environment.py

import gym
import my_gym  # Necessary to see CartPoleContinuous, though PyCharm does not understand this
import numpy as np
from wrappers import BetaWrapperCartP,BetaWrapperPendul,FeatureInverter, BinaryShifter, BinaryShifterDiscrete, ActionVectorAdapter, \
    PerfWriter, PendulumWrapper, MountainCarContinuousWrapper
from gym.wrappers import TimeLimit

# to see the list of available gym environments, type:
# from gym import envs
# print(envs.registry.all())


class Simulator(object):
    """
    Object used by the multi-threading to give each worker an environnement
    """
    def __init__(self, args):
        self.env = make_env(args.env_name, args.policy_type,
                            args.max_episode_steps, args.env_obs_space_name)
        self.env.reset()

    def step(self, action):
        return self.env.step(action)


def make_env(env_name,
             policy_type,
             max_episode_steps,
             env_obs_space_name=None):
    """
    Wrap the environment into a set of wrappers depending on some hyper-parameters
    Used so that most environments can be used with the same policies and algorithms
    :param env_name: the name of the environment, as a string. For instance, "MountainCarContinuous-v0"
    :param policy_type: a string specifying the type of policy. So far, "bernoulli" or "normal"
    :param max_episode_steps: the max duration of an episode. If None, uses the default gym max duration
    :param env_obs_space_name: a vector of names of the environment features. E.g. ["position","velocity"] for MountainCar
    :return: the wrapped environment
    """
    env = gym.make(env_name)
    # tests whether the environment is discrete or continuous
    # if not env.action_space.contains(np.array([0.5])):
    #    assert policy_type == "bernoulli" or policy_type =="discrete", 'cannot run a continuous action policy in a discrete action environment'

    if max_episode_steps is not None:
        env = TimeLimit(env, max_episode_steps)
    if env_name == "CartPole-v0" or env_name == "CartPoleContinuous-v0":
        env = FeatureInverter(env, 1, 2)
        env = ActionVectorAdapter(env)
        if policy_type == "beta":
            env = BetaWrapperCartP(env)

    #### MODIF : added actionVectorAdapter on Mountain car.
    if env_name == "MountainCar-v0":
        env = ActionVectorAdapter(env)
    #### MODIF

    env.observation_space.names = env_obs_space_name

    if policy_type == "bernoulli":
        # tests whether the environment is discrete or continuous
        if env.action_space.contains(np.array([0.5])):
            env = BinaryShifter(env)
        else:
            env = BinaryShifterDiscrete(env)

    if env_name == "Pendulum-v0":
        env = PendulumWrapper(env)
        if policy_type == "beta":
            env = BetaWrapperPendul(env)

    if env_name == "MountainCarContinuous-v0":
        env = MountainCarContinuousWrapper(env)


    env = PerfWriter(env)
    #print(env)
    return env