Wednesday, April 12, 2000
4:15 PM (refreshments 4:00)
Room NE43-518
EECS Special Seminar
Abstract
Probabilistic methods are well suited for dealing with the uncertainties involved in sensing and acting in the real world. In this talk I will focus on the application of probabilistic methods to fundamental problems in mobile robotics. The first problem is estimating the position of a mobile robot within its environment. I will present Monte-Carlo localization, a sample-based variant of a general framework for mobile robot localization. An extension of Monte-Carlo localization allows to transfer information between different robots, thereby amortizing high-cost sensors across multiple robot platforms. Another problem I will discuss in this talk is the task of building a map of a robot's environment. This problem is more complex than the first one since it involves the concurrent estimation of a model of the robot's environment and its position therein. Finally, I will present applications of probabilistic methods to mobile robot control. The techniques described in this talk have been implemented and tested in several real-world applications of mobile robots, including the deployments of two mobile robots as interactive museum tour-guides.
|
Modified: Mar 1, 2000|
Current events|
Your comments
and inquiries are welcome.