coverxit / npuzzle

CS 170: Project 1

N-Puzzle %Problem

Introduction

Using the following three kinds of alogrithms to solve the N-Puzzle problem:

• Uniform Cost Search.
• A* with the Misplaced Tile heuristic.
• A* with the Manhattan Distance heuristic.

The main "driver" funtion should match this pseudo code:

function general-search(problem, QUEUEING-FUNCTION)
	nodes = MAKE-QUEUE(MAKE-NODE(problem.INITIAL-STATE)) 
	loop do
		if EMPTY(nodes) then return "failure" 
			node = REMOVE-FRONT(nodes)
		if problem.GOAL-TEST(node.STATE) succeeds then 
			return node 
		nodes = QUEUEING-FUNCTION(nodes, EXPAND(node, problem.OPERATORS))
	end

Structure

Since the assignment requires the code should be kept as general as possible, we first introduce classes needed for a (abstract) general searcher:

• [Problem<StateT, ExpandCostT>](@ref Problem), @copybrief Problem

  @copydetails Problem

• [PriorityQueue](@ref PriorityQueue), @copybrief PriorityQueue

  @copydetails PriorityQueue

• [GeneralSearcher<StateT, NodeT, ExpandCostT>](@ref GeneralSearcher), @copybrief GeneralSearcher

  @copydetails GeneralSearcher

For a specific problem, which is N-Puzzle problem for this project, we defines use a namespace NPuzzle.

Then we set the fundamental arguments:

• NPuzzle::demonstration, @copybrief NPuzzle::demonstration

  • NPuzzle::matrixDemonstration, @copybrief NPuzzle::matrixDemonstration

• NPuzzle::moveCost, @copybrief NPuzzle::moveCost

In addition, we define some structs and typedefs:

• NPuzzle::NPuzzleState, @copybrief NPuzzle::NPuzzleState
• NPuzzle::NPuzzleNode, @copybrief NPuzzle::NPuzzleNode
• NPuzzle::NPuzzleCostFunction, @copybrief NPuzzle::NPuzzleCostFunction

Having all these definitions, we can safely define the searcher for N-Puzzle problem:

• NPuzzle::NPuzzleProblem, @copybrief NPuzzle::NPuzzleProblem
• NPuzzle::NPuzzleSearcher, @copybrief NPuzzle::NPuzzleSearcher

For the three kinds of searching alogrithms, we have the following corresponding functions:

• NPuzzle::GetUniformHeuristicCost, @copybrief NPuzzle::GetUniformHeuristicCost
• NPuzzle::GetMisplacedTileCount, @copybrief NPuzzle::GetMisplacedTileCount
• NPuzzle::GetManhattanDistance, @copybrief NPuzzle::GetManhattanDistance

Because the assignment requires recording the total amount of nodes expanded and the max length of the search queue, we construct a solver:

• NPuzzle::NPuzzleSolver, @copybrief NPuzzle::NPuzzleSolver

Now we have all the necessary classes, then we can solve the N-Puzzle problem by calling NPuzzle::NPuzzleSolver::solve() with an initialState.