CS 470 Artificial Intelligence

Winter 2006, Instructor: Jeffrey Horn NAME: _______________________

QUIZ 5: Stimulus-Response (S-R) versus Finite State Machines

Handed out/Assigned: Tuesday, April 25, 2006

Due: Thursday, April 27, 2006 in class!

VEHICLE:

We have two binary outputs, one for each motor. LM is left motor, RM is right. A "0" output means no power to the motor, while a "1" means go forward. (so an output of 01, which is 0 to the left and 1 to the right motor, would result in an arcing turn to the left, while 10 would be a right turn)

Let there be two input bits from the left and right sensors pictured (LS and RS): a "01" input vector (left sensor, right sensor) would mean something detected to the right, but nothing center or left, while "11" would mean something in the center (or something left and right!). Also assume a finite detector range, defined by an arc (not shown above) with its center at the robot's center.

Let there be a third sensor, "GO", (not shown) which indicates whether the object detected (if any!) is a Goal or an Obstacle (hence the name GO). Let us say you are seeking light sources, but avoiding obstacles, and you have these two combo detectors, LS and RS, that can detect both obstacles and light sources (if a detector detects both, it will register only the closest, so note that ONLY ONE OBJECT WILL BE DETECTED AT ANY ONE TIME!). The detectors will use the GO bit to indicate whether the nearest detected object is a light source (GO == "1") or an obstacle (GO == ")"). Note that if no obstacle is detected, then the value of GO is meaningless.

(1) S-R: Fill in a truth table for a strategy to move TOWARD goals (e.g., light sources), while avoiding obstacles:

INPUTS			OUTPUTS
GO	LS	RS	LM	RM
0	0	0
0	0	1
0	1	0
0	1	1
1	0	0
1	0	1
1	1	0
1	1	1

FINITE STATE MACHINE: Now let's see if we can improve on the S-R behavior above by increasing the computational power of its brain. Let us add memory, two bits worth: M0 and M1. You can think of these as internal sensors that detect the state of two "internal environment" sensors. Furthermore, you can change the state of these two bits each time step. Thus the states of the variables M0 and M1 in the INPUTS column below are read at time t, then you specify the outputs M0 and M1 to be read in at time t+1 (i.e., the next time the truth table is read).

(2) FSM: Fill in a truth table for the same goal-seeking, obstacle-avoiding strategy as in (1) above, but now trying to take advantage of the "two-bit" memory you have!

INPUTS						OUTPUTS
Num₁₀	M0	M1	GO	LS	RS	LM	RM	M0	M1
0	0	0	0	0	0
1	0	0	0	0	1
2	0	0	0	1	0
3	0	0	0	1	1
4	0	0	1	0	0
5	0	0	1	0	1
6	0	0	1	1	0
7	0	0	1	1	1
8	0	1	0	0	0
9	0	1	0	0	1
10	0	1	0	1	0
11	0	1	0	1	1
12	0	1	1	0	0
13	0	1	1	0	1
14	0	1	1	1	0
15	0	1	1	1	1
16	1	0	0	0	0
17	1	0	0	0	1
18	1	0	0	1	0
19	1	0	0	1	1
20	1	0	1	0	0
21	1	0	1	0	1
22	1	0	1	1	0
23	1	0	1	1	1
24	1	1	0	0	0
25	1	1	0	0	1
26	1	1	0	1	0
27	1	1	0	1	1
28	1	1	1	0	0
29	1	1	1	0	1
30	1	1	1	1	0
31	1	1	1	1	1

(3) EVOLUTION: The goal here is to let evolution fill in the truth table in (2) above. We are going to do this in our 3DGS simulator (yay!). Help design this research experiment:

(3a.) Let's assume a bit-string encoding of the output bits above. What is the size of the search space?

(3b.) Let's assume we want to evolve a behavior that seeks out the player's avatar (the goal for the GOA sensor) while avoiding obstacles. Describe a fitness function that you think might get us there: