A ROS package for tracking using ultra-wideband (UWB) radios. The target needs one UWB tag and is localized by a tracker with multiple UWB radios (i.e. a robot).
The package is made up of three nodes:
- uwb_serial: Reads binary messages from a serial port. The corresponding UWB driver for an embedded board are provided at https://developer.mbed.org/users/bhepp/code/AIT_UWB_Tracker/.
- uwb_multi_range: Processes the multi-range timestamps from uwb_serial and publishes calibrated and uncalibrated ranges.
- uwb_tracker: Processes the calibrated ranges from uwb_multi_range and tracks the target position using an Extended Kalman Filter. It publishes the state and covariance of the filter but also a corresponding transform.
- Boost (system and thread module)
- numpy
- scipy
- rospy
Checkout the repository in your ROS catkin workspace and build the workspace as usual.
You can start all three nodes using
roslaunch uwb uwb.launch
Many parameters like the serial port, baudrate, transform frames and parameter file can be specified as arguments. See the launch file for details. More low-level parameters can be defined via rosparam (the default parameters should be fine in most cases). Check out the code of the nodes for details.
The folder utils contains the MATLAB script calibrate.m. This generates a YAML file containing offsets for each UWB unit in the tracker coordinate frame as well as linear coefficients for the calibrated range-measurements. The uwb_multi_range node uses the YAML file to generate calibrated range measurements.
A motion-capture system is useful for generating the calibration data but it could also be measured by hand. The file calibrate_example.m in the utils folder shows how the data for calibrate.m can be generated from motion-capture data.
Note: When recording data for calibration make sure to use the uncalibrated measurements from the uwb_multi_range node (otherwise you might use already calibrated values and get a wrong calibration).