+#!/usr/bin/python -u
+
+#NOTE: The -u option is required for unbuffered stdin/stdout.
+# If stdin/stdout are buffered, the manager program will not recieve any messages and assume that the agent has timed out.
+
+'''
+ khaos.py - A sample Stratego AI for the UCC Programming Competition 2012
+
+ The name describes the state of this file :S
+
+ Written in python, the slithery language
+
+ author Sam Moore (matches) [SZM]
+ website http://matches.ucc.asn.au/stratego
+ git git.ucc.asn.au/progcomp2012.git
+'''
+
+import os
+
+from basic_python import *
+from path import *
+
+def OppositeDirection(direction):
+ if direction == "UP":
+ return "DOWN"
+ elif direction == "DOWN":
+ return "UP"
+ elif direction == "LEFT":
+ return "RIGHT"
+ elif direction == "RIGHT":
+ return "LEFT"
+ else:
+ assert(False)
+ return "ERROR"
+
+class Hunter(BasicAI):
+ " Python based AI of DEATH "
+ def __init__(self, scoresFilename=None):
+ if scoresFilename == None:
+ scoresFilename = "default.scores"
+ BasicAI.__init__(self)
+
+ scoresFile = open(scoresFilename, "r")
+ self.scoreTable = []
+ for i in scoresFile.readline().strip().split(' '):
+ self.scoreTable.append(float(i))
+ scoresFile.close()
+
+ self.maxdepth = 1
+ self.recursiveConsider = {"allies" : 2, "enemies" : 2}
+ self.paths = {}
+
+
+ def PositionLegal(self, x, y, unit = None):
+ if x >= 0 and x < len(self.board) and y >= 0 and y < len(self.board[x]):
+ if unit == None:
+ return True
+ else:
+ return self.board[x][y] == None or self.board[x][y].colour == oppositeColour(unit.colour)
+ else:
+ return False
+
+ def BestMove(self, maxdepth = 1):
+
+ moveList = []
+
+
+ if maxdepth < self.maxdepth:
+ #sys.stderr.write("Recurse!\n")
+ considerAllies = self.recursiveConsider["allies"]
+ considerEnemies = self.recursiveConsider["enemies"]
+ else:
+ considerAllies = len(self.units)+1
+ considerEnemies = len(self.enemyUnits)+1
+
+ for enemy in self.enemyUnits[0:considerEnemies]:
+ for ally in self.units[0:considerAllies]:
+ moveList.append(self.DesiredMove(ally, enemy))
+
+ for desiredMove in moveList:
+ if desiredMove[0] == "NO_MOVE" or desiredMove[2] == None:
+ desiredMove[1] = -2.0
+
+
+
+
+ if maxdepth > 1:
+ for desiredMove in moveList:
+ if desiredMove[2] == None or desiredMove[1] < 0.0:
+ continue
+ p = move(desiredMove[3].x, desiredMove[3].y, desiredMove[2][0], 1)
+ if self.board[p[0]][p[1]] == None:
+ x = desiredMove[3].x
+ y = desiredMove[3].y
+ result = desiredMove[0] + " OK"
+ self.InterpretResult(result)
+ bestRecurse = self.BestMove(maxdepth-1)
+ if bestRecurse != None:
+ desiredMove[1] += bestRecurse[1]# / float(max(1.0, maxdepth))
+ self.board[desiredMove[3].x][desiredMove[3].y] = None
+ self.board[x][y] = desiredMove[3]
+ desiredMove[3].x = x
+ desiredMove[3].y = y
+
+
+
+
+ if len(moveList) <= 0:
+ return None
+ moveList.sort(key = lambda e : e[1], reverse = True)
+ return moveList[0]
+
+
+ def GetPath(self, ally, enemy):
+ #Attempts to do the minimum required work to reconstruct a path
+ return PathFinder().pathFind((ally.x, ally.y), (enemy.x, enemy.y), self.board)
+ if (ally in self.paths.keys()) == False:
+ self.paths.update({ally : {}})
+ #sys.stderr.write("Update keys are " + str(self.paths.keys()) + "\n")
+ #sys.stderr.write("Keys are " + str(self.paths.keys()) + "\n")
+
+ if (enemy in self.paths[ally].keys()) == False: #No path exists; compute a new one
+ path = PathFinder().pathFind((ally.x, ally.y), (enemy.x, enemy.y), self.board)
+ if path != False:
+ self.paths[ally].update({enemy : [path, (ally.x, ally.y), (enemy.x, enemy.y)]})
+ return path
+
+ oldPath = self.paths[ally][enemy]
+ if oldPath[1][0] != ally.x or oldPath[1][1] != ally.y or oldPath[2][0] != enemy.x or oldPath[2][1] != enemy.y:
+ #The pieces involved have moved. Recompute the path
+ path = PathFinder().pathFind((ally.x, ally.y), (enemy.x, enemy.y), self.board)
+ if path != False:
+ self.paths[ally][enemy] = [path, (ally.x, ally.y), (enemy.x, enemy.y)]
+ return path
+
+ if len(oldPath[0]) > 1:
+ #The pieces involved haven't moved, check to see if the path is blocked
+ p = move(ally.x, ally.y, oldPath[0][0], 1) #Look forward one move
+ if self.PositionLegal(p[0], p[1]) and self.board[p[0]][p[1]] != None: #If the position is blocked...
