RECEDING HORIZON CONTROL FOR SELECTION OF FOCUS OF ATTENTION

M. Lauri, R. Ritala

Abstract

We consider the problem how a mobile robot equipped with a sensory subsystem with multiple operation modes should focus its attention to provide optimal support for finding the shortest path to a given target. The path planning problem is considered as a search task in a directed graph. The path planning is linked to the focus of attention problem by noting that any observations made of the environment ultimately reveal us information on the arc weights in the graph, i.e. the cost function. The formulation allows us to describe consistently various phenomena affecting optimal paths in the environment, for example moving obstacles that the robot may not collide with. We give a formulation for a cost function that allows trade-off between exploration of the environment and exploitation of current information by a scaling parameter.

Keywords

focus of attention, partially observable Markov decision process, receding horizon control, dynamic programming

Download

IMEKO-WC-2012-TC7-P8.pdf

DOI

-

IMEKO TC

TC7 - Measurement Science

Back to the proceedings