Page 448 - Introduction to AI Robotics

431
11.10 Exercises
sor readings. The features can be either local features such as a corner, or
more topologically significant, such as a landmark or gateway. The choice
of features gives rise to two problems: the certainty with which a feature
can be detected (how likely is the robot to miss a hallway on its left) and
reasoning about indistinguishable locations (one hallway intersection looks
like another). Topological localization can be done by matching the gate-
ways encountered with the expected gateways given on a topological map.
All existing map-making methods tend to break in the presence of changing
environments.
Formal exploration strategies for mapping new areas fall into two cate-
gories of techniques: frontier-based and GVG. Frontier-based methods look as
the perimeter of the of sensed region on an occupancy grid and then rank
areas to explore. Voronoi methods represent the explored world as a Voronoi
diagram and use geometric properties to ensure that open areas are oppor-
tunistically explored as the robot moves.

11.10 Exercises

Exercise 11.1
An important property of an occupancy grid is that it supports sensor fusion.Define
sensor fusion in one or two sentences and give two examples.

Exercise 11.2
What is the difference between an occupancy grid, a certainty grid, and an evidence
grid?
Exercise 11.3
What is the difference between iconic and feature-based localization?
Exercise 11.4
List the desirable attributes of a landmark. Discuss why or why not each of the fol-
lowing might make a good landmark for feature-based localization:

a. Corner
b. Intersection of hallways
c. Open doorway

Exercise 11.5
Name three applications where metric map-making might be useful.

Exercise 11.6
Why is localization important in map-making?