Home > Seminars > On Vehicle Localization: From Geometry to Topology

On Vehicle Localization: From Geometry to Topology

Start:

8/4/2014 at 3:00PM

End:

8/4/2014 at 4:00PM

Location:

117I Cushing Hall

Host:

College of Engineering close button
headerbottom

Vijay Gupta

Vijay Gupta

VIEW FULL PROFILE Email: vgupta2@nd.edu
Phone: 574-631-2294
Website: http://ee.nd.edu/faculty/vgupta/
Office: 270 Fitzpatrick Hall

Affiliations

Department of Electrical Engineering Professor and Associate Chair of Graduate Studies
Research Interests: Dr. Gupta's current research interests are in the analysis and design of cyberphysical systems. Such systems are the next generation of engineering systems and involve tightly coupled control, communication, and processing algorithms. Applications include structural health ...
Click for more information about Vijay
574-631-2294
Add to calendar:
iCal vCal
In this talk we will describe methodologies to localize both a single and a team of vehicles navigating in a complex environment without GPS. During the first part of the talk, we will consider the situation when vehicles (or a single vehicle navigating in an environment with multiple beacons) can measure their relative (inter-vehicle) distances. In this case, the problem can be posed as a distributed graph embedding problem. The convergence speed of this type of algorithms strongly depends on the spectral property of the underlying network (graph), and we will discuss a distributed clustering algorithm that can accelerate the distributed embedding problem for slowly varying networks. Next, we will consider situations where vehicle teams can only coarsely compare relative distance measurements. We will show that the problem can be posed, once again, as a graph embedding problem and demonstrate how this information can be fused with inertial sensors. Finally, we discuss the limiting case when a vehicle navigating in an environment, with multiple beacons, can coarsely localize itself by leveraging visibility (binary detection) information as well as a low-grade inertial sensor. Key to this final result is leveraging local homology computations to enable successful loop closure in a probabilistic filter. Initial experimental results for this last scenario will be presented.

Seminar Speaker:

Alberto Speranzon

Alberto Speranzon

United Technologies Research Center

Alberto Speranzon received the ‘‘Laurea’’ degree in computer engineering from University of Padova, Italy in 2000, the Tech. Lic. and Ph.D. in automatic control from the School of Electrical Engineering, Royal Institute of Technology, Stockholm, Sweden in 2004 and 2006, respectively. Since 2008 he is a Research Scientist at United Technologies Research Center (UTRC), in East Harford, CT, USA. At UTRC, Alberto served as project manager and principal investigator for DARPA's ASPN program on novel methods for distributed localization of multiple vehicles in GPS degraded/denied environments combining adaptive filtering, graph theoretical and algebraic topological methods. His research interests are mainly in the area of distributed control, estimation and optimization, with particular focus on multi-vehicle systems and wireless sensor networks. Alberto’s received the Outstanding Achievement Award in 2009, highest award given by UTRC. Before joining UTRC, between October 2006 and September 2008, he was a Marie Curie Research Fellow at Unilever R&D, Port Sunlight, UK. During 2006-2008 he was a regular visitor at University of California at Berkeley, USA working on decentralized estimation over sensor networks.