EECS 391/491: PROBLEM SET #2

Reading:

Problem 2.1 [10 points]

Problem 2.5 of R&N (c) and (d) only. Put your answers in a grid as in Figure 2.5 (page 40).

Problem 2.2 [10 points]

Problem 2.6 of R&N for (c) and (d) only. Put your answers to the first part in a grid as in Figure 2.6 (page 43).

Problem 2.3 [10 points]

The discrete photovore we talked about in class had a one-dimensional percept with two values: {NoLight, Light}. Thus, using the notation in the middle of page 45 of R&N it has |P|=2 (that is, the total number of percepts is two). It also had two actions A={Stay, Move}, so |A|=2. Thus, there are a total of 22 stimulus/response (S/R) agent programs (also known as look-up tables or LUTs): 
NoLight| Stay
Light  | Stay

NoLight| Stay
Light  | Move

NoLight| Move
Light  | Stay

NoLight| Move
Light  | Move

How many S/R agents are there in general? Write your answer in terms of the number of percepts and actions, |P| and |A|. Hint: You may wish to test your answer in special cases, such as |A|=1 or |A|=2 while |P|=n.

Use you answer to compute the number of S/R agents for the vacuum-cleaner agent pictured in Figure 2.2 (and described on page 33) of R&N.

Problem 2.4 [10 points]

Draw a state machine that captures the behavior of the sphex wasp described on page 37 of R&N.

Problem 2.5 [10 points]

Design (the stimulus response curves and connections for) a vehicle that circles a source at a fixed distance, as does the middle vehicle in Braitenberg's Figure 7 (page 18).

Problem 2.6 [10 points]

Recall the equations we wrote in class for Figure 10b on page 23 of Braitenberg's book. Do the same for the circuit in Figure 10a, calling the input s, and the outputs of the successive threshold units (left to right) v, w, x, y, and z.

You may do this by hand or write a computer program to figure these out. Your choice. Whatever choice, make sure the answer is organized in a well-labeled table as we did in class.

Problem 2.7 [10 points]

Draw the memory circuit described for Vehicle 5 (at the top of page 24). Draw one that will remember if it has seen more than one red light. Exactly 2 red lights.

Note: we will use the discrete-time model of Problem 2.6 above. Thus, "2 lights" means a red light was seen at two different integer times (which might or might not be contiguous). Hint: "2 or more" is the same "see one now AND saw one in my history".

Problem 2.8 [20 points]

Problem 2.8 of R&N. Note: this requires building the simulator of Problem 2.7 of R&N. You may write it yourself or use the code mentioned. If you write your own, you do not have to make a modular implementation in terms of size and shape, only initial configurations of dirt and robot. You only have to simulate the specific size and shape of the environment depicted in Figure 2.2.

There is no need to turn in any code. Just turn in a short explanation of how your program is structured. Cite any repository code used and note significant modifications.

You must report the S/R table used and the performance score for each configuration and its overall average score.

Problem 2.9 [10 points]

Problem 2.9 of R&N. Note: there is no need to simulate/code for this problem.

Note: This problem--and all GX.Y problem--are for 491 students only.

Problem G2.1 [20 points]

State of the Art AI Agents

This answer must be typed and on a separarate sheet from the above.

Pick a current, state-of-the-art agent (a DARPA Urban Challenge Vehicle, iRobot's Roomba, etc.) for which you can find materials regarding its agent problem and its agent architecture.

Created: 2008-01-25. Modified: 2008-01-25.