Skip to main content

Andrew Symington

Personal photo - Andrew Symington

Andrew Symington

Doctoral Student

Leaving date: 14th July 2013

Themes:

Completed Projects:

Interests

My research focuses on exploiting opportunistic encounters between pairs of sensors to track mobile objects as they move through two-dimensional space. The key idea is that whenever two sensors encounter one another it relates them in space and time, and we can use this information to improve the accuracy of tracking over long periods. I have proposed a novel algorithm, Encounter Based Tracking (EBT) that models the encounters between pairs of sensors, or between sensors and anchors, as a connected graph. The model uses planar graph realization to obtain the relative two-dimensional positions of the vertices from graph edges alone. EBT uses this realized graph to drift-correct a sensor’s estimated trajectory, which may be estimated by inertial, visual or mechanical odometry.

I am currently working on an extension to EBT, in which a map is used to improve the performance of tracking. In some scenarios the encounter graph has an insufficient number of edges to be uniquely realized in two dimensions. In this case the configuration space represents all valid realizations of the graph, given the edges. I believe that it is possible to use graph rigidity theory to efficiently optimize over this configuration space. The goal of the proposed algorithm, EBT-MAP, will be to identify a candidate realization in this configuration space, that agrees with the free space in the given map.

Biography

MSc (Computer Science), University of Cape Town, 2009

BBusSci (Computer Science) Hons., University of Cape Town, 2006

Selected Publications

View AllManage publications

Activities

Completed Projects

Supervisor