School of Computer Science
The University of Adelaide
Artificial Intelligence
Assignment 1
Semester 1 2022
Due 11:59pm Wednesday 23 March 2022
1 Pathfinding
Pathfinding is the problem of finding a path between two points on a plane. It is a
fundamental task in robotics and AI. Perhaps the most obvious usage of pathfinding is
in computer games, when an object is instructed to move from its current position to a
goal position, while avoiding obstacles (e.g., walls, enemy fire) along the way.
Pathfinding in commercial games is frequently accomplished using search algorithms1
.
We consider a simplified version in this assignment. The following shows a 2D map
drawn using ASCII characters:
1 1 1 1 1 1 4 7 8 X
1 1 1 1 1 1 1 5 8 8
1 1 1 1 1 1 1 4 6 7
1 1 1 1 1 X 1 1 3 6
1 1 1 1 1 X 1 1 1 1
1 1 1 1 1 1 1 1 1 1
6 1 1 1 1 X 1 1 1 1
7 7 1 X X X 1 1 1 1
8 8 1 1 1 1 1 1 1 1
X 8 7 1 1 1 1 1 1 1
Given a start position and an end position on the map, our aim is to find a path from the
start position to the end position. The character ‘X’ denotes an obstacle that cannot be
traversed by a path, while the digits represent the elevation at the respective positions.
Any position is indicated by the coordinates (i, j), where i is the row number (ordered
top to bottom) and j is the column number (ordered left to right). For example, the
1http://theory.stanford.edu/~amitp/GameProgramming/
Semester 1 2022 Page 1 by Tat-Jun Chin
top left position is (1, 1), the bottom right is (10, 10), while the position with elevation
‘3’ is (4, 9). Given start position (1, 1) and end position (10, 10), a possible path is
* * * 1 1 1 4 7 8 X
1 1 * 1 1 1 1 5 8 8
1 1 * * * * * * * 7
1 1 1 1 1 X 1 1 * 6
1 1 1 1 1 X 1 * * 1
1 1 1 1 1 1 1 * 1 1
6 1 1 1 1 X 1 * * *
7 7 1 X X X 1 1 1 *
8 8 1 1 1 1 1 1 1 *
X 8 7 1 1 1 1 1 1 *
Note that we use 4-connectedness for paths, which means any two points on the path
are only connected if they are vertically or horizontally (but not diagonally!) adjacent.
1.1 Problem formulation
Following the lecture notes, we formulate the problem as follows:
• States: Any obstacle-free position (i, j) on the map.
• Initial state: A position (i0, j0) given by the user.
• Actions: Since we consider 4-connectedness, only four actions are available: Up,
down, left and right (your program must expand each node in this order).
Available actions for positions adjacent to the map boundary or obstacles are
reduced accordingly.
• Transition model: Moving left moves the current position to the left, etc.
• Goal test: Check if the current state is the end position (i∗, j∗) given by the user.
• Path cost: Given a map M and a path P = {(i0, j0),(i1, j1), . . . ,(iN , jN )}, the
cost of the path is calculated as
g(P) = X
N
k=1
c(ik−1, jk−1, ik, jk, M),
Semester 1 2022 Page 2 by Tat-Jun Chin
where
c(a, b, c, d, M) = (
1 + M(c, d) − M(a, b) if M(c, d) − M(a, b) > 0
1 otherwise
and M(a, b) is the elevation at position (a, b). In words, the cost of a path is the
sum of the costs between two adjacent points of the path, and the cost between
adjacent points is 1 plus the difference between the elevation of the two points if
we climb “uphill”, or simply 1 if we stay “level” or slide “downhill”.
This means shorter paths which avoid climbing cost less. As an example, the cost
in the path in the previous page is 25. What is the optimal (cheapest) path?
