object_tracking ekf_multi_class_object_tracking.launch 

Configuration

The behavior of the ekf_multi_object_tracking node can be customized using the config/config.yaml file. Below is a description of the key configuration parameters:

Topic Names

• lidar_objects: Topic for LiDAR object data. Topic type is jsk_recognition_msgs::BoundingBoxArray. Default: /hmi/perc/jsk_pillar_objects

Configuration Options

• input_localization: Selects the input localization source.
  • 0: None
  • 1: nav_msgs::Odometry
  • 2: NavSatFix (TODO)
• global_coord_track: Whether the tracker is performed in global coordinates
  • true: Global coordinate tracker
  • false: Local coordinate tracker
• output_local_coord: Determines if the output is in local coordinates.
  • true: Output local coordinate
  • false: Output global coordinate
• output_period_lidar: Synchronizes output with LiDAR.
  • true: LiDAR synced output
  • false: Motion synced output
• output_confirmed_track: Outputs only confirmed tracks if set to true.
• use_predefined_ref_point: Uses a predefined reference point if set to true.
• reference_lat, reference_lon, reference_height: Predefined reference point coordinates.
• cal_detection_individual_time: Enables LiDAR motion compensation.
• lidar_rotation_period: Sets the LiDAR rotation period.
• lidar_sync_scan_start: Synchronizes LiDAR time to scan start if set to true.
• max_association_dist_m: Maximum distance for track association.
• prediction_model: Selects the prediction model.
  • 0: CV
  • 1: CTRV
  • 2: CA
  • 3: CTRA
• system_noise_std_xy_m, system_noise_std_yaw_deg, etc.: System noise parameters.
• meas_noise_std_xy_m, meas_noise_std_yaw_deg: Measurement noise parameters.
• dimension_filter_alpha: Filtering parameter for object dimensions.
• use_kinematic_model: Aligns velocity direction to heading if set to true.
• use_yaw_rate_filtering: Restricts yaw rate based on velocity.
• max_steer_deg: Maximum steering angle.
• visualize_mesh: Enables mesh visualization if set to true.

Vehicle Origin

• vehicle_origin: Defines the vehicle origin.
  • 0: Rear Axle
  • 1: C.G.

Transformations

• rear_to_main_lidar: Transformation from rear axle to main LiDAR.
  • parent_frame_id: Parent frame ID
  • child_frame_id: Child frame ID
  • transform_xyz_m: Translation in meters
  • rotation_rpy_deg: Rotation in degrees
• cg_to_main_lidar: Transformation from C.G. to main LiDAR.

Users can modify these parameters in the config/config.yaml file to suit their specific requirements.