Files
PPCA-AIPacMan-2024/reinforcement/deepQLearningAgents.py
2024-07-22 19:10:40 +08:00

357 lines
15 KiB
Python

import model
from qlearningAgents import PacmanQAgent
from backend import ReplayMemory
import layout
import copy
import torch
import numpy as np
import os
from util import *
import util
from pacman import GameState
from game import Directions, Actions
class Agent:
"""
An agent must define a getAction method, but may also define the
following methods which will be called if they exist:
def registerInitialState(self, state): # inspects the starting state
"""
def __init__(self, index=0):
self.index = index
def getAction(self, state):
"""
The Agent will receive a GameState (from either {pacman, capture, sonar}.py) and
must return an action from Directions.{North, South, East, West, Stop}
"""
raiseNotDefined()
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>
"""
# Useful information you can extract from a GameState (pacman.py)
if currentGameState.isLose():
return -500
if currentGameState.isWin():
return 500
kInf=1e100
capsules_position = currentGameState.getCapsules()
current_pos = currentGameState.getPacmanPosition()
food_positions = currentGameState.getFood().asList()
# print(f"action:{action}, action_vec:{action_vec}")
current_ghost_positions = currentGameState.getGhostPositions()
current_ghost_scared_times = [ghostState.scaredTimer for ghostState in currentGameState.getGhostStates()]
not_scared_ghosts_positions = [current_ghost_positions[i] for i in range(len(current_ghost_positions)) if current_ghost_scared_times[i]==0]
scared_ghosts_positions = [current_ghost_positions[i] for i in range(len(current_ghost_positions)) if current_ghost_scared_times[i]>0 and current_ghost_scared_times[i]<=1.2*util.manhattanDistance(current_ghost_positions[i],current_pos)+2]
edible_ghosts_positions = [current_ghost_positions[i] for i in range(len(current_ghost_positions)) if current_ghost_scared_times[i]>1.2*util.manhattanDistance(current_ghost_positions[i],current_pos)+2]
current_self_position = currentGameState.getPacmanPosition()
def DotProduct(a,b):
return a[0]*b[0]+a[1]*b[1]
def CrossProduct(a,b):
return a[0]*b[1]-a[1]*b[0]
def EuclideanDistance(a,b):
return ((a[0]-b[0])**2+(a[1]-b[1])**2)**0.5
def DistanceAnalysis(current_self_position,object_postion_list,flag="None"):
if len(object_postion_list)==0:
return 0
if current_self_position in object_postion_list:
return kInf
res=0
for obj_pos in object_postion_list:
if flag=="Ghost" and util.manhattanDistance(current_self_position,obj_pos)>=6:
continue
distance_to_obj=util.manhattanDistance(current_self_position,obj_pos)
res=max(res,1/distance_to_obj)
return res
da_for_foods=DistanceAnalysis(current_self_position,food_positions)
da_for_unscared_ghosts=DistanceAnalysis(current_self_position,not_scared_ghosts_positions,"Ghost")
da_for_scared_ghosts=DistanceAnalysis(current_self_position,scared_ghosts_positions)
da_for_capsules=DistanceAnalysis(current_self_position,capsules_position)
da_for_edible_ghosts=DistanceAnalysis(current_self_position,edible_ghosts_positions)
res=da_for_capsules*2-da_for_unscared_ghosts*2-da_for_scared_ghosts*0.2+da_for_foods*0.2+da_for_edible_ghosts*1
if da_for_unscared_ghosts<1/6:
res+=(da_for_foods*0.2+da_for_edible_ghosts*1)*5
# res*=random.uniform(0.9, 1.1)
res*=100
global last_score
res+=(currentGameState.getScore()-last_score)*10
# print(f"res:{res}")
return res
class MultiAgentSearchAgent(Agent):
"""
This class provides some common elements to all of your
multi-agent searchers. Any methods defined here will be available
to the MinimaxPacmanAgent, AlphaBetaPacmanAgent & ExpectimaxPacmanAgent.
You *do not* need to make any changes here, but you can if you want to
add functionality to all your adversarial search agents. Please do not
remove anything, however.
Note: this is an abstract class: one that should not be instantiated. It's
only partially specified, and designed to be extended. Agent (game.py)
is another abstract class.