+ path = PathFinder().pathFind((ally.x, ally.y), (enemy.x, enemy.y), self.board) #Compute new path
+ if path != False:
+ self.paths[ally][enemy] = [path, (ally.x, ally.y), (enemy.x, enemy.y)]
+ return path
+ return False
+
+ def DesiredMove(self, ally, enemy):
+ """ Determine desired move of allied piece, towards or away from enemy, with score value """
+ scaleFactor = 1.0
+ if ally.rank == 'F' or ally.rank == 'B':
+ return ["NO_MOVE", 0, None, ally, enemy]
+
+ actionScores = {"ATTACK" : 0, "RETREAT" : 0}
+ if enemy.rank == '?':
+ for i in range(0, len(ranks)):
+ prob = self.rankProbability(enemy, ranks[i])
+ if prob > 0:
+ desiredAction = self.DesiredAction(ally, ranks[i])
+ actionScores[desiredAction[0]] += prob* (desiredAction[1] / 2.0)
+ if len(enemy.positions) <= 1 and ally.rank != '8':
+ scaleFactor *= (1.0 - float(valuedRank(ally.rank)) / float(valuedRank('1')))**2.0
+ elif len(enemy.positions) > 1 and ally.rank == '8':
+ scaleFactor *= 0.05
+ #elif len(enemy.positions) > 1:
+ # scaleFactor *= (1.0 - float(valuedRank(ally.rank)) / float(valuedRank('1')))**0.25
+ # scaleFactor = max(0.05, scaleFactor)
+ else:
+ desiredAction = self.DesiredAction(ally, enemy.rank)
+ actionScores[desiredAction[0]] += desiredAction[1]
+
+
+ desiredAction = sorted(actionScores.items(), key = lambda e : e[1], reverse = True)[0]
+ direction = None
+ path = self.GetPath(ally, enemy)
+
+
+ if path != False and len(path) > 0:
+ if desiredAction[0] == "RETREAT":
+ #sys.stderr.write("Recommend retreat! "+ally.rank + " from " + enemy.rank+"\n")
+ direction = OppositeDirection(path[0])
+ p = move(ally.x, ally.y, direction, 1)
+ if self.PositionLegal(p[0], p[1], ally) == False:
+ path = None
+ scaleFactor = 0.05 * scaleFactor
+ else:
+ direction = path[0]
+ if desiredAction[1] > 0.0 and path != None:
+ scaleFactor = scaleFactor / float(len(path))
+ return [str(ally.x) + " " + str(ally.y) + " " + direction, desiredAction[1] * scaleFactor, path, ally, enemy]
+
+ #directions = {"RIGHT" : enemy.x - ally.x, "LEFT" : ally.x - enemy.x, "DOWN" : enemy.y - ally.y, "UP" : ally.y - enemy.y}
+ #if desiredAction[0] == "RETREAT":
+ # for key in directions.keys():
+ # directions[key] = -directions[key]
+
+ #while direction == None:
+ # d = sorted(directions.items(), key = lambda e : e[1], reverse = True)
+ # p = move(ally.x, ally.y, d[0][0], 1)
+ # if self.PositionLegal(p[0], p[1]) and (self.board[p[0]][p[1]] == None or self.board[p[0]][p[1]] == enemy):
+ # direction = d[0][0]
+ # scaleFactor *= (1.0 - float(max(d[0][1], 0.0)) / 10.0)**2.0
+ # else:
+ # del directions[d[0][0]]
+ # if len(directions.keys()) <= 0:
+ # break
+
+ if abs(enemy.x - ally.x) >= abs(enemy.y - ally.y):
+ if enemy.x > ally.x:
+ direction = "RIGHT"
+ elif enemy.x < ally.x:
+ direction = "LEFT"
+ else:
+ if enemy.y > ally.y:
+ direction = "DOWN"
+ elif enemy.y < ally.y:
+ direction = "UP"
+ if direction == None:
+ return ["NO_MOVE", 0, [], ally, enemy]
