A Pragmatic Trade-off Between Deployment Cost and Location Accuracy for Indoor Tracking in Real-Life Environments
This repository contains the data used in the evaluation of our paper "A Pragmatic Trade-off Between Deployment Cost and Location Accuracy for Indoor Tracking in Real-Life Environments" accepted for publication in the proceedings of the 2024 IEEE International Conference on Localization and GNSS (ICL-GNSS 2024).
Indoor Tracking Systems (ITS) are based on different technologies that often require infrastructure challenging to deploy and maintain, such as Bluetooth Low Energy (BLE) or Wi-Fi beacons or gateways. Because of this challenge, infrastructure-less ITS based on inertial and/or magnetic sensors are emerging as an alternative approach.
While this approach reduces the cost of deploying and maintaining infrastructure, it suffers from an accumulation of errors impacting their accuracy over time. In this paper, we propose a solution that balances location accuracy and infrastructure cost by using a small number of beacons to correct the positioning errors of the Pedestrian Dead Reckoning (PDR) algorithm used at the core of most inertial-based ITS.
Our main contribution is an optimized placement of beacons using a computationally efficient geospatial indexing technique, which relies on real-life data to compute the most frequent location of mobile phone users in a given environment. Our solution allows us to position the beacons at the most visited locations and to observe how they impact the location estimates of mobile devices.
We demonstrate that our approach significantly decreases the density of beacons compared to other solutions proposed in the literature. With an optimized placement of beacons, we also improve the tracking accuracy compared to the standalone PDR algorithm. With only three beacons, in a test area of 8400 m2, we achieve an accuracy increase of 9%, and up to 26% with six beacons.
Each evaluation folder contains three subfolders:
inertials: contains the trajectories generated by the PDR algorithm alone (without beacons) and using the raw inertial data collected by participants.groundtruths: contains the ground truth trajectories of the participants.corrected_trajectories: contains the trajectories generated by the PDR algorithm and corrected using beacons.
The number of top beacons used in the evaluation is indicated in the folder name.
Each trajectory is made of a set of tuples composed of timestamps, latitudes and longitudes.
The training-set folder contains the trajectories used to compute the most frequent locations of the participants.
Beacons are placed at the top-6 most visited locations defined in the beacons-locations.csv file. In this file, the beacons are sorted by the number of visits.
If you use this data in your research, please cite the following paper:
Coming soon