NSF_Logoumblogocp_logo_lrg
This project is funded by the NSF Award CNS-1116430 (2011-2014)

Research Description

We are interested in enabling techniques for target monitoring using low-cost sensor networks. A “sensor” is a small computing device that can read the environment and process information; it can be a mobile phone or an autonomous underwater vehicle with sensing capabilities, or typical sensors as in a conventional sensor network. We want to devise a solution that can tell us whether a special event of interest happens, when and where, quickly, accurately, and efficiently.

The key feature of our development is “low-cost”. A sensor node needs not built-in capability to determine its location (e.g., by having a GPS) and there should not require any expensive external assisting device such as a robot to help with the tracking. Our vision is that given today’s ubiquitous communications infrastructure and increasingly advanced mobile computing at cheaper price, we could achieve the quality of an expensive target monitoring solution, which is often centralized, by utilizing low-cost easy-to-deploy sensing devices and easy-to-obtain information about the environment such as radio signal strength, sound, and geomagnetic field. We want to support a wide range of sensor networks, which can have modest resource capacities or be deployed in non-conventional physical settings such as under the water or on wet ground.

Especially, we want to apply our research to tracking of Autonomous Underwater Vehicles (AUVs) deployed in Boston Harbor, in collaboration with UMass Center for Coastal Environmental Sensing Networks. When at the surface AUVs utilize GPS for position tracking, but when underwater they rely entirely on an inertial navigation system for localization.

ThimbleIsland

Here is a Google earth image of 5 AUV missions to go on the surface in a prescribed rectangular pattern near Thimble Island (leaf-like island seen in picture); the AUV path in each mission is represented by a color. Although they were each just repeats of the same mission and we can see a lot of variability in the actual trajectory even with the AUV remaining on the surface the whole time. We expect much worse variability while the AUV is under the surface. Our approach is based on the observation that the aquatic environment should offer a rich fingerprint that the AUV can sense in real time and utilize for better tracking.

Results

We have proposed and evaluated a set of techniques to serve the purposes described above.

  • A decentralized framework for notification services in sensor networks. This is useful for target monitoring applications as it allows for fast notification upon triggering of an event.
  • A regularization framework for offline and online localization and tracking using fingerprint data.
  • Tensor decomposition approach for localization using multi-modal fingerprint data.
  • A hierarchical machine learning approach for fast and efficient localization.
  • Low-cost techniques for human activity recognition using smartphones and accelerometers.

Inspired by the research, we would like to continue exploring the algebraic approach, not only to the localization of a target, but also to recognition of the activities the target is doing.

Participants


Senior Personnel

  • PI: Prof. Duc A. Tran, University of Massachusetts at Boston
  • Co-PI: Prof. Bridget Benson, California Polytechnic State University
  • Technical support: Francesco Peri, Manager, UMass Center for Coastal Environmental Sensing Networks

PhD Students

  • Pejman Ghorbanzade (in progress)
  • Ting Zhang (in progress)
  • Siyuan Gong (in progress)
  • Quynh Vo (in progress)
  • Cuong Pham (graduated, 12/2012)
  • Khanh Nguyen (graduated, 12/2012)

Undergraduate Students

  • Vy Thao Nguyen (summer 2012)
  • Vy Thuy Nguyen (spring 2013)
  • Phong Truong (spring 2013, summer 2013)
  • Kwaku Farkye (summer 2013, visiting from Cal Poly)
  • Ken Ugo (summer 2013, visiting from Cal Poly)

photo1
RA students working in Summer 2013; left to right: Ken Ugo (undergrad), Kwaku Farkye (undergrad), and Ting Zhang (PhD student)

Software


  • We have developed eTrack, a prototype for indoor localization and tracking. Below is a demo showing it work in our lab. For more information, visit eTrack website.


  • We have developed an Android app for sample collection for indoor localization and tracking.
  • Graduate student Ching-Kai Huang has developed a prototype that allows remote control of a low-cost iRobot by an Android phone or a PC

Pasted Graphic

Data



OldHarbor-Undulate-2auv_tracking

  • A mission trajectory for the AUV is shown in picture above, consisting of 18 waypoints. Fingerprint data are collected as the AUV navigates.
  • A rich set of bathymetric and fingerprint data (courtesy of UMass Boston School for the Environment)

umbcs_mapumbcs.trajectory
  • UMBCS Dataset: A fingerprint dataset covering the third floor (Computer Science dept) of the Old Science building at UMass Boston. This dataset includes a set of individual Wi-Fi, FM, and Geomagnetic fingerprints with associated locations and the specification for a test trajectory. This dataset was obtained in 2013.
    • umbcs.validpoints.csv: location coordinates for each sample location where sensor data is collected
    • Wifi_avg.txt: wifi data averaged over multiple readings per location
    • FM_avg.txt: fm data averaged over multiple readings per location
    • Geo_Mag_avg.txt: geomagnetic data averaged over multiple readings per location
    • Trajectories.xlsx: trajectory file
    • *.map.*: floorpan
    • *.trajectory.*: trajectory T1

umbccul.mapIumbccul.trajectory
  • UMBCCUL Dataset: A Wi-Fi fingerprint dataset covering the upper-level floor of the Campus Center at UMass Boston. This dataset was obtained in 2013.
    • umbccul.validpoints.csv: location coordinates for each sample location where sensor data is collected
    • *.map.*: floorpan
    • *.trajectory.*: trajectory
    • umbccul.wifi.scaled.txt: wifi fingerprint data


umbwheatley.map umbwheatley.trajectory
  • UMBWheatley Dataset: A fingerprint dataset covering the first floor of the Wheatley building at UMass Boston. This dataset includes a set of individual Wi-Fi, FM, and Geomagnetic fingerprints with associated locations and the specification for a test trajectory. This dataset was obtained in Spring 2015
    • points.txt: location coordinates for each sample location where sensor data is collected
    • *wifi*: wifi data
    • *.fm.*: fm data
    • *.trajectory.*: trajectory file
    • *.map.*: floorpan
    • *featurelist*: list of feature IDs for the fingerprint
    • raw_reading.zip: all the raw readings (wifi, fm, geomagnetic) for each sample location

Presentations


  • 11/3/2014: Dr. Duc Tran's seminar talk on "Fast and Efficient Fingerprint-based Localization and Tracking", Department Seminar, Department of EECS, University of Central Florida, Orlando

  • 10/30/2014: Dr. Duc Tran's conference talk on "An Online Algorithm for Fingerprint-based Location Tracking" at IEEE MASS 2014, Philadelphia

  • 10/28/2014: Dr. Duc Tran's conference talk on "Fast and Accurate Indoor Localization based on Spatially Hierarchical Classification" at IEEE MASS 2014, Philadelphia

  • 9/16/2014: Dr. Duc Tran's seminar talk on "Fast and Efficient Fingerprint-based Localization and Tracking" at Raytheon BBN Technologies, Waltham, MA

  • 7/30/2013: Dr. Duc Tran's keynote talk on "Fast and Accurate Indoor Localization" at WiMan 2013 (Nassau, Bahamas)


Papers

(those names marked with * are student participants in the project)



Contact Information

  • UMass Boston site: Prof. Duc A. Tran (duc.tran@umb.edu)
  • Cal Poly site: Prof. Bridget Benson (bbenson@calpoly.edu)