"""
def __init__(self, evalFn = 'scoreEvaluationFunction', depth = '2'):
self.index = 0 # Pacman is always agent index 0
self.evaluationFunction = betterEvaluationFunction
self.depth = int(depth)
last_score=0
class ExpectimaxAgent(MultiAgentSearchAgent):
"""
Your expectimax agent (question 4)
"""
def ExpectMaxSearch(self,gameState: GameState,depth_remain:int,agentIndex:int) -> tuple[int, list[Actions]]:
if depth_remain==0:
# print(f"depth_remain:{depth_remain}")
# print(f"returning leaf {self.evaluationFunction(gameState)}, {[]}")
return self.evaluationFunction(gameState),[]
legal_actions = gameState.getLegalActions(agentIndex)
if len(legal_actions)==0:
# print(f"depth_remain:{depth_remain}")
# print(f"returning leaf {self.evaluationFunction(gameState)}, {[]}")
return self.evaluationFunction(gameState),[]
kInf=1e100
res_action=[]
res_val=0
if agentIndex==0:
# Max
res_val = -kInf
for action in legal_actions:
successorGameState = gameState.generateSuccessor(agentIndex,action)
nxt_depth=depth_remain-1 if agentIndex==gameState.getNumAgents()-1 else depth_remain
val,action_list=self.ExpectMaxSearch(successorGameState,nxt_depth,(agentIndex+1)%gameState.getNumAgents())
if action=="Stop":
val-=100
if val>res_val:
res_val=val
# print(f"action:{action}, action_list:{action_list}")
res_action=[action]+action_list
else:
# Mins
res_val = kInf
val_list=[]
for action in legal_actions:
successorGameState = gameState.generateSuccessor(agentIndex,action)
nxt_depth=depth_remain-1 if agentIndex==gameState.getNumAgents()-1 else depth_remain
val,action_list=self.ExpectMaxSearch(successorGameState,nxt_depth,(agentIndex+1)%gameState.getNumAgents())
val_list.append(val)
if val<res_val:
res_val=val
res_action=[action]+action_list
res_val=sum(val_list)/len(val_list)
# print(f"depth_remain:{depth_remain}")
# print(f"returning {res_val}, {res_action}")
return res_val,res_action
def getAction(self, gameState: GameState):
"""
Returns the expectimax action using self.depth and self.evaluationFunction
All ghosts should be modeled as choosing uniformly at random from their
legal moves.
"""
global last_score
last_score=gameState.getScore()
stat = self.ExpectMaxSearch(gameState,self.depth,0)
# print(f"stat:{stat}")
return stat[1][0]
class PacmanDeepQAgent(PacmanQAgent):
def __init__(self, layout_input="smallGrid", target_update_rate=300, doubleQ=True, **args):
PacmanQAgent.__init__(self, **args)
self.model = None
self.target_model = None
self.target_update_rate = target_update_rate
self.update_amount = 0
self.epsilon_explore = 1.0
self.epsilon0 = 0.4
self.minimal_epsilon = 0.01
if model.kProductionMode:
self.epsilon_explore=0.01
self.epsilon0=0.01
self.minimal_epsilon=0.01
print("in production mode, epsilon set to 0.01")
self.epsilon = self.epsilon0
self.discount = 0.95
self.update_frequency = 3
self.counts = None
self.replay_memory = ReplayMemory(50000)
self.min_transitions_before_training = 10000
self.td_error_clipping = 10
# Initialize Q networks:
if isinstance(layout_input, str):
layout_instantiated = layout.getLayout(layout_input)
else:
layout_instantiated = layout_input
self.state_dim = self.get_state_dim(layout_instantiated)
self.initialize_q_networks(self.state_dim)
self.doubleQ = doubleQ
if self.doubleQ:
self.target_update_rate = -1
self.guiding_agent = ExpectimaxAgent()
def get_state_dim(self, layout):
pac_ft_size = 2
ghost_ft_size = 2 * layout.getNumGhosts()
food_capsule_ft_size = layout.width * layout.height
return pac_ft_size + ghost_ft_size + food_capsule_ft_size
def get_features(self, state):
pacman_state = np.array(state.getPacmanPosition())
ghost_state = np.array(state.getGhostPositions())
capsules = state.getCapsules()
food_locations = np.array(state.getFood().data).astype(np.float32)
for x, y in capsules:
food_locations[x][y] = 2
return np.concatenate((pacman_state, ghost_state.flatten(), food_locations.flatten()))
def initialize_q_networks(self, state_dim, action_dim=5):
import model
self.model = model.DeepQNetwork(state_dim, action_dim)
self.target_model = model.DeepQNetwork(state_dim, action_dim)
if os.path.exists('para.bin'):
print("Loading model parameters from para.bin")
checkpoint = torch.load('para.bin')
self.model.load_state_dict(checkpoint['model_state_dict'])
self.target_model.load_state_dict(checkpoint['target_model_state_dict'])
self.model.optimizer.load_state_dict(checkpoint['model_optimizer_state_dict'])
