def betterEvaluationFunction(currentGameState: GameState):
"""
Your extreme ghost-hunting, pellet-nabbing, food-gobbling, unstoppable
evaluation function (question 5).
DESCRIPTION: <write something here so we know what you did>
"""
"*** YOUR CODE HERE ***"
pos = currentGameState.getPacmanPosition()
foods = currentGameState.getFood()
walls = currentGameState.getWalls()
capsules = currentGameState.getCapsules()
ghosts = currentGameState.getGhostStates()
score = currentGameState.getScore()
foods_cost = _food_cost(foods.asList(), pos, walls)
capsules_cost = _food_cost(capsules, pos, walls)
ghosts_dis, s_ghosts_dis = _ghost_cost(ghosts, pos, walls)
def d(x):
if x == 0:
return float('inf')
return 9 / (x**2)
ghosts_cost = sum(map(d, ghosts_dis))
s_ghosts_cost = sum(map(lambda x: x[0], filter(lambda x: x[0] < x[1], s_ghosts_dis)))
return score - (2 * foods_cost) - capsules_cost - s_ghosts_cost - ghosts_cost
def betterEvaluationFunction(currentGameState: GameState) -> float:
"""
Your extreme, unstoppable evaluation function (problem 4). Note that you can't fix a seed in this function.
"""
# BEGIN_YOUR_CODE (our solution is 13 lines of code, but don't worry if you deviate from this)
def getDistFromPacman(x, y):
pacmanPos = currentGameState.getPacmanPosition()
return abs(pacmanPos[0] - x) + abs(pacmanPos[1] - y)
def exponentiallyWeightedScore(objectives):
return sum(2**(-1 *
getDistFromPacman(objectivePos[0], objectivePos[1]))
for objectivePos in objectives)
def getFoodScore(foodGrid, pacmanPosition):
foodsPos = [(x, y) for y in range(foodGrid.height)
for x in range(foodGrid.width) if foodGrid[x][y] == True]
return exponentiallyWeightedScore(foodsPos)
food = 9 * getFoodScore(currentGameState.getFood(),
currentGameState.getPacmanPosition())
ghostStates = currentGameState.getGhostStates()
scary = sum(ghostState.scaredTimer for ghostState in ghostStates) > 0
capsule = 100 if scary else 200 * exponentiallyWeightedScore(
currentGameState.getCapsules())
scaredGhostsPos = [
ghostState.getPosition() for ghostState in ghostStates
if ghostState.scaredTimer > 0
]
ghost = 200 * exponentiallyWeightedScore(scaredGhostsPos)
score = currentGameState.getScore()
return food + capsule + ghost + score
def betterEvaluationFunction(currentGameState:GameState):
"""
Your extreme ghost-hunting, pellet-nabbing, food-gobbling, unstoppable
evaluation function (question 5).
DESCRIPTION: <write something here so we know what you did>
"""
features = [0.]*8
weights = [20000,10,1,1,45,0.5,3,1]
#1/number_food,1/cloest_food_dis,
#1/num_pallets, 1/cloest_pallet, if_empty_pallet(encourage to eat the last pallet), scared_time(encourage the pacman to eat pallet to increase scared_time)
#encourage to eat the ghost, some scared time becomes 0, some are not
#1/distance_to_ghost
#weights = []
if currentGameState.isWin():
#print(currentGameState.getFood().asList())
return float("inf")
if currentGameState.isLose():
return float("-inf")
pac_pos = currentGameState.getPacmanPosition()
food_pos = currentGameState.getFood().asList()
num_food = len(food_pos)
ghost_states = currentGameState.getGhostStates()
ghost_scared_times = [ghostState.scaredTimer for ghostState in ghost_states]
ghost_pos = [g.getPosition() for g in ghost_states]
pallets = currentGameState.getCapsules()
features[0] = 1/num_food
food_dis = sorted([util.manhattanDistance(fp,pac_pos) for fp in food_pos])
features[1] = 1/food_dis[0]
pallets_dis = sorted([util.manhattanDistance(p,pac_pos) for p in pallets])
if pallets_dis:
features[2] = 1/len(pallets_dis)
features[3] = 1/pallets_dis[0]
features[4] = int(not pallets) # if pallets become zero, we need to encourage this
features[5] = sum(ghost_scared_times)
features[6] = 1 if 0 in ghost_scared_times and sum(ghost_scared_times)>0 else 0
# if some ghost's scared time becomes zero, some are not, then some ghost is eaten
# for those states after the ghost is eaten, this will always be 1, but does not afftect the action selection, because we compare the relative relation
ghost_dis = [util.manhattanDistance(g, pac_pos) for g in ghost_pos]
minDis = min(ghost_dis)
min_dis_ghost_inds = [i for i in range(len(ghost_dis)) if ghost_dis[i] == minDis]
min_dis_ghost_scared_times = [ghost_scared_times[i] for i in min_dis_ghost_inds]
if any(st < minDis for st in min_dis_ghost_scared_times):
features[7] = -1.0 / minDis # if any closet ghost has no enough scared time, then the agent is in danger
else:
features[7] = 1.0 / minDis # otherwise, there is no danger
return sum(w*f for w,f in zip(weights,features))