1.2 Your tasks
Solve pathfinding using Breadth-First Search (BFS), Uniform-Cost Search (UCS) and
A* Search. You should base your program on the pseudocode GRAPH-SEARCH in the
lecture slides, and carefully think about the appropriate data structures to use. For A*
Search, you must implement two heuristics:
• Euclidean distance between current position and end position.
• Manhattan distance between current position and end position.
For the map in Page 1 with start position (1, 1) and end position (10, 10), your program
should help you answer these questions:
1. Are the paths returned by the three methods different?
2. What about the optimality of the returned paths?
3. Which method is the most computationally and memory efficient?
4. Do the two heuristics for A* Search provide different solutions?
5. Does checking for repeated states matter in this problem?
Semester 1 2022 Page 3 by Tat-Jun Chin
1.3 Deliverables
Write your pathfinding program in Python 3 in a file called pathfinder.py. Your
program must be able to be run as follows:
$ python pathfinder.py [map] [algorithm] [heuristic]
The inputs/options to the program are as follows.
• [map] specifies the path to map, which is a text file formatted according to this
example (see next page):
10 10
1 1
10 10
1 1 1 1 1 1 4 7 8 X
1 1 1 1 1 1 1 5 8 8
1 1 1 1 1 1 1 4 6 7
1 1 1 1 1 X 1 1 3 6
1 1 1 1 1 X 1 1 1 1
1 1 1 1 1 1 1 1 1 1
6 1 1 1 1 X 1 1 1 1
7 7 1 X X X 1 1 1 1
8 8 1 1 1 1 1 1 1 1
X 8 7 1 1 1 1 1 1 1
The first line indicates the size of the map (rows by columns), while the second
and third line represent the start and end positions respectively. The map data
then follows, where all elevation values are integers from 0 to 9 inclusive.
• [algorithm] specifies the search algorithm to use, with the possible values of bfs,
ucs, and astar.
• [heuristic] specifies the heuristic to use for A* search, with the possible values
of euclidean and manhattan. This input is ignored for BFS and UCS.
Your program must then print to standard output the path returned by the
search algorithm, in the following format:
Semester 1 2022 Page 4 by Tat-Jun Chin
* * * 1 1 1 4 7 8 X
1 1 * 1 1 1 1 5 8 8
1 1 * * * * * * * 7
1 1 1 1 1 X 1 1 * 6
1 1 1 1 1 X 1 * * 1
1 1 1 1 1 1 1 * 1 1
6 1 1 1 1 X 1 * * *
7 7 1 X X X 1 1 1 *
8 8 1 1 1 1 1 1 1 *
X 8 7 1 1 1 1 1 1 *
where the path is indicated by asterisks ‘*’ superimposed on the original map beginning
from the start position and leading to the end position. Do not include extraneous
spaces or other characters in the output.
If the given map or problem does not have a feasible path, your program must print
null
Again, do not include extraneous spaces or other characters in the output.
1.3.1 Python libraries
You are allowed to use NumPy to write your pathfinding program. The marking program
will not be able to run your program to completion if other Python libraries are used.
1.4 Submission
You must submit your program files on Gradescope. Instructions on accessing Gradescope
and submitting assignments are provided at https://help.gradescope.com/
article/5d3ifaeqi4-student-canvas. Please use the course code X3ZJZE to enrol
into the course. For undergraduates, please submit your pathfinding program
(pathfinder.py) to Assignment 1 - Undergraduates. If there are any questions or
issues with Gradescope, please contact Andrew via email at andrew.du@adelaide.edu.au.
1.5 Assessment
I will compile and run your code on several test problems. If it passes all tests you will
get 15% (undergrads) or 12% (postgrads) of the overall course mark.
There will be no further manual inspection/grading of your program to award marks
on the basis of coding style, commenting or “amount of code written.
Semester 1 2022 Page 5 by Tat-Jun Chin
1.6 Using other source code
You may not use other source code for this assignment. You should personally and
carefully implement the search algorithms to fully understand the concept.
