Logo: University of Southern California

Events Calendar



Select a calendar:



Filter September Events by Event Type:


SUNMONTUEWEDTHUFRISAT
31
1
2
3
4
5
6

7
8
9
12
13

14
15
16
17
18
19
20

21
22
23
25
26
27

28
29
30
1
2
3
4


University Calendar
Events for September

  • PhD Defense- Andrew Goodney

    Thu, Sep 11, 2014 @ 11:00 AM - 01:00 PM

    Thomas Lord Department of Computer Science

    University Calendar


    Multipulse Acoustic Tomography

    PhD Candidate: Andrew Goodney

    Date: Thursday Sept. 11th, 11am
    Location: EEB132
    Committee:

    Dr. Young H. Cho (chair)
    Dr. Shanghua Teng
    Dr. Bhaskar Krishnamachari (outside member)

    Abstract:

    Water temperature sensing in the marine environment is an important task motivated by use of the oceans for industrial purposes, national defense and ecological research. Scientists and engineers desire real-time, high resolution temperature maps so that the dynamic nature of underwater processes can be tracked and understood with high fidelity. Point sensors, which only sense in one location, require dense deployments of sensor nodes to provide high resolution water temperature maps and thus spatial resolution is constrained by the cost of deploying additional nodes. Robotic boats and autonomous underwater vehicles (AUVs) can sample at high spatial resolutions, however they can only sample at one location at a given time. Therefore such water craft can miss dynamic water temperature events that might occur in the area of interest, but not directly at the current location of the robot or AUV. These water craft are also known for being fabulously expensive, with a per unit costs on the order of $100k.

    Point sensors can be deployed as underwater wireless sensor network where the communication links are formed through the water using acoustic modems. Our initial work for this dissertation made the observation that since speed-of-sound in water is dependent on the water temperature, the acoustic communications links could be used for sensing between the nodes by measuring the time-of-flight for a signal between two nodes. Known as acoustic tomography, this technique had previously been applied in the ocean with internode distances measuring 100's to 1000's of kilometers. In order to implement acoustic tomography with a sensor network we developed two novel and highly precise time-of-flight measurement techniques on top of a novel coding scheme that allows underwater sensor networks to combine data and sensing transmissions.

    Acoustic tomography increases the resolution of water temperature maps over those drawn from just point sensors by reconstructing the sound-speed field located between the sensor nodes by using the time-of-flight data for signals sent between the nodes. However, current travel-time tomography techniques make the assumption that for each node pair only one data point can be observed: the average speed-of-sound along the path between the nodes. The resolution of a reconstruction is thus limited by the number of paths that cross through an area of interest, and therefore to increase the resolution of an acoustic tomography system more nodes must be added in strategic locations, which may not always be possible due to cost and/or other limitations.

    Our third contribution, and the topic of todays defense talk is a travel-time acoustic tomography technique that breaks the "average assumption." With our technique we are able to derive a spatial distribution of water temperature along each acoustic path, and the spatial resolution of the distribution is not limited by node density, but by the time-of-flight measurement precision. In today's talk I will introduce the technique, which we call multipulse acoustic tomography, and show that if a stream of bidirectional pulses (signals) is sent between two nodes, subtle differences in time-of-flight can be used to build a spatial distribution of temperature along the acoustic path. We show through simulation that the technique works under a set of conditions and constraints that resembles those found in our maria testbed. We also show that the technique improves the resolution and accuracy of reconstructions by disambiguating data that would otherwise pose a problem for traditional acoustic tomography techniques. Finally, multipulse tomography generalizes to any travel-time tomography domain and we discuss how it could have important applications in other fields such as power grid monitoring

    Location: Hughes Aircraft Electrical Engineering Center (EEB) - 132

    Audiences: Everyone Is Invited

    Contact: Lizsl De Leon

    Add to Google CalendarDownload ICS File for OutlookDownload iCal File