RSSI as a Long-Term Localization Metric

By Nathan Hurtig, Maren Sorber, and Artemis Pados
Slide 1: Title slide presenting RSSI as a Long-Term Localization Metric research by Nathan Hurtig, Maren Sorber, Artemis Pados under Dr. Jason Hallstrom

Slide-1

Research presented by Nathan Hurtig, Maren Sorber, and Artemis Pados under the guidance of Dr. Jason Hallstrom.

Slide 2: Context of the Problem discussing benefits of studying person movement and mobility intelligence systems

Slide-2

Context of the Problem

Benefits of studying person movement

  • City planning and businesses
  • Navigation
  • Tourism and advertisement
  • Real estate

How: mobility intelligence system [1]

  • Implementation issues - cameras
  • Use RSSI
Slide 3: Our Contributions outlining four key research contributions including RSSI data collection and model analysis

Slide-3

Our Contributions

Research on RSSI is incomplete

Gap in long-term analysis [2]

  1. Collection of a large sample of RSSI data
  2. Proof that models retain their accuracy
  3. Analysis of RSSI's limits
  4. Distance-flexible models
Slide 4:

Slide-4

The slide shows a diagram and three photos of an experimental setup. The top diagram shows two squares labeled S1 and S2, separated by a line labeled 50', with another square labeled E and a wheeled object between them. Arrows point to the left and right. The height of S1 is labeled as 32". The bottom of the image has three photographs. The left photo shows a wheeled robot with exposed electronics, wires, and a battery on a track made of two white PVC pipes. The middle photo shows the same robot from a different angle, with a gray pillar in the background. The right photo shows a longer view of the track and the experimental area, including chairs and tables in the background.

Slide 5: Ground Truth Evaluation describing robotic platform using LiDAR with error less than 10 cm

Slide-5

Ground Truth Evaluation

Robotic platform finds its location with LiDAR optical distance sensors

We conducted an experiment to measure the robot's accuracy

Error less than 10 cm

Slide 6: Data Collection showing 24/7 robotic platform operation with over 300 hours and 35 million RSSI measurements

Slide-6

Data Collection

  • Robotic platform ran 24/7
  • 6 days for 30' trial
  • 11 days for 50' trial
  • Over 300 hours of data collection
  • 35 million RSSI measurements
Slide 7: Data Processing describing machine learning applications and statistical analysis results

Slide-7

Data Processing

Applied machine learning to data

Best performers: decision tree, linear regression

Statistical analysis of data

  • Average coefficient of variation: 24%
  • Resolution error: 20 cm

There is a bar chart titled "Power values recorded at single location." The x-axis is "Power (normalized)" from 0.0 to 1.0, and the y-axis is "Frequency" from 0 to 250.

Slide 8: Results showing model accuracy stability over time and distance flexibility

Slide-8

Results

Model accuracy doesn't degrade over time

Apply models at different distances

There is a CDF plot titled "CDF of models tested on 50' data," with four colored lines. The x-axis is "Error defined as proportion of track," and the y-axis is "CDF" from 0.0 to 1.0. The plot on the right is a bar chart titled "Time Analysis of 50' trial." The x-axis is labeled with numbers 1 through 8, and the y-axis is "Accuracy (MAE, proportion of track)" from 0.00 to 0.12.

Slide 9: Key Takeaways highlighting model stability and distance flexibility with RSSI localization accuracy bounds

Slide-9

Key Takeaways

  • Our models are stable enough to remain accurate over long periods of time
  • We can train models at one distance between sensors, and apply them to another distance
  • RSSI localization accuracy has an upper bound
Slide 10: Limitations and Future Work discussing ideal environment testing and one-dimensional experiment constraints

Slide-10

Limitations and Future Work

  • Tests were conducted in a close to ideal environment
  • One-dimensional experiment meant that anisotropic antenna patterns had no impact
Slide 11: References listing two academic sources about mobility intelligence and RSSI localization

Slide-11

References

[1] S. Mazokha, F. Bao, J. Zhai, and J. O. Hallstrom, "Mobintel: Sensing and analytics infrastructure for urban mobility intelligence," in 2020 IEEE International Conference on Smart Computing (SMARTCOMP), 2020, pp. 106–113.

[2] F. Bao, S. Mazokha, and J. O. Hallstrom, "Mobintel: Passive outdoor localization via RSSI and machine learning," Florida Atlantic University, 2020

Last slide: Contains plain text stating 'End of presentation. Click the right arrow to return to beginning of slide show.'

End of Presentation

Click the right arrow to return to the beginning of the slide show.

For a downloadable version of this presentation, email: I-SENSE@FAU.

Additional Information
The Institute for Sensing and Embedded Network Systems Engineering (I-SENSE) was established in early 2015 to coordinate university-wide activities in the Sensing and Smart Systems pillar of FAU’s Strategic Plan for the Race to Excellence.
Address
Florida Atlantic University
777 Glades Road
Boca Raton, FL 33431
i-sense@fau.edu