The Impact of Using Multiple Antennas on Wireless Localization - - PowerPoint PPT Presentation
The Impact of Using Multiple Antennas on Wireless Localization Konstantinos Kleisouris Computer Science Department Rutgers University Joint work with: Prof. Yingying Chen, Jie Yang, Prof. Richard P. Martin (advisor) Localization Office
Joint work with:
Chen, Jie Yang,
variety of computing devices to communicate wirelessly
for communication but for localization of devices in 2D and 3D
(X, Y) Localization
Office Floor
use landmarks and a training set
traffic at known positions
radio properties and locations
Strength (Si), Angle of Arrival (AoA), Time of Arrival (ToA)
Strengths (Si) measured at some location Office Floor
landmark landmark landmark [X, Y, S1, S2, S3] S1 S2 S3 [X?, Y?, S1’, S2’, S3’] S1’ S3’ S2’ fingerprint
Received Signal Strength (RSS) is affected indoors by
Difficult to associate signal strength to location Can we alleviate the impact of RSS variability on the
Investigated signal strength variability when employing
Investigated the effects of using multiple antennas on
Multiple antennas can average out environmental effects
Multiple antennas can improve the localization accuracy
Methodology RSS Variability Study Stability & Accuracy Results Conclusions & Future Work
fading effects
location
an opportunity to smooth out these effects
antennas at a given landmark location
conducted in the yellow area
stars) different locations
antennas per landmark location (1-2 ft from each
spots where we collected data
219 ft 169 ft
Coordinates (in feet), Description (x, y, 0) Center (x, y, 3) East (x-1, y, 3) West (x+1, y, 3) North (x, y+1, 3) South (x, y-1, 3) Vertical (x, y, 3), monitor vertical to the floor Parallel (x, y, 3), monitor parallel to the floor (x, y, 5.16) Desk Placement Floor Shoulder
in response to small-scale movements of a mobile device
locations p2, p3, …, pn
RSS Variability Study Stability & Accuracy Results Conclusions & Future Work
variations on signal strength?
antennas to a theoretical propagation model
determination R2
1
Free Space Model
the RSS for all 3 antennas (3-antenna-avg) achieves the best fit
does improve the data fit to a simple free-space model
Algorithms
Results
(x, y) plane: Center, North, South, East, West, Vertical, Parallel z-axis: Center, Floor, Shoulder Center placement is always the original p1 position
Desk, Center
(x, y) plane z-axis
(30%)
Desk, Center, No Train., Test.=51
(x, y) plane, No Train., Test.=51 z-axis, No Train., Test.=51
(44%)
(34%)
landmarks, although the impact is not huge
better for localization algorithms
what the limiting number is where improvements tail off
Using Monopole Antennas
Desk, Center Gaussian
performance when compared to real data
Floor Shoulder
(28%)
21.7ft (29%)
Desk, Center Gaussian
(75%)
compared to real data
Floor Shoulder
Center
Center
(x, y) plane z-axis
(74%)
Desk, Center, Train.=100, Test.=1 Gaussian, Train.=100, Test.=1
curves
the RSS has the same performance
Desk, Center, No Train., Test.=51 Gaussian, No Train., Test.=51
(48%)
the RSS has the same performance
Floor, N=NA=51 Shoulder, N=NA=51
(34%)
(42%)
(x, y) plane, Train.=100, Test.=1 (x, y) plane, No Train., Test.=51
(33%)
(44%)
M1 M2 M3 Basic Similar Coefficients Corridor
distance between training/testing fingerprints on the results when averaging and not averaging the RSS
what the limiting number is where improvements tail off
cheap wireless communication in computing devices
traditional communication networks
(X, Y) Localization
Office Floor
Algorithms
(IMG)
Floor is divided into tiles of size 10in × 5in Derives an expected fingerprint for each tile
Results
Real data Gaussian data set
(x, y) plane: Center, North, South, East, West, Vertical, Parallel z-axis: Center, Floor, Shoulder
Algorithms
Results
Real data Gaussian data set
(x, y) plane: Center, North, South, East, West, Vertical, Parallel z-axis: Center, Floor, Shoulder
movements around a testing spot
antenna and averaging the RSS from different antenna combinations
distribution