+ return [str(ally.x) + " " + str(ally.y) + " " + direction, desiredAction[1], None, ally, enemy]
+
+
+ def DesiredAction(self, ally, enemyRank):
+ if enemyRank == 'F':
+ return ["ATTACK", 1.0]
+ if ally.rank == '8' and enemyRank == 'B':
+ return ["ATTACK", 0.9]
+ if ally.rank == '1' and enemyRank == 's':
+ return ["RETREAT", 0.9]
+ if ally.rank == 's' and enemyRank == '1':
+ return ["ATTACK", 0.6]
+ if enemyRank == 'B':
+ return ["RETREAT", 0.0]
+ if ally.rank == enemyRank:
+ return ["ATTACK", 0.1]
+ if valuedRank(ally.rank) > valuedRank(enemyRank):
+ return ["ATTACK", float(self.scoreTable[ranks.index(enemyRank)]) * (0.1 + 1.0/float(self.scoreTable[ranks.index(ally.rank)]))]
+ else:
+ return ["RETREAT", float(self.scoreTable[ranks.index(ally.rank)]) / 10.0]
+
+
+ def MakeMove(self):
+ if len(self.units) < 20:
+ self.maxdepth = 1
+ bestMove = self.BestMove(self.maxdepth)
+
+
+ if bestMove == None:
+ #sys.stderr.write("Khaos makes random move!\n")
+ return BasicAI.MakeMove(self)
+
+ #sys.stderr.write("Board state before move: \n")
+ #self.debugPrintBoard()
+
+ #sys.stderr.write("Best move is \"" + bestMove[0] + "\" with score " + str(bestMove[1]) + " as part of path " +str(bestMove[2]) + " ...\n")
+ #sys.stderr.write(" Ally with rank " + bestMove[3].rank + " is targeting unit at " + str((bestMove[4].x, bestMove[4].y)) + " rank " + bestMove[4].rank + "\n")
+
+ sys.stdout.write(bestMove[0] + "\n")
+ #self.paths[bestMove[3]][bestMove[4]].pop(0)
+
+ return True
+
+
+
+ def rankProbability(self, target, targetRank):
+
+ if targetRank == '+' or targetRank == '?':
+ return 0.0
+ if target.rank == targetRank:
+ return 1.0
+ elif target.rank != '?':
+ return 0.0
+
+ total = 0.0
+ for rank in ranks:
+ if rank == '+' or rank == '?':
+ continue
+ elif rank == 'F' or rank == 'B':
+ if target.lastMoved < 0:
+ total += self.hiddenEnemies[rank]
+ else:
+ total += self.hiddenEnemies[rank]
+
+ if total == 0.0:
+ return 0.0
+ return float(float(self.hiddenEnemies[targetRank]) / float(total))
+
+ def InterpretResult(self, string=None):
+ if BasicAI.InterpretResult(self, string) == False:
+ return False
+
+
+ if self.maxdepth > 1:
+ if self.lastMoved != None and self.lastMoved.colour == self.colour and self.lastMoved.alive == False:
+ self.units.sort(key = lambda e : valuedRank(e.rank), reverse = True)
+ elif self.lastMoved != None and self.lastMoved.colour == oppositeColour(self.colour) and self.lastMoved.alive == True:
+ oldRank = self.lastMoved.rank
+ self.lastMoved.rank = '1'
+ self.enemyUnits.sort(key = lambda e : valuedRank(e.rank), reverse = True)
+ self.lastMoved.rank = oldRank
+
+
+ return True
+
+
+if __name__ == "__main__":
+ if len(sys.argv) > 1:
+ hunter = Hunter(sys.argv[1])
+ else:
+ string = ""
+ path = sys.argv[0].split('/')
+ for i in range(0, len(path)-1):
+ string += path[i] + "/"
+ string += "default.scores"
+
+
+ hunter = Hunter(string)
+ if hunter.Setup():
+ while hunter.MoveCycle():
+ pass
+
git.ucc.asn.au / progcomp2012.git / blobdiff
UCC git Repository :: git.ucc.asn.au