Page 392 - Introduction to AI Robotics

11 Localization and Map Making

Chapter Objectives:
Describe the difference between iconic and feature-based localization.

Be able to update an occupancy grid using either Bayesian, Dempster-
Shafer,or HIMMmethods.
Describe the two types of formal exploration strategies.

11.1 Overview

The two remaining questions of navigation are: where am I? and wherehaveI
been? The answers to these questions are generally referred to as localization
and map-making, respectively. Both are closely related, because a robot cannot
create an accurate map if it does not know where it is. Fig. 11.1 shows a
hallway in black in a building. The hallway makes a complete circuit around
the center of the building. The gray shows the hallway as sensed by a mobile
robot. The mobile robot senses, updates the map with the portions of the
hallway that have come into view, then moves, updates, and so on. In this
case, it uses shaft encoders to determine where it has moved to and how to
update the map.
As can been seen from the figure, as well as discussions in Ch. 6, shaft
encoders are notoriously inaccurate. Worse yet, the inaccuracies are highly
dependent on surfaces. For example, the robot’s wheel will slip differently
on carpet than on a polished floor. Developing an error model to estimate the
slippage is often unrealistically difficult. The shaft encoder problem might
appear to be eliminated by new hardware technology, especially GPS and
MEMS (micro electrical-mechanical systems) inertial guidance systems (INS).