Quarterly Progress Report on Contract N00014-91-J-1577 (Yale University) .

Prof. Drew McDermott and Sean Engelson focused on developing a heuristic model of place perception for mobile robots. The model differs from past approaches in its emphasis on error correction. As a network of places is developed, the system keeps statistics that tell it when to merge two locations...

Full description

Bibliographic Details
Main Author: McDermott, Drew
Other Authors: YALE UNIV NEW HAVEN CT DEPT OF COMPUTER SCIENCE
Format: Text
Language:English
Published: 1991
Subjects:
Cybernetics
*PERCEPTION
*DECISION MAKING
*ROBOTS
NETWORKS
LAYERS
MOTION
MOBILE
ERRORS
HEURISTIC METHODS
MAPS
LEARNING
NONLINEAR ALGEBRAIC EQUATIONS
MODELS
CORRECTIONS
ALGORITHMS
POSITION(LOCATION)
RESEARCH MANAGEMENT
UNIVERSITIES.
Hager
sami
Online Access:http://www.dtic.mil/docs/citations/ADA240995
http://oai.dtic.mil/oai/oai?&verb=getRecord&metadataPrefix=html&identifier=ADA240995
Description
Summary:Prof. Drew McDermott and Sean Engelson focused on developing a heuristic model of place perception for mobile robots. The model differs from past approaches in its emphasis on error correction. As a network of places is developed, the system keeps statistics that tell it when to merge two locations previously thought distinct, and when to split one supposed location into two. Prof. Greg Hager has been working together with Sami Atiya of Karlsruhe University to extend their work in robot localization. The latest version of their algorithm includes techniques for models of robot motion and map learning. IEEE Journal of Robotics and Automation. Hager's work in interval-based model recovery and decision making has been reimplemented in C++. The new implementation includes all of the results from his previous work, and includes several extensions. In particular, the new system can solve general systems of nonlinear equations in a very natural and clean fashion. Prof. McDermott also spent time designing and implementing the interface between the reactive interpreter and plan layer for the XFRM planner.

Similar Items