import itertools import numpