Python planner.RoutePlanner方法代码示例

# 需要导入模块: import planner [as 别名]
# 或者: from planner import RoutePlanner [as 别名]
def __init__(self, env):
        super(LearningAgent, self).__init__(env)  # sets self.env = env, state = None, next_waypoint = None, and a default color
        self.color = 'red'  # override color
        self.planner = RoutePlanner(self.env, self)  # simple route planner to get next_waypoint
        # TODO: Initialize any additional variables here
        self.possible_actions = ['forward', 'left', 'right', None]
        self.possible_weights = [0, 0, 0, 0]
        self.Qvalues = {}
        self.initialQvalue = 10
        self.Qiterations = {}
		# Constants
        self.alpha = 1 #learning rate, will decrease with iterations
        self.gamma = .1 #discount factor 

# 需要导入模块: import planner [as 别名]
# 或者: from planner import RoutePlanner [as 别名]
def __init__(self, env):
        super(LearningAgent, self).__init__(env)  # sets self.env = env, state = None, next_waypoint = None, and a default color
        self.color = 'red'  # override color
        self.planner = RoutePlanner(self.env, self)  # simple route planner to get next_waypoint
        # TODO: Initialize any additional variables here
        self.Qdf = self.init_Q()
        self.past_state = 0
        self.past_reward = 0
        self.past_action = 'None'
        self.epsilon_iter = 1
        self.state_visit_hist=np.zeros(10,dtype=np.int) 

# 需要导入模块: import planner [as 别名]
# 或者: from planner import RoutePlanner [as 别名]
def __init__(self, env, learning=False, epsilon=1.0, alpha=0.5):
        super(LearningAgent, self).__init__(env)     # Set the agent in the evironment 
        self.planner = RoutePlanner(self.env, self)  # Create a route planner
        self.valid_actions = self.env.valid_actions  # The set of valid actions

        # Set parameters of the learning agent
        self.learning = learning # Whether the agent is expected to learn
        self.Q = dict()          # Create a Q-table which will be a dictionary of tuples
        self.epsilon = epsilon   # Random exploration factor
        self.alpha = alpha       # Learning factor

        ###########
        ## TO DO ##
        ###########
        # Set any additional class parameters as needed 

# 需要导入模块: import planner [as 别名]
# 或者: from planner import RoutePlanner [as 别名]
def __init__(self, env):
        super(LearningAgent, self).__init__(env)  # sets self.env = env, state = None, next_waypoint = None, and a default color
        self.color = 'red'  # override color
        self.planner = RoutePlanner(self.env, self)  # simple route planner to get next_waypoint
        self.trial = 0 # [0,99]
        self.unlocked = False
        self.trial_end = False
        # TODO: Initialize any additional variables here
        # Valid actions
        self.actions = [None, 'forward', 'left', 'right']

        # Q-Learning
        # Alpha (learning rate)
        self.alpha = get_simulation_params(0)[0] # should decay with t too?
        # Gamma (discount factor)
        self.gamma = get_simulation_params(0)[1] #
        self.epsilon = get_simulation_params(0)[2] # equal chance 0.5, progressive decay with t value
        self.decay_factor = 1.0
        self.Q = {}
        self.Q_default_value = 0.0 #not learning yet

        # Report
        self.total_reward = []
        self.total_penalties = []
        self.trial_reward = 0
        self.trial_penalty = 0
        self.success = 0
        self.failure = 0
        self.last_failure = 0 

# 需要导入模块: import planner [as 别名]
# 或者: from planner import RoutePlanner [as 别名]
def __init__(self, env, init_value=0, gamma=0.90, alpha=0.20, epsilon=0.10,
                 discount_deadline=False, history=0):
        super(LearningAgent, self).__init__(env)  # sets self.env = env, state = None, next_waypoint = None, and a default color
        self.color = 'red'  # override default color
        self.planner = RoutePlanner(self.env, self)  # simple route planner to get next_waypoint
        