1.7 Due date and late submission policy
This assignment is due by 11:59pm Wednesday 23 March 2022. If your submission is
late, the maximum mark you can obtain will be reduced by 25% per day (or part thereof)
past the due date or any extension you are granted.
Continues next page for postgraduate section.
Semester 1 2022 Page 6 by Tat-Jun Chin
2 Pathfinding by direct optimisation
For postgraduate students, completing this section successfully will give you the remaining
3% of the marks.
Here we shall attempt to directly optimise the path instead of step-by-step searching.
We consider the simulated annealing algorithm shown in Algorithm 1. For more
background on simulated annealing, see Section 4.1 of Russell and Norvig (3rd ed.).
Algorithm 1 Simulated annealing for path optimisation
1: input Initial path P
0
, initial temperature Tini, final temperature Tf in, cooling rate
α, segment length d.
2: output Optimised path P.
3: Initialise T ← Tini, P ← P
0
.
4: while T > Tf in do
5: P
h ← rand-local-adjust(P, d).
6: ∆g ← g(P) − g(P
h
)
7: if ∆g > 0 then
8: P ← P
h
9: else
10: With probability e
∆g/T
, P ← P
h
.
11: end if /* Record T and g(P) here for bookkeeping.*/
12: T ← αT
13: end while
14: return P
The algorithm receives as input a feasible (but non-optimal) path joining a start
position and an end position on a map. The core idea is to iteratively perform random
local adjustments to the path, and accept the new path if the adjustments improve the
path cost (defined in Sec. 1.1), or accept it probabilistically if the cost is not improved.
The process is repeated until the annealing temperature T falls below a small value
Tf in given by the user. The temperature reduction is conducted as T = αT , where
0 < α < 1 is the cooling rate (also supplied by the user). See Section 4.1 of Russell and
Norvig (3rd ed.) for more details.
The main body of the algorithm is conceptually simple — the hardest part is the routine
to perform the random adjustments. Fortunately we can rely on the BFS program
written in the previous section. The method is shown in Algorithm 2.
Note that the adjustments cannot make the path infeasible, i.e., any resulting path
still joins the original start position and end position required by the user.
Semester 1 2022 Page 7 by Tat-Jun Chin
Algorithm 2 Make random local adjustment on path
1: input Path P, segment length d.
2: output Adjusted path P
h
.
3: Random pick a point (u, v) on P.
4: Pick as (x, y) the point of d positions away from (u, v) along P towards the end
position. If such a point does not exist, use the end position for (x, y).
5: Find a random path S joining (u, v) and (x, y) using randomised BFS (see text
below).
6: Replace path segment in P between (u, v) and (x, y) with S. Store new path as P
h
.
7: return P
h
.
To perform randomised BFS, only a minor tweak to the original BFS algorithm is
required — the order of actions for expanding each node in the search tree is randomised
every time. For example, while in Sec. 1.1 the order is fixed as UDLR (up, down, left,
right), we randomise this at every instance to be LURD, DLUR, etc. The following
shows randomised adjustments with d = 5, and (u, v) = (8, 1) and (x, y) = (10, 4).
* 1 8 1 1 2 4 7 8 X * 1 8 1 1 2 4 7 8 X * 1 8 1 1 2 4 7 8 X
* 1 1 5 1 5 1 5 8 8 * 1 1 5 1 5 1 5 8 8 * 1 1 5 1 5 1 5 8 8
* 4 2 2 1 6 1 4 6 7 * 4 2 2 1 6 1 4 6 7 * 4 2 2 1 6 1 4 6 7
* 5 1 7 0 3 5 1 1 6 * 5 1 7 0 3 5 1 1 6 * 5 1 7 0 3 5 1 1 6
* 7 8 1 2 6 8 1 5 1 * 7 8 1 2 6 8 1 5 1 * 7 8 1 2 6 8 1 5 1
* 7 4 1 1 4 2 2 4 2 * 7 4 1 1 4 2 2 4 2 * 7 4 1 1 4 2 2 4 2
* 5 1 2 1 2 7 5 1 6 * 5 1 2 1 2 7 5 1 6 * 5 1 2 1 2 7 5 1 6
* 7 1 3 4 2 0 4 2 1 * * 1 3 4 2 0 4 2 1 * 7 1 3 4 2 0 4 2 1
* * 1 1 1 5 1 1 9 1 8 * 1 1 1 5 1 1 9 1 * * * * 1 5 1 1 9 1
X * * * * * * * * * X * * * * * * * * * X 8 7 * * * * * * *
2.1 Your tasks
Implement simulated annealing for path optimisation. As a sanity check, test your
program on the following map with start position (1, 1) and end position (10, 10), with
the initial path given by your (deterministic) BFS method from the previous section.
