Robot Navigation: Integrating Perception, Environmental Constraints and Task Execution Within a Probabilistic Framework
Abstract
This paper proposes an integrated approach to robot navigation that incorporates task-related information needs, perceptual capabilities, robot knowledge metrics and spatial characteristics of the environment into the motion planning process. Autonomous robots are modelled as discrete-time dynamic systems that implement optimal or suboptimal control policies in their choice of appropriate control actions. A stochastic lattice model, the Inference Grid, is used to represent spatially distributed information. Various information metrics are defined to measure the extent, accuracy and complexity of the robot's world model, and to quantify the information needs of a task. A dual control architecture allows the robot to servo on the information required to solve a given task, and employs multi-objective optimization methods to plan the robot's perceptual and motor actions in an integrated manner.
Invited Paper
BibTeX
@workshop{Elfes-1995-16212,author = {Alberto Elfes},
title = {Robot Navigation: Integrating Perception, Environmental Constraints and Task Execution Within a Probabilistic Framework},
booktitle = {Proceedings of International Workshop on Reasoning with Uncertainty in Robotics (RUR '95)},
year = {1995},
month = {December},
editor = {L. Dorst, M. van Lambalgen, and F. Voorbraak},
pages = {93 - 130},
publisher = {Springer-Verlag},
address = {Berlin, Germany},
}