Microsoft Indoor Localization Competition – IPSN 2016

1. Marinis et al. in the 2D category was not considered for the awards. The system required a significantly different evaluation process compared to the official one. The team required to go over specific evaluation points multiple times, and to also record the exact evaluation point they were at, at any given time. Given that no other teams followed this process, this team was not included in the awards.
2. Elias et al. competed with two versions of the same system: one that requires explicit initialization to a known ground truth location before the evaluation, and one where the initial location was automatically inferred by the system based on WiFi and magnetic fingerprints. Both versions of the system complied to the rules of the competition, and the version of the system with initialization clearly achieved the lowest error. However, given that the system from Su et al. that achieved very similar accuracy to that of Elias et al. did not use explicit initialization, we believe that a fair comparison can only be made among the Elias et al. without initialization and the rest of the teams. Given this, Su et al. got the first place, and Elias et al. the second place. Given the technology leveraged by both teams, the achieved accuracy by these teams is practically identical.
3. Morin et al. in the 3D category was an exhibition team from the competition’s sponsor Leica. They used a backpack equipped with 5 high-end camera modules, 2 LIDARs and a commercial-grade GPS receiver. This system was not a real-time system as they had to post-process image data and manually select the evaluation points in these images. As a result, it was not considered for the awards.
4. Flores et al. in the 3D category: during the evaluation process, the organizers wrote down the wrong x-coordinate for evaluation point 15. Even though the team reported (13.38, 0.44, 3.65), the organizers wrote down (30.38, 0.44, 3.65). With this correction, the average location error for this team is 1.74m and not 2.86m.
5. Kulm et al. in the 3D category reported coordinates without adjusting for their 1.5m tall tripod they used to attach their tag to. Unfortunately, they informed the organizers about this after the competition was completed. When adjusting for the 1.5m fixed offset of the tripod, the location error of this team comes down to 0.43m from the 1.43m shown in the graphs above.

We are happy to announce we have received 49 submissions this year! All 49 submissions have been classified into two categories: 3D and 2D localization. All 27 teams in the 3D localization category require to deploy custom hardware, and will have to provide 3D locations during evaluation. All 22 teams in the 2D localization category rely on already available signals such as WiFi and FM, and will have to report 2D locations during evaluation.

Teams in green fonts have at least one team member officially registered for the competition, and will be allowed to deploy their systems during the competition. Teams in red fonts do not have any of their team members officially registered, and therefore they will NOT be allowed to deploy their systems during the competition. (based on registration data as of Monday March 21st).

The detailed schedule for both days of the competition: