Page 168 - Introduction to AI Robotics

4.7 Exercises
Exercise 4.14 151
How does the Reactive Paradigm handle the frame problem and the open world as-
sumption?
Exercise 4.15
An alternative RUNAWAY behavior is to turn 90 (either left or right, depending on
whether its “left handed” or “right handed” robot) rather than 180 .Can this be
represented by a potential field?
Exercise 4.16
Using rules, or if-then statements, is a method for representing and combining pro-
gramming units which are often called behaviors; for example “if OBSTACLE-ON-
LEFT and HEADING-RIGHT, then IGNORE.” Can the layers in subsumption for
hall-following be written as a series of rules? Can the potential fields? Are rules
equivalent to these two methods? Do you think rules are more amenable to good
software engineering practices?
Exercise 4.17
Some researchers consider random wandering as a primitive potential field. Recall
that random wandering causes the robot to periodically swap to a new vector with a
random direction and magnitude. How can a wander field be represented? Does the
array of the field represent a physical area or time? Unlike regular potential fields, the
vector is computed as a function of time, every n minutes, rather than on the robot’s
relationship to something perceivable in the world.
Exercise 4.18 [Programming]
Design and implement potential fields:
a. Construct a potential field to represent a “move through door” behavior from
primitive potential fields. Why won’t a simple attractive field work? ANS: if the
robot is coming from a side, it will graze the door frame because the robot is not a
point, it has width and limited turning radius.
b. What happens if a person is exiting the door as the robot enters? Design an appro-
priate “avoid” potential field, and show the emergent potential field when AVOID
and MOVE-THRU-DOOR are activated at the same time.
c. Simulate this using the Khepera simulator for Unix systems found at:
http://www.k-team.com.
d. Runthis onareal khepera.
Exercise 4.19 [Programming]
Program two versions of a phototropic behavior using the Khepera simulator. Both
versions should use the same motor schema, an attractive field, but different percep-
tual schemas. In one version, the perceptual schema processes light from a single
sensor and the behavior is instantiated 8 times. In the second version, the perceptual
schema processes light from all sensors and returns the brightest. Set up five inter-
esting “worlds” in the simulator with different placements of lights. Compare the
emergent behavior for each world.