This repository contains the trajectories used for the paper "Decentralized Collaborative Indoor Tracking".
Each folder contains the trajectories for a specific scenario. The folder name is made of the following elements separated with hyphens:
- Algorithm used for collaboration
- Lower threshold
- Upper threshold
- Starting interval (number of seconds before each device starts moving)
- Number of devices in the experiment
The trajectories are grouped into three folders:
- groundtruths: predefined trajectories executed by each device
- inertials: trajectories obtained using the local PDR algorithm
- corrected_trajectories: trajectories obtained after running the collaborative algorithm presented in the paper.
In addition, parameters.json contains more details on the experiment such as the devices used with their respective number of interactions.
results.csv contains the results of the evaluation for each of the possible scenarios. The metrics that are been used as benchmarks are:
dfd_pdr: Discrete Fréchet Distance of the trajectory generated by the PDR algorithm compared to the groundtruth.dfd_algorithm: Discrete Fréchet Distance of the trajectory generated by the collaborative algorithm compared to the groundtruth.mean_localisation_error_pdr: mean localisation error of the PDR algorithm compared to the groundtruth.mean_localisation_error_algorithm: mean localisation error of the collaborative algorithm compared to the groundtruth.third_quantile_localisation_error_pdr: third quantile of the localisation error of the PDR algorithm compared to the groundtruth.third_quantile_localisation_error_algorithm: third quantile of the localisation error of the collaborative algorithm compared to the groundtruth.