self.target_model.optimizer.load_state_dict(checkpoint['target_model_optimizer_state_dict'])
self.replay_memory = checkpoint['memory']
print(self.model.state_dict())
else:
print("Initializing new model parameters")
def save_model(self, filename="para.bin"):
if model.kProductionMode:
print("in production mode, not saving model")
return
print(f"Saving model parameters to {filename}")
torch.save({
'model_state_dict': self.model.state_dict(),
'target_model_state_dict': self.target_model.state_dict(),
'model_optimizer_state_dict': self.model.optimizer.state_dict(),
"target_model_optimizer_state_dict": self.target_model.optimizer.state_dict(),
"memory": self.replay_memory
}, filename)
print(self.model.state_dict())
def getQValue(self, state, action):
"""
Should return Q(state,action) as predicted by self.model
"""
feats = self.get_features(state)
legalActions = self.getLegalActions(state)
action_index = legalActions.index(action)
state = torch.tensor(np.array([feats]).astype("float64"), dtype=torch.double)
return self.model.run(state).data[0][action_index]
def shape_reward(self, reward):
if reward > 100:
reward = 10
elif reward > 0 and reward < 10:
reward = 2
elif reward == -1:
reward = 0
elif reward < -100:
reward = -10
return reward
def compute_q_targets(self, minibatch, network = None, target_network=None, doubleQ=False):
"""Prepare minibatches
Args:
minibatch (List[Transition]): Minibatch of `Transition`
Returns:
float: Loss value
"""
if network is None:
network = self.model
if target_network is None:
target_network = self.target_model
states = np.vstack([x.state for x in minibatch])
states = torch.tensor(states, dtype=torch.double)
actions = np.array([x.action for x in minibatch])
rewards = np.array([x.reward for x in minibatch])
next_states = np.vstack([x.next_state for x in minibatch])
next_states = torch.tensor(next_states)
done = np.array([x.done for x in minibatch])
Q_predict = network.run(states).data.detach().cpu().numpy()
Q_target = np.copy(Q_predict )
state_indices = states.int().detach().numpy()
state_indices = (state_indices[:, 0], state_indices[:, 1])
exploration_bonus = 1 / (2 * np.sqrt((self.counts[state_indices] / 100)))
replace_indices = np.arange(actions.shape[0])
action_indices = np.argmax(network.run(next_states).data.cpu(), axis=1)
target = rewards + exploration_bonus + (1 - done) * self.discount * target_network.run(next_states).data[replace_indices, action_indices].detach().cpu().numpy()
Q_target[replace_indices, actions] = target
if self.td_error_clipping is not None:
Q_target = Q_predict + np.clip(
Q_target - Q_predict, -self.td_error_clipping, self.td_error_clipping)
return Q_target
def update(self, state, action, nextState, reward):
legalActions = self.getLegalActions(state)
action_index = legalActions.index(action)
done = nextState.isLose() or nextState.isWin()
reward = self.shape_reward(reward)
if self.counts is None:
x, y = np.array(state.getFood().data).shape
self.counts = np.ones((x, y))
state = self.get_features(state)
nextState = self.get_features(nextState)
self.counts[int(state[0])][int(state[1])] += 1
transition = (state, action_index, reward, nextState, done)
self.replay_memory.push(*transition)
if len(self.replay_memory) < self.min_transitions_before_training:
self.epsilon = self.epsilon_explore
else:
self.epsilon = max(self.epsilon0 * (1 - self.update_amount / 20000), self.minimal_epsilon)
if len(self.replay_memory) > self.min_transitions_before_training and self.update_amount % self.update_frequency == 0:
minibatch = self.replay_memory.pop(self.model.batch_size)
states = np.vstack([x.state for x in minibatch])
states = torch.tensor(states.astype("float64"), dtype=torch.double)
Q_target1 = self.compute_q_targets(minibatch, self.model, self.target_model, doubleQ=self.doubleQ)
Q_target1 = torch.tensor(Q_target1.astype("float64"), dtype=torch.double)
if self.doubleQ:
Q_target2 = self.compute_q_targets(minibatch, self.target_model, self.model, doubleQ=self.doubleQ)
Q_target2 = torch.tensor(Q_target2.astype("float64"), dtype=torch.double)
self.model.gradient_update(states, Q_target1)
if self.doubleQ:
self.target_model.gradient_update(states, Q_target2)
if self.target_update_rate > 0 and self.update_amount % self.target_update_rate == 0:
self.target_model.set_weights(copy.deepcopy(self.model.parameters))
self.update_amount += 1
def final(self, state):
"""Called at the end of each game."""
PacmanQAgent.final(self, state)