7 p help
- 1. Building a State Tree • Node associated with a state of the game • has 1 child node for each valid move from the node • score of the node is the best score of its children • Leafno valid moves node • has • leaf score determined by end state 1
- 2. Building a State Tree void buildTree(): root = makeNode( startState, null ) • Node associated with a Node makeNode( GameState s, Node parent ): state of the game node = new Node( s, parent ) • has 1 child node for each moves = findAllMoves( s ) if moves.size == 0 // base case: leaf node valid move from the node • score of the node is the best node.score = getScore( s ) else score of its children foreach move in moves • Leaf node save state s • has no valid moves do the move • leaf score determined by end child = makeNode( s, node ) node.children.add( child ) state if child.score is best so far bestChild = child restore state s node.score = bestChild.score 1
- 3. Using a Move object 2
- 4. Using a Move object void doMove( Move mv ) { CardStack startPile = mv.getStartPile(); CardStack endPile = mv.getEndPile(); Card c = startPile.pop(); endPile.push( c ); } void undoMove( Move mv ) { CardStack toPile = mv.getStartPile(); CardStack fromPile = mv.getEndPile(); if ( fromPile.size() == 0 ) print error msg else toPile.push( fromPile.pop() ); update(); } 2
- 5. Build State tree with Move Alternatively, save differences between states (the move!) • the “inverse” move restores the previous state • much more efficient 3
- 6. Build State tree with Move void buildTree(): root = makeNode( null, new Move( _drawPile, _discards ), 0 ); Node makeNode( Node parent, Move m, int depth): Alternatively, save node = new Node( parent, m, depth ) differences between states moves = findAllMoves( ) (the move!) if moves.size == 0 // base case: // leaf node game ends • the “inverse” move node.score = getScore() restores the previous else state foreach move in moves • much more efficient doMove( move ) do the move child = makeNode( node, move, depth+1 ) node.children.add( child ) if child.score is best so far bestChild = child undoMove( move ) node.score = bestChild.score 3