feat: alpha-beta prunning

multiAgents.py

@@ -72,12 +72,12 @@ class ReflexAgent(Agent):
         newPos = successorGameState.getPacmanPosition()
         newFood = successorGameState.getFood()
         newGhostStates = successorGameState.getGhostStates()
-        newScaredTimes = [ghostState.scaredTimer for ghostState in newGhostStates]
 
         "*** YOUR CODE HERE ***"
         ans = successorGameState.getScore()
-        AVOID_GHOST_SCORE = 10
+        AVOID_GHOST_SCORE = 25
         EAT_FOOD_SCORE = 20
+        EAT_CAPSULE_SCORE = 70
 
         nearest_ghost_distance = 1e9
         for ghostState in newGhostStates:
@@ -88,12 +88,16 @@ class ReflexAgent(Agent):
         for foodPos in newFood.asList():
             nearest_food_distance = min(nearest_food_distance, util.manhattanDistance(foodPos, newPos)+1)
 
+        nearest_capsule_distance = 1e9
+        for capsulePos in successorGameState.getCapsules():
+            nearest_capsule_distance = min(nearest_capsule_distance, util.manhattanDistance(capsulePos, newPos)+1)
+
         ans -= AVOID_GHOST_SCORE * (1/nearest_ghost_distance)
         ans += EAT_FOOD_SCORE * (1/nearest_food_distance)
-
+        if nearest_capsule_distance != 1e9:
-
+            ans += EAT_CAPSULE_SCORE * (1/nearest_capsule_distance)
-        
+        else:
-
+            ans += 500
 
         return ans
 
@@ -158,7 +162,29 @@ class MinimaxAgent(MultiAgentSearchAgent):
         Returns whether or not the game state is a losing state
         """
         "*** YOUR CODE HERE ***"
-        util.raiseNotDefined()       
+        return self.result(gameState, 0, self.depth)[1]
+
+    def result(self, state, agentIndex, depth):
+        if depth == 0 or state.isLose() or state.isWin():
+            return self.evaluationFunction(state), None
+        elif agentIndex == 0:               # maximize
+            (nextAgentIndex, nextDepth) = (0, depth-1) if agentIndex == (state.getNumAgents()-1) else (agentIndex+1, depth)
+            max_reward, max_action = -1e9, None
+
+            for action in state.getLegalActions(agentIndex):
+                reward = self.result(state.generateSuccessor(agentIndex, action), nextAgentIndex, nextDepth)[0]
+                if reward > max_reward:
+                    max_reward, max_action = reward, action
+            return max_reward, max_action
+        else:                               # minimize
+            (nextAgentIndex, nextDepth) = (0, depth-1) if agentIndex == (state.getNumAgents()-1) else (agentIndex+1, depth)
+            min_reward, min_action = 1e9, None
+
+            for action in state.getLegalActions(agentIndex):
+                reward = self.result(state.generateSuccessor(agentIndex, action), nextAgentIndex, nextDepth)[0]
+                if reward < min_reward:
+                    min_reward, min_action = reward, action
+            return min_reward, min_action
 
 class AlphaBetaAgent(MultiAgentSearchAgent):
     """
@@ -170,7 +196,35 @@ class AlphaBetaAgent(MultiAgentSearchAgent):
           Returns the minimax action using self.depth and self.evaluationFunction
         """
         "*** YOUR CODE HERE ***"
-        util.raiseNotDefined()
+        return self.result(gameState, 0, self.depth, -1e9, 1e9)[1]
+
+    def result(self, state, agentIndex, depth, alpha, beta):
+        if depth == 0 or state.isLose() or state.isWin():
+            return self.evaluationFunction(state), None
+        elif agentIndex == 0:               # maximize
+            (nextAgentIndex, nextDepth) = (0, depth-1) if agentIndex == (state.getNumAgents()-1) else (agentIndex+1, depth)
+            max_reward, max_action = -1e9, None
+
+            for action in state.getLegalActions(agentIndex):
+                reward = self.result(state.generateSuccessor(agentIndex, action), nextAgentIndex, nextDepth, alpha, beta)[0]
+                if reward > max_reward:
+                    max_reward, max_action = reward, action
+                if reward > beta:
+                    return reward, action
+                alpha = max(alpha, max_reward)
+            return max_reward, max_action
+        else:                               # minimize
+            (nextAgentIndex, nextDepth) = (0, depth-1) if agentIndex == (state.getNumAgents()-1) else (agentIndex+1, depth)
+            min_reward, min_action = 1e9, None
+
+            for action in state.getLegalActions(agentIndex):
+                reward = self.result(state.generateSuccessor(agentIndex, action), nextAgentIndex, nextDepth, alpha, beta)[0]
+                if reward < min_reward:
+                    min_reward, min_action = reward, action
+                if reward < alpha:
+                    return reward, action
+                beta = min(beta, min_reward)
+            return min_reward, min_action
 
 class ExpectimaxAgent(MultiAgentSearchAgent):
     """