Study Aid - Pop Quiz 1998

This is a "pop" (unannounced) quiz that I gave to a robotics class in 1998. It is up to you to determine which, if any, of these questions are relevant to the material we are currently studying.


Question 1. Learning in Robotic Systems.

Why is it difficult to learn with a simple success/failure signal when many actions take place before the feedback signal is given?

What is the basic method used by Hougen et al. to overcome this difficulty?


Question 2. Sensing.

Consider the dichotomy of active versus passive sensing. Give one advantage of active sensing over passive sensing.

Give one advantage of passive sensing over active sensing.


Question 3. Computer Vision.

The basic template matching algorithm we discussed was for black and white images (that is, images with each pixel having a value of either 0 or 1). How did that algorithm determine the best match between templates and new images?

How could we extend that pattern matching algorithm to work on gray-scale images? (For example, the gray-scale image might have an intensity value of 0-255 for each pixel.)


Question 4. Control Architectures.

What differences should there be between the control architecture of a robot designed to work in a fully known and predictable environment (such as painting a car in a factory) and a robot designed to be used in a less structured environment (such as driving a car in traffic)?


Question 4. Stereo Vision.

What advantage does stereo vision have over monocular (one camera) vision? How is this advantage realized?

What is the primary difficulty to implementing stereo vision?