        ## Initialize the Q-function as a dictionary (state) of dictionaries (actions)
        self.q_function = {}
        self.history = history
        if self.history > 0:
            self.init_q_function()
        
        ## Initial value of any (state, action) tuple is an arbitrary random number
        self.init_value = init_value
        ## Discount factor gamma: 0 (myopic) vs 1 (long-term optimal)
        self.gamma = gamma
        self.discount_deadline = discount_deadline
        
        ## Learning rate alpha: 0 (no learning) vs 1 (consider only most recent information)
        ## NOTE: Normally, alpha decreases over time: for example, alpha = 1 / t
        self.alpha = alpha
        ## Parameter of the epsilon-greedy action selection strategy
        ## NOTE: Normally, epsilon should also be decayed by the number of trials
        self.epsilon = epsilon
        
        ## The trial number
        self.trial = 1
        ## The cumulative reward
        self.cumulative_reward = 0 

# 需要导入模块: import planner [as 别名]
# 或者: from planner import RoutePlanner [as 别名]
def __init__(self, env):
        super(LearningAgent, self).__init__(env)  # sets self.env = env, state = None, next_waypoint = None, and a default color
        self.color = 'red'  # override color
        self.planner = RoutePlanner(self.env, self)  # simple route planner to get next_waypoint
        # TODO: Initialize any additional variables here

        # How likely we are gone explore new paths?
        self.epsilon = 0.05

        # Q learning update formula:
        # https://en.wikipedia.org/wiki/Q-learning
        # Good tutorial to start:
        # http://mnemstudio.org/path-finding-q-learning-tutorial.htm
        self.learning_rate = 0.90

        #the initial value of Q value
        self.default_q = 0

        # discount factor
        self.gamma = 0.10

        self.Q_values = {}
        
        self.prev_state = None
        self.prev_action = None
        self.prev_reward = None

        self.penalty_num = 0
        self.move_num = 0 

# 需要导入模块: import planner [as 别名]
# 或者: from planner import RoutePlanner [as 别名]
def __init__(self, env):
        super(LearningAgent, self).__init__(env)  # sets self.env = env, state = None, next_waypoint = None, and a default color
        self.color = 'red'  # override color
        self.planner = RoutePlanner(self.env, self)  # simple route planner to get next_waypoint
        # TODO: Initialize any additional variables here
        self.Q = np.zeros((9,4))
        self.Q[1][2] = 1
        self.Q[3][3] = 1
        self.Q[6][3] = 1
        self.Q[7][1] = 1
        self.Q[0][2] = -1
        self.Q[2][2] = -1
        self.Q[4][3] = -1
        self.Q[5][3] = -1
        self.Q[8][1] = -1
        self.gamma = 0.8
        self.alpha = 0.3
        self.epsilon = 0.6
        self.epsilon_start = 0.6
        self.n_updates = 0
        # possible states, these 9 states represents all the 96 possible states given 
        # the values of the variables used to define a state
        self.states = np.array([
            ['left', 'green', 'forward', None],
            ['left', 'green', None, None],
            ['left', 'red', None, None],
            ['right', 'green', None, None],
            ['right', 'red', None, 'forward'],
            ['right', 'red', 'left', None],
            ['right', 'red', None, None],
            ['forward', 'green', None, None],
            ['forward', 'red', None, None]
        ]) 

# 需要导入模块: import planner [as 别名]
# 或者: from planner import RoutePlanner [as 别名]
def __init__(self, env):
        super(LearningAgent, self).__init__(env)  # sets self.env = env, state = None, next_waypoint = None, and a default color
        self.color = 'red'  # override color
        self.planner = RoutePlanner(self.env, self)  # simple route planner to get next_waypoint
        self.actions = [None, 'forward', 'left', 'right']
        self.learning_rate = 0.3
        self.state = None
        self.q = {}
        self.trips = 0