Python rl.agents.DDPGAgent() Examples

The following are 3 code examples of rl.agents.DDPGAgent(). You can vote up the ones you like or vote down the ones you don't like, and go to the original project or source file by following the links above each example. You may also want to check out all available functions/classes of the module rl.agents , or try the search function .
Example #1
Source File: test_continuous.py    From keras-rl2 with MIT License 6 votes vote down vote up 
def test_ddpg():
    # TODO: replace this with a simpler environment where we can actually test if it finds a solution
    env = gym.make('Pendulum-v0')
    np.random.seed(123)
    env.seed(123)
    random.seed(123)
    nb_actions = env.action_space.shape[0]

    actor = Sequential()
    actor.add(Flatten(input_shape=(1,) + env.observation_space.shape))
    actor.add(Dense(16))
    actor.add(Activation('relu'))
    actor.add(Dense(nb_actions))
    actor.add(Activation('linear'))

    action_input = Input(shape=(nb_actions,), name='action_input')
    observation_input = Input(shape=(1,) + env.observation_space.shape, name='observation_input')
    flattened_observation = Flatten()(observation_input)
    x = Concatenate()([action_input, flattened_observation])
    x = Dense(16)(x)
    x = Activation('relu')(x)
    x = Dense(1)(x)
    x = Activation('linear')(x)
    critic = Model(inputs=[action_input, observation_input], outputs=x)
    
    memory = SequentialMemory(limit=1000, window_length=1)
    random_process = OrnsteinUhlenbeckProcess(theta=.15, mu=0., sigma=.3)
    agent = DDPGAgent(nb_actions=nb_actions, actor=actor, critic=critic, critic_action_input=action_input,
                      memory=memory, nb_steps_warmup_critic=50, nb_steps_warmup_actor=50,
                      random_process=random_process, gamma=.99, target_model_update=1e-3)
    agent.compile([Adam(lr=1e-3), Adam(lr=1e-3)])

    agent.fit(env, nb_steps=400, visualize=False, verbose=0, nb_max_episode_steps=100)
    h = agent.test(env, nb_episodes=2, visualize=False, nb_max_episode_steps=100)
    # TODO: evaluate history
Example #2
Source File: test_continuous.py    From keras-rl with MIT License 6 votes vote down vote up 
def test_ddpg():
    # TODO: replace this with a simpler environment where we can actually test if it finds a solution
    env = gym.make('Pendulum-v0')
    np.random.seed(123)
    env.seed(123)
    random.seed(123)
    nb_actions = env.action_space.shape[0]

    actor = Sequential()
    actor.add(Flatten(input_shape=(1,) + env.observation_space.shape))
    actor.add(Dense(16))
    actor.add(Activation('relu'))
    actor.add(Dense(nb_actions))
    actor.add(Activation('linear'))

    action_input = Input(shape=(nb_actions,), name='action_input')
    observation_input = Input(shape=(1,) + env.observation_space.shape, name='observation_input')
    flattened_observation = Flatten()(observation_input)
    x = Concatenate()([action_input, flattened_observation])
    x = Dense(16)(x)
    x = Activation('relu')(x)
    x = Dense(1)(x)
    x = Activation('linear')(x)
    critic = Model(inputs=[action_input, observation_input], outputs=x)
    
    memory = SequentialMemory(limit=1000, window_length=1)
    random_process = OrnsteinUhlenbeckProcess(theta=.15, mu=0., sigma=.3)
    agent = DDPGAgent(nb_actions=nb_actions, actor=actor, critic=critic, critic_action_input=action_input,
                      memory=memory, nb_steps_warmup_critic=50, nb_steps_warmup_actor=50,
                      random_process=random_process, gamma=.99, target_model_update=1e-3)
    agent.compile([Adam(lr=1e-3), Adam(lr=1e-3)])

    agent.fit(env, nb_steps=400, visualize=False, verbose=0, nb_max_episode_steps=100)
    h = agent.test(env, nb_episodes=2, visualize=False, nb_max_episode_steps=100)
    # TODO: evaluate history
Example #3
Source File: keras_ddpg.py    From costar_plan with Apache License 2.0 5 votes vote down vote up 
def __init__(self, env, *args, **kwargs):
        super(KerasDDPGAgent, self).__init__(*args, **kwargs)
        self.env = env
        
        #assert len(env.action_space.shape) == 1
        #TODO: is there a way to output a tuple (6,1)
        nb_actions = sum(sum(1 for i in row if i) for row in self.env.action_space.sample())
        
        
        #TODO: terminology? feature or observation?
        observation = env.reset()
        
        print ">>>>>>>>>>>>>>>>>>>", observation.shape

        # TODO: find a way to customize network
        actor = Sequential()
        actor.add(Flatten(input_shape=(1,) + observation.shape))
        actor.add(Dense(16))
        actor.add(Activation('relu'))
        actor.add(Dense(16))
        actor.add(Activation('relu'))
        actor.add(Dense(16))
        actor.add(Activation('relu'))
        actor.add(Dense(nb_actions))
        actor.add(Activation('tanh'))
        actor.add(Lambda(lambda x: x * 3.14159))

        print(actor.summary())
        
        action_input = Input(shape=(nb_actions,), name='action_input')
        
        observation_input = Input(shape=(1,) + observation.shape, name='observation_input')
        flattened_observation = Flatten()(observation_input)
        x = merge([action_input, flattened_observation], mode='concat')
        x = Dense(32)(x)
        x = Activation('relu')(x)
        x = Dense(32)(x)
        x = Activation('relu')(x)
        x = Dense(32)(x)
        x = Activation('relu')(x)
        x = Dense(1)(x)
        x = Activation('linear')(x)
        critic = Model(input=[action_input, observation_input], output=x)
        print(critic.summary())
        
       
        memory = SequentialMemory(limit=500000, window_length=1)
        random_process = OrnsteinUhlenbeckProcess(size=nb_actions, theta=.15, mu=0., sigma=.3)
        self.agent = DDPGAgent(nb_actions=nb_actions, actor=actor, critic=critic, critic_action_input=action_input,
                          memory=memory, nb_steps_warmup_critic=1000, nb_steps_warmup_actor=1000,
                          random_process=random_process, gamma=.99, target_model_update=1e-3)
        self.agent.compile(Adam(lr=.001, clipnorm=1.), metrics=['mae'])