Source code for machina.envs.continuous2discrete_env

Continuous to discrete.

import gym
import numpy as np

[docs]class C2DEnv(object): """ Wrapper environment for converting continuous action space to multi discrete action space. Parameters ---------- env : gym.Env n_bins : int Number of bins for converting continuous to discrete. e.g. continuous action space is 0 ~ 1 and n_bins=5, action space is converted to [0, 0.25, 0.5, 0.75, 1] """ def __init__(self, env, n_bins=30): assert isinstance(env.ac_space, gym.spaces.Box) assert len(env.ac_space.shape) == 1 self.env = env self.n_bins = n_bins self.ac_space = gym.spaces.MultiDiscrete( env.ac_space.shape[0] * [n_bins]) if hasattr(env, 'original_env'): self.original_env = env.original_env else: self.original_env = env @property def observation_space(self): return self.env.ob_space @property def action_space(self): return self.ac_space @property def horizon(self): if hasattr(self.env, 'horizon'): return self.env._horizon
[docs] def reset(self): return self.env.reset()
[docs] def step(self, action): continuous_action = [] for a, low, high in zip(action, self.env.ac_space.low, self.env.ac_space.high): continuous_action.append(np.linspace(low, high, self.n_bins)[a]) action = np.array(continuous_action) next_obs, reward, done, info = self.env.step(action) return next_obs, reward, done, info
[docs] def render(self): self.env.render()
[docs] def terminate(self): self.env.terminate()