CS 470  Artificial Intelligence   Winter 2006,  Instructor:  Jeffrey Horn

PROJECT 2:  (program 2):  Heuristic Planner (A* Search) in Java
 

Handed out/Assigned:	March 7, 2006
Due:	March 21, 2006

• Inspiration:  Pathfinding in computer games, for example:  (from Bethesda Softwork's Morrowind''s AI):

  "The AStar algorithm is a common, but effective and efficient, path-finding algorithm enhanced in Morrowind with the use of pre-generated grids. "  (Kerry Sergent, AI Programmer, TheElderScrolls:  Morrowind, Bethesda Softworks)

• Overview:
  the goal is to build an A* planner that finds the optimal plan to get from an initial to a goal state for one of three problem domains (you choose!)

• PROBLEM DOMAIN 1	PROBLEM DOMAIN 2	PROBLEM DOMAIN 3
  Pacman-like maze	Outer Space Maneuvering	Path-finding on Open Terrain
  2D, dynamic, with moving opponents	2D, dynamic, with moving opponents and moving goal	3D, static, using heightmaps
  click here for details	click here for details	click here for details

• Make sure you have read chapters 7 and 8 on search, and looked through ch. 9 on heuristic search (by this I mean that you do not have to follow every step of every proof; just understand what is being proven and why...).
• Some excellent web sites on PathFinding and A*:
  • http://www.cove.org/default.aspx?id=1&sid=2&mid=7  (great INTERACTIVE animated PathFinding in 2D):
  • http://www.gameai.com/pathfinding.html
  • Justin Heyes-Jones
  • 1996 GameDeveloper article on Intelligent Path Finding, now hosted on Gamasutra site (REALLY HELPFUL)
  • Amit's list of PathFinding tutorials online (extensive list)
  • Illustrative little applet using 8-puzzle (no source code)
  • http://www.student.nada.kth.se/~f93-maj/pathfinder/4.html#2.2
  • pseudocode for A*  from Justin's old A* tutorial
  • more pseudocode for A*  from Justin's newer A* tutorial

   

• CODE:    You have choices.

   

  1. STRANGERS' CODE:
     • Geert-Jan's A* from www.gameai.com   (looks good, but I have not tried it myself...)
     • Some XML-controlled Java code
     • David Brackeen's article has code in it
     • Here is a paper with Java A* code applied to path finding.

      

  2. OUR CODE

  • Start with our Java search code, because it works!  (It compiles and runs, on the Farmer-Goat-Wolf-Cabbage problem...)
  • Modify it to implement A* search on one of the three problem domains above.
  • Try to stick to this modified interface for State (that way maybe, just maybe, we can interchange each other's search code and problem code [dream on...]):