Localization Teams - Track 1

Team name

ZETE


Corresponding author

Matteo Tomasi

Affiliation

Zetesis S.r.l (Italy)

Description

Aiming to develop a smartphone app for indoor localization, we propose using the embedded motion sensors to improve scene recognition. In our method, pairs of snapshots taken in different directions are matched to a set of 360º panoramic images, each representing a location of an indoor environment. The consistency between calculated snapshot rotation based on panorama and rotation measured by motion sensors helps reduce false matches. The feature matching is based on FREAK (Fast Retina Keypoint) descriptor, which is suitable for smartphone implementation.

 


 

 

Team name

NAIN

 

Corresponding author
J. C. Aguilar Herrera

Affiliation
Navix Indoor Navigation (Mexico)

Description

WiFi beacons are deployed in two levels in a testbed indoor environment using fingerprinting and trilateration signal-based positioning methods in order to evaluate the results on different densities of beacon arrays and positioning methods. The positioning estimate obtained from the signal-based method is then integrated with smartphone sensors PDR (Pedestrian Dead Reckoning), probability map based on pedestrians tracked location and map information of the indoor environment (e.g. walls and obstacles). These positioning methods have been implemented in a smartphone application to evaluate their performance in the first and second floor in the ITESM Technology Park. In order to use the tri-lateration method delivering high position accuracy and avoid the fingerprinting time expensive periodic calibration and maintenance, an optimal beacons distribution is needed, the comparative results of this work provide a method to optimize the installation in new indoor environments by using the building layout and beacons propagation model.


 

Team name

MMSS

Corresponding author

You Li

Affiliation

University of Calgary (Canada) and Wuhan University (China)

Description

The proposed system is an indoor navigation algorithm that uses multiple kinds of sensors and technologies, such as 9- axis sensors (i.e., 3D gyros, accelerometers, and magnetometers), WiFi, and magnetic matching. The corresponding real-time software on smartphones includes modules such attitude determination and gyro bias estimation, pedestrian deadreckoning (PDR), WiFi positioning, and magnetic matching. The heading from the attitude-determination module is feed into the PDR-based position-tracking module. Then, PDR is used for providing continuous position solutions and for the blunder detection of both WiFi fingerprinting and magnetic matching. Meanwhile, WiFi fingerprinting utilizes a point-by-point matching technology, while magnetic matching is based on profile-matching. Finally, WiFi and magnetic matching results are passed into the position-tracking module as updates.


 

 

Team name

SAMS

 

Corresponding author

Pawel Wilk

Affiliation

Samsung R&D (Poland)

Description

Indoor Positioning Engine provides real time indoor location to all interested location aware Smartphone applications. The system relies on information provided by device’s commonly found sensors: Wi-Fi, BLE and motion sensors. Information is fused in order to determine position of people holding phone devices. Additional constraints taken from floor plan data are used to refine and smooth the walked paths. Calculated position is provided to all registered listeners. The solution implements Wi-Fi fingerprinting based approach consisting of two phases:

  • Offline training data collection served to build mapping between geometric coordinates and radio signal readings
  • Online position determination from best matched trained data with actually seen fingerprint.

 

Localization Teams - Track 2

Team name

NESL

nesl logo

Corresponding author

Chan Gook Park

Affiliation

Seoul National University(South Korea)

Description

In Pedestrian Dead Reckoning (PDR) system, heading error is one of the main factors that cause estimated position error. In order to reduce heading error, several methods are proposed by using an assumption that generally walls and corridors are straight and either parallel or orthogonal to each other in manmade building. They called the typical directions of walls and corridors as the dominant directions or cardinal directions. However these methods have limitation when the pedestrian is not walking along the corridors for a long period of time because these methods can cause a new azimuth error by matching the closed dominant direction. To overcome these limitations, we implemented INS-EKF-ZUPT (IEZ) based Advanced Heuristic Drift Elimination (AHDE) which can remove azimuth drift error in indoor environments.

 


 

Team name

DATA


Corresponding author

Rinara Woo

Affiliation

Center for Embedded Software Technology Daegu (South Korea)

Description

The proposed position tracking system with foot-mounted IMU sensor by using the dominant acceleration tracking algorithm (DATA) which compensates positioning error using the dominant acceleration direction of pedestrian walking. In DATA, it analyzes the pedestrian step and extracts dominant direction from acceleration data in one step. We essentially obtain orientation from Madgwick algorithm and compensate orientation using DATA. We use ZVU to minimize the accumulated position error. These methods were tested on real and they have shown improvements over the Madgwick algorithm without DATA.

 


 

Localization Teams - Track 3

Team name

RTLS@UM

Corresponding author
Adriano Moreira

Affiliation
University of Minho (Portugal)

Description
Among the several techniques being proposed for indoor positioning, solutions based on WiFi fingerprinting are the most popular since they exploit existing WLAN infrastructures to support software-only positioning, tracking and navigation applications. Despite the enormous research efforts in this domain, and despite the existence of some commercial products based on WiFi fingerprinting, it is still difficult to compare the performance of the several existing solutions.


 

Team name

HFTloc

Corresponding author
S. Knauth

Affiliation
Stuttgart University of Applied Sciences (Germany)

Description
We apply our "Fingerprint Calibrated Centroid Localization" FCWC for the given task. FCWC comprises two tasks: the first task is to determine “virtual positions” of access points out of calibration measurements, i.e. RSSI Fingerprints. The fingerprint data is taken from the UJIIndoorLoc database and comprises RSSI information and meta information like measurement position, building, floor, device model etc.  The AP positions are estimated by reverse positioning using weighted centroid. Once the APs are determined, rover positioning can be performed. This is again performed by applying weighted centroid, this time the position is obtained via the weighted vector sum of the virtual positions of the access points. A second alternative approach is based on a fingerprinting algorithm with a comparison norm based on the scalar product over weighted RSSI values.


 

Team name

MOSAIC

mosaic_eng_rgb

Corresponding author
Rafael Berkvens

Affiliation
University of Antwerp  (Belgium)

Description
We are interested in quantifying the localization performance of exteroceptive sensors solely by virtue of their sensor model using information theory. For the third track of the EvAAL competition, we propose a probabilistic WiFi measurement model. The Maximum Likelihood Estimation and k Nearest Neighbor are used as a localization scheme to our measurement model.


 

Team name

ICSL

Corresponding author
H. Jin Kim

Affiliation
Seoul National University (Korea)

Description
The proposed localization method can solve step-by-step the localization problem, first with a building classification then with a floor classification and finally with a position regression. Dimensionality of this dataset is high, but the most of elements in a data vector are empty. Dimensionality reduction algorithm not only can extract important WiFi feature, but also reduce much dimensionality of the raw dataset.


 

 
Joomla SEO powered by JoomSEF