Source code for nnabla_rl.environment_explorers.gaussian_explorer

# Copyright 2020,2021 Sony Corporation.
# Copyright 2021,2022,2023 Sony Group Corporation.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
#     http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.

from dataclasses import dataclass
from typing import Dict, Tuple, cast

import numpy as np

from nnabla_rl.environment_explorer import EnvironmentExplorer, EnvironmentExplorerConfig
from nnabla_rl.environments.environment_info import EnvironmentInfo
from nnabla_rl.typing import ActionSelector


@dataclass
class GaussianExplorerConfig(EnvironmentExplorerConfig):
    """List of configurations for gaussian explorer.

    Args:
        action_clip_low (float): Minimum noise value. Noise below this value will be clipped.
            Defaults to np.finfo(np.float).min
        action_clip_high (float): Maximum noise value. Noise above this value will be clipped.
            Defaults to np.finfo(np.float).max
        sigma (float): Standard deviation of gaussian noise. Must be positive. Defaults to 1.0.
    """

    action_clip_low: float = cast(float, np.finfo(np.float32).min)
    action_clip_high: float = cast(float, np.finfo(np.float32).max)
    sigma: float = 1.0

    def __post_init__(self):
        self._assert_positive(self.sigma, 'sigma')


[docs]class GaussianExplorer(EnvironmentExplorer): """Gaussian explorer. Explore using policy's action with gaussian noise appended to it. Policy's action must be continuous action. Args: policy_action_selector (:py:class:`ActionSelector <nnabla_rl.typing.ActionSelector>`): callable which computes current policy's action with respect to current state. env_info (:py:class:`EnvironmentInfo <nnabla_rl.environments.environment_info.EnvironmentInfo>`): environment info config (:py:class:`LinearDecayEpsilonGreedyExplorerConfig\ <nnabla_rl.environment_explorers.LinearDecayEpsilonGreedyExplorerConfig>`): the config of this class. """ # type declarations to type check with mypy # NOTE: declared variables are instance variable and NOT class variable, unless it is marked with ClassVar # See https://mypy.readthedocs.io/en/stable/class_basics.html for details _config: GaussianExplorerConfig def __init__(self, policy_action_selector: ActionSelector, env_info: EnvironmentInfo, config: GaussianExplorerConfig = GaussianExplorerConfig()): super().__init__(env_info, config) self._policy_action_selector = policy_action_selector
[docs] def action(self, step: int, state: np.ndarray, *, begin_of_episode: bool = False) -> Tuple[np.ndarray, Dict]: (action, info) = self._policy_action_selector(state, begin_of_episode=begin_of_episode) return self._append_noise(action, self._config.action_clip_low, self._config.action_clip_high), info
def _append_noise(self, action, low, high): noise = np.random.normal(loc=0.0, scale=self._config.sigma, size=action.shape).astype(np.float32) return np.clip(action + noise, low, high)