Description
For this problem, handin a lab report pdf (include name, date, assignment and class number
in pdf) which studies statistics for a Wumpus World exploring agent. You should handin
the report pdf as well as the source code used in the study. The code should conform to the
style requested in the class materials. You will find the Wumpus World Matlab code in the
class code link, in subdir A1; use CS4300 WW2 (with a fixed board layout). In addition,
please turn in a hardcopy of the report in class before the start of class on September 1,
2015. Two-person teams need to have one person do report sections 1,3,5, and the other
do 2,4, and 6, and each their own part of section 7; these sections need to be attributed to
the person doing them. Write a lab report in the format (please do not deviate from this
format!) described in the course materials. Discuss the statistical framework to establish a
confidence interval on the means, and any hypothesis tests.
Part I
Provide a mathematical analysis of the percent of solvable 4×4 Wumpus boards for the
number of pits 0,1 and 14 (using only actions FORWARD, RIGHT, LEFT); note that the
gold and Wumpus should be placed on the board first. The gold, Wumpus and pits can go
in any cell, but no two can be in the same cell. Provide an exact analysis (e.g., for 0 pits, all
but one boards should be solvable). In addition, calculate a statistical answer as follows:
1. For the number of pits, p, ranging from 0 to 14
(a) Generate some number, N, of boards with p pits randomly distributed (also
include the Wumpus and the gold, neither of which can be in a pit cell)
(b) Determine if the board is solvable (use Matlab function: CS4300 Wumpus solvable).
1
2. Using the above data, compute the mean percent solvable boards when choosing
actions from the set A = {F ORW ARD, RIGHT, LEF T}, as well as the variance
and a 95% confidence interval.
Part II
For the Wumpus board given below, determine the likelhood that the agent described below
arrives at the square with the gold.
1. Develop an agent (named CS4300 agent1.m) function that randomly (uniformly) selects actions from FORWARD, RIGHT, LEFT in the Wumpus World. The starting
location for each trial should by x = 1, y = 1 and facing right (toward square [2, 1]).
2. Run 2000 trials and determine the mean and variance of the number of steps the agent
survives and the percentage of times the agent arrives at square [3, 4]. Also give the
95% confidence intervals for these.
The board layout is:
+——–+——–+——–+——-+
| | | | |
4 | | Pit | Gold | |
| | | | |
+——–+——–+——–+——-+
| | | | |
3 | Pit | | | |
| | | | |
Y +——–+——–+——–+——-+
| | | | |
2 | | | Pit | Pit |
| | | | |
+——–+——–+——–+——-+
| | | | |
1 | | | | |
| | | | |
+——–+——–+——–+——-+
1 2 3 4
X
2
Develop the agent function according to the following header:
function action = CS4300_agent1(percept)
% CS4300_agent1 – random agent example
% It randomly either changes direction or moves forward
% On input:
% percept (1×5 Boolean vector): percept values
% (1): Stench
% (2): Pit
% (3): Glitters
% (4): Bumped
% (5): Screamed
% On output:
% action (int): action selected by agent
% FORWARD = 1;
% ROTATE_RIGHT = 2;
% ROTATE_LEFT = 3;
% GRAB = 4; — NOT USED
% SHOOT = 5; — NOT USED
% CLIMB = 6; — NOT USED
% Call:
% a = CS4300_agent1([0,1,0,0,0]);
% Author:
%
% UU
%
%
3 CS 4300
