Note: This content is accessible to all versions of every browser. However, this browser does not seem to support current Web standards, preventing the display of our site's design details.

  

Decentralized Estimation and Control for Information Gathering Robots: Searching for Targets Lost in a Bayesian World

Back
Abstract:
This seminar addresses the problem of decentralized estimation and control in networks of sensors and autonomous agents involved in information gathering tasks. Despite the increased level of complexity associated with integrating multiple autonomous agents into a network to share information and collaborate, networked robots have the potential to increase efficiency, robustness, and responsiveness in time critical missions. A novel decentralized Bayesian approach to the problem of coordinating teams of heterogeneous sensor platforms tasked with searching for lost targets in a dynamic environment will be presented. This architecture combines both decentralized sensor measurements fusion and decentralized platforms control into a unified framework. In a first time, a General Decentralized Data Fusion (GDDF) algorithm allows each sensor node, via anonymous communication with their network neighbors, to build an equivalent representation of the complete Probability Density Function (PDF) of the targets states. Search control decisions are made based on this information. The same underlying mathematics used for the information fusion process are also used to create a negotiation algorithm referred to as General Decentralized Control Fusion (GDCF) algorithm which allows the decision makers to iteratively converge, on the fly, to a Nash equilibrium corresponding to the team optimal search trajectory for a given time horizon. Communication and coupled utility between decision makers are considered to be the fundamental mechanisms underlying cooperation. Several simulation results of realistic search scenarios will be shown to illustrate these mechanisms and to demonstrate the efficiency of the approach. This sort of decentralized architecture offers increased reactivity, robustness and scalability by avoiding the overheads, bottlenecks and single points of failure associated with centralized structures. The improve efficiency and rate of success should directly translate into a higher rate of success in saving human lives. As well, the application domain is open to a wide range of scientific and industrial applications in environmental monitoring, surveillance, planetary exploration, bushfire fighting, border protection, and more.

Type of Seminar:
Public Seminar
Speaker:
Frédéric Bourgault, Postdoctoral Research Associate
Sibley School of Mechanical and Aerospace Engineering, Cornell University
Date/Time:
Oct 22, 2007   17:15
Location:

ETH Zentrum, Gloriastr. 35, Building ETZ, Room E7
Contact Person:

File Download:

Request a copy of this publication.
Biographical Sketch:
Frédéric Bourgault received a B.Eng. in Mechanical Engineering (Aeronautics option) from École Polytechnique de Montréal, Canada, in 1996, a S.M. degree in Aeronautics and Astronautics from the Massachusetts Institute of Technology, U.S.A., and a Ph.D. from the Australian Centre for Field Robotics, the University of Sydney, Australia. For the past year, he has been a postdoctoral research associate at Cornell University, NY, USA. His research interests are in networked robotics; non-linear dynamics, estimation and control; and scalable adaptive human-robot interactions. The focus has been on Bayesian approaches to decentralized control and information fusion in the areas of search and rescue, exploration, surveillance and tracking.