Skip to content

Non-parametric Error Modeling for Ultra-wideband Localization Networks

License

Notifications You must be signed in to change notification settings

acshi/non-parametric-uwb

Repository files navigation

Data and source code files for "Non-parametric Error Modeling for Ultra-wideband Localization Networks"

This project is built using the meson/ninja build system, but the main code has no other dependencies.
Figures where made in Python with numpy and matplotlib.

Dataset5.txt is very similar to dataset4.txt and is only included because it had a particularly
interesting/convoluted distribution used in one of the figures.

To recreate figures from the paper, run `bash do_everything.sh`, which will install dependencies,
make the program, preprocess the datasets, perform the evaluations, and create the figures.

Notes on the datasets:
The robot has node id 16.
Ground truth positions are in feet and are (by node id):
8: (0, 0)
12: (15, 1)
14: (2, 9)
16: N/A
54: (20, 3)
61: (10, 0)
81: (17, 7)
87: (8, 8)

Datasets have the following columns:
time_of_propagation node_id_a node_id_b temperature_1 pp_amp_1 impulse_pwr_1 std_noise_1 fp_power_1 rx_power1
                                        temperature_2 pp_amp_2 impulse_pwr_2 std_noise_2 fp_power_2 rx_power2
                                        temperature_3 pp_amp_3 impulse_pwr_3 std_noise_3 fp_power_3 rx_power3
We only used the first three columns of information for our paper, but the rest is included for completeness.
The meaning of these additional data is best described by the DW1000 user manual.
There are three sets of columns for the later parameters. As we use double-sided two-way-ranging,
communication goes as follows. (1)A -> B. (2)B -> A. (3)A -> B. (4)B -> A.
The three sets of columns correspond to the receive (rx) information of the messages sent in (1), (2), and (3).
The only reason message (4) is sent is to relay receive information of message (3) back to node A.

Dataset 1 actually has an abundance of data: roughly 4096 measurements per link, instead of 1024.
Datasets 1, 2, and 3 also contain extra measurements from node 16 to all the other nodes.
Each time these repeat, the robot has moved to a new position.
While we thought this extra information might have improved results,
we didn't actually see that happen, and so left that out of the paper.

About

Non-parametric Error Modeling for Ultra-wideband Localization Networks

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published