2 1 8 1 1 2 4 7 8 X
3 1 1 5 1 5 1 5 8 8
4 4 2 2 1 6 1 4 6 7
8 5 1 7 0 3 5 1 1 6
Semester 1 2022 Page 8 by Tat-Jun Chin
3 7 8 1 2 6 8 1 5 1
2 7 4 1 1 4 2 2 4 2
6 5 1 2 1 2 7 5 1 6
7 7 1 3 4 2 0 4 2 1
8 8 1 1 1 5 1 1 9 1
X 8 7 1 3 1 7 1 0 0
Use parameter values Tini = 10, Tf in = 0.001, α = 0.99, and d = 5. Your program
should help you answer the following questions:
1. Does simulated annealing find the optimal path every time?
2. How important is it to be able to accept an inferior path? Investigate by disabling
Step 10 in Algorithm 1.
3. How sensitive is the performance to the parameter settings? Investigate by changing
the values of Tini, Tf in, α and d.
2.2 Deliverables
Write your simulated annealing pathfinder program in Python 3 in a file called sapathfinder.py.
Your program must be able to be run as follows:
$ python sapathfinder.py [map] [init] [tini] [tfin] [alpha] [d]
The test program will assume that you would use the same programming language as in
Sec. 1, and that you have a working program (pathfinder.py) for the tasks in Sec. 1.
The inputs/options to the program are as follows.
• [map] specifies the path to a map, formatted according to Sec. 1.3.
• [init] specifies the path to an initial path, encoded according to the output of
the program in Sec. 1.3.
• [tini] and [tfin] specifies the initial and final temperature respectively.
• [alpha] specifies the cooling rate.
• [d] specifies the segment length for random local path adjustments.
Your program must then print to standard output the optimised path, as well as
the evolution of the temperature and path cost, in the manner of this example:
Semester 1 2022 Page 9 by Tat-Jun Chin
* 1 8 1 1 2 4 7 8 X
* 1 1 5 1 5 1 5 8 8
* 4 2 2 1 6 1 4 6 7
* 5 1 7 0 3 5 1 1 6
* 7 8 1 2 6 8 1 5 1
* 7 4 1 1 4 2 2 4 2
* * * * 1 2 7 5 1 6
7 7 1 * 4 2 0 4 2 1
8 8 1 * 1 5 1 1 9 1
X 8 7 * * * * * * *
T = 10.000000, cost = 38
T = 9.900000, cost = 44
T = 9.801000, cost = 42
.
.
.
T = 5.151371, cost = 40
T = 5.099857, cost = 40
T = 5.048859, cost = 41
.
.
.
T = 0.001014, cost = 23
T = 0.001004, cost = 23
Do not include extraneous spaces or other characters in the output.
Submit your program in the same way as the submission for Sec. 1. For postgraduates,
please submit your pathfinding programs (pathfinder.py and sapathfinder.py)
to Assignment 1 - Postgraduates. The due date, late submission policy and code
reuse policy are also the same as in Sec. 1.
∼∼∼ The End ∼∼∼
Semester 1 2022 Page 10 by Tat-Jun Chin