PhD Defense: "Distributed Estimation on The Graph Cycle Space"

Wm. Joshua Russell

September 1st (Wednesday), 10:30am
HFH 4146

As large scale sensor networks have become more prevalent in military and consumer applications, increasing interest has been paid to distributed estimation and its applications. In this talk we consider estimation from measurements of relative differences in state. A typical example of a “relative differences” problem is estimating the absolute position of a set of sensors from measurements of displacement between sensors.

We present a method where the network is represented as a graph and the network’s state is estimated on the cycle space of the graph. We show that this estimation method is distributable and for a large class of graphs the distributed algorithm converges with fewer communications then a pre-existing algorithm.

Finally, we show that the existing distributed estimation methods can be extended to the problem of estimating the positions of a group of vehicles from inertial measurements and relative position measurements between vehicles. We show that using little communication, the distributed algorithm approaches solutions achieved by a Kalman Filter.

About Wm. Joshua Russell:

Wm. Josh Russell completed his undergraduate education at the University of Washington in 2006 and subsequently enrolled as a graduate student at UCSB. He earned a MS from the ECE department in 2008. As well as working toward a Ph.D. in the ECE department, Josh is currently working on a MA in economics. He hopes one day to apply his knowledge on estimation and optimization to the financial world.

Hosted by: Joao Hespanha