def _reward(self, gameState, action):
score_before = gameState.data.score
gameState = gameState.generateSuccessor(self.index, self.actions[action])
score_after = gameState.data.score
redFoodEaten = self.numRedFood - len(gameState.getRedFood().asList())
blueFoodEaten = self.numBlueFood - len(gameState.getBlueFood().asList())
if gameState.isOnRedTeam(self.index):
reward = blueFoodEaten - redFoodEaten + CaptureAgent.getScore(self,gameState)*100
else:
reward = redFoodEaten - blueFoodEaten + CaptureAgent.getScore(self,gameState)*100
return reward
def getScore(self, state):
"""
Override the getScore function to include some more Information
"""
score = CaptureAgent.getScore(self, state)
score -= len(self.getCapsules(state))
return score
def observationFunction(self, gameState):
"""
This is where we ended up after our last action.
The simulation should somehow ensure this is called
"""
if not CaptureAgent.getPreviousObservation is None:
reward = gameState.getScore() - CaptureAgent.getScore(
CaptureAgent, CaptureAgent.getPreviousObservation)
self.observeTransition(CaptureAgent.getPreviousObservation,
self.lastAction, gameState, reward)
return gameState
def get_score_factor(self):
score_diff = CaptureAgent.getScore(self.root_agent_object,
self.node_state)
return 10000 * score_diff
