HW4 : Planning and Neural Network Review

Submission Guidelines

DUE Thursday May 17th IN CLASS

(Total Points 8)

Questions

A. Neural Network (4 points)

We need to Review Neural Networks.

Here is the excell sheet for an AND gate neural network.

Extend it to solve Problem 2 on HW 2.

B. Planning Homework (4 points)

An expendable Martian robot plans a one-way trip to a Martian mountain 20 kilometers from its base camp. (It need not return to base camp.) The robot can travel exactly 10 kilometers on a pellet of fuel. It can also carry a pellet of fuel as freight. Base camp is named B, the mountain is named M, and a staging site (10 kilometers from base camp and from the mountain) is named S. There are three pellets of fuel, P1, P2, and P3, at B, and the robot is initially unfueled.

The robot has four actions:

1. goto(x,y), where x is the initial place (one of B,M,S) and the y is the destination place (one of B,M,or S). The robot must have a pellet of fuel in its fuel hopper in order to execute this action, and it can only go between B and S and between S and M on one pellet of fuel.

2. pickup(u,x), where u is a pellet of fuel deposited at place x. The effect of this action is that the robot is carrying u as freight.

3. putdown(u,x), where u is a pellet of fuel deposited at place x. The robot has to be carrying u in order to execture putdown.

4. refuel(u,x), where u is a pellet of fuel at place x that is put in the robot's fuel hopper. (Assume that it takes no fuel to execture the pickup, putdown, and refuel actions.)

We use the following predicate forms (with obvious intended meanings):

AtRobot(x), where x is one of the places, B, M, or S.

At(r,x), where x is one of the places, B, M, or S, and r is one of the fuel pellets. (We do not consider a pellet that is on baord the robot to be at x even if the robot is at x).

Carrying(u) states that the robot is carrying fuel pellet u (as freight)

Fueled states that the robot has a pellet of fuel (ready to burn) in its fuel hopper.

CanGo(x,y) gives the range of the robot on one pellet of fuel. It states that the robot can go from x to y (if it is fueled).

We want to set up this problem so that a STRIPS-type problem solver can produce a place to get the robot to the mountain.

1. What is the initial state description and the goal condition?

2. What are the operators and their descriptions (preconditions, delete list, add list) needed to solve this problem? To make the problem a bit simpler, you do not need to include predicates like Place(x) or Pellet(x) in your fomulas.

3. Give a solution plan to the preceding problem that a STRIPS system (using the algorithm we discussed in class) might produce. Write you solution plan in the following format:

State 0 (S0) - Literals in the initial state
First Operator (i.e. action)
State 1 (S1) - Literals in the Subsequent state
Second Operator
etc... until a state satisfying the goal is reached.

updated 5/10/2007