We provide a UWB time-difference-of-arrival(TDOA) measurement dataset. We collected the UWB measurements based on the Loco Position System(LPS) from Bitcraze using TDOA2 mode. The groundtruth is provided by a motion capture system with 10 Vicon cameras. The TDOA measurements were logged from a Crazyflie 2.0 nano quadcopter. The groundtruth measurements from Vicon system were synchronized with the UWB TDOA measurements so as to compute the TDOA measurement biases. A training dataset was created with measurements collected from seven different UWB anchor constellations. A neural network is used to capture and correct the TDOA measurement bias for accurate state estimation.
- Python 3.x
- matplotlib
- numpy, panda
- scipy, sklearn
- joblib
- pytorch (version 1.6.0)
- tkinter
Run meas_visual.py
and select the measurement data csv
Run bias_visual.py
and select the UWB TDOA measurement bias data csv. To visualize the NN regression, set 'ShowNN' to 'True' and select network with/without anchor orientation by setting 'WithAn' to 'True' or 'False'.
- /training_csv/meas_data/#Date/#Trajectory
- /training_csv/bias_data/#Data/#Trajectory
- /testing_csv/meas_data/#Date/#Trajectory
- /testing_csv/bias_data/#Data/#Trajectory
- /testing_csv/bias_data/#Data/#Trajectory/AnchorPos_#Date.npy
- /testing_csv/bias_data/#Data/#Trajectory/AnchorQuat_#Date.npy
- row i: position of UWB anchor i in the inertial frame x [m] | y [m] | z [m]
- row i: unit quaternion of UWN anchor i in the inertial frame q.w | q.x | q.y | q.z
x [m] | y [m] | z [m] | Vicon measurement [m] | UWB TDOA measurement [m] | time [s]
Delta x i [m] | Delta y i [m] | Delta z i [m] | Delta x j [m] | Delta y j [m] | Delta z j [m]
azimuth angle cf i [deg.] | elevation angle cf i [deg.] | azimuth angle cf j [deg.] | elevation angle cf j [deg.]
azimuth angle an i [deg.] | elevation angle an i [deg.] | azimuth angle an j [deg.] | elevation angle an j [deg.]
UWB TDOA bias [m] | time [s]