Toward task oriented localization

Abstract

In the course of building a fully autonomous robot platform it is important to look at the computational resources spent by the indi- vidual modules. Each of them cannot be greedy, or the overall demand for computational power will be beyond what can be handled on-board. Maintaining an estimate of the pose of a mobile robot is a typical exam- ple where we might not always need to run the algorithm at the highest possible rate. This paper deals with the problem of determining how much e ort is needed in order to accomplish the localization part of a task. The approach we have taken to the problem is to optimize a cost function that accounts for the cost of sensing and the growth of the uncertainty.