From GICL Wiki
Revision as of 00:14, 24 September 2007 by Pwt23 (Talk | contribs)

(diff) ← Older revision | Latest revision (diff) | Newer revision → (diff)
Jump to: navigation, search


Assignment 3: Path Planning

We will not be using Roombas in this assignment, rather we will simulate a point robot. In this assignment you will write a program to calculate a path through an occupancy map using the Wavefront algorithm, a breadth first search. Once the path has been planned, this turns into a dead reckoning problem, which we have already seen, and since we know how temperamental dead reckoning is, this will introduce a wrinkle. We know with dead reckoning the path your robot takes is not exactly the path you specify, and over the course of the path your robot's error grows. If your robot hits an obstacle, this will throw off the path that you had told your robot to execute and your robot may not finish the course. Before creating the path, your program must first adapt the map so that your robot will not come close to any obstacle or any wall except for the start and end positions. Then you should calculate the possible paths in the map using the Wavefront algorithm, and finally calculate a best path for your robot given the Wavefront map.


  • rows and columns were written backwards, the first number is columns, the second number is rows
  • All parts of the assignment should be done in your code, your code can be whatever language you like
  • The number of rows and columns start from 1, so if rows = 9, then there are 9 rows
  • For lettering your path, rather than A-Z,AA-AZ you can just start over at A again.

Description of Assignment

Your program should take in a text file called path_input.txt which will look like:

Media:Path input.txt

Note that this is an example input, which will be followed in the assignment. The actual input file will be provided below.

The first line tells how many columns (9), the second line tells how many rows (8) the map has. The starting point for the path will have an S, and the ending point will have an E. All obstacles will have a 1, and open cells will have a 0.

We will assume that your robot can move one square at a time in any direction (up, down, left, right, all four diagonals), as long as it does not go off the map or into an occupied cell. We also assume that your robot takes up 1 cell. To provide your robot with a buffer so that it does not run into an obstacle or a wall given the error in its path, you will add extra occupied cells to the map. You should first increase the size of all obstacles by the size of your robot in all directions by adding a 1 in all directions around all obstacles.

You will not turn this step in, but here is a representation so that you can see what the input map would look like after your program expands the obstacles:


Next you will expand all walls by adding a 1 above the lower wall, to the right of the left wall, below the upper wall, and to the left of the right wall, do this for all cells that are not currently occupied and are not the start and end positions (i.e. for any cell that has a 0).

You will not turn this step in, but here is a representation so that you can see what the input map would look like after your program expands the walls:


Now we perform the Wavefront algorithm. This is a breadth first search that starts at the end position, and expands out from there, noting how far every other cell is from the end position. You should replace the E (end position) with a 2, then until your algorithm encounters the S (but will not replace the S), you will expand out in a wave from the end position, setting all adjacent cells to 3, then the adjacent cells of the 3's to 4, etc.

You will turn this step in, so your program should output a file named wavefront_path_map.txt that looks like the following:


Finally you will calculate the shortest path from the S to the 2. Note that this minimal distance path may not be unique. Your path will replace the S with an A, then increment through alphabet (B..Z, then AA, AB, AC..AZ if needed) with one letter per cell to indicate your path. Since some environments don't allow more than one character in a position, you may start your path over at A, i.e. A-Z, then A-Z again.

You will turn this step in, so your program should output a file named wavefront_path_plan.txt that looks like the following:


Note that there is an alternative path with F the cell below the displayed F, and then another choice for G, as all of these paths will have the same length.


You will be given the Media:A3_path_input.txt file (and change the name to path_input.txt), and your code should execute on that input file, producing the two output files.

You should turn in:

  • your code
  • wavefront_path_map.txt
  • wavefront_path_plan.txt

For gradudate students, you must implement the Wavefront algorithm in addition to one other path-planning algorithm that you find (other than a Random Walk). For some examples, see chapter 2 of Steven LaValle's Planning Algorithms book at http://planning.cs.uiuc.edu/. Please provide a short description of what other algorithm you choose and how it works. Graduate students must then turn in:

  • your code
  • wavefront_path_map.txt
  • wavefront_path_plan.txt
  • the output file of your additional algorithm
  • any other support files for your additional algorithm
  • README file, containing a short description of your additional algorithm and how it works, and an explanation of what additional files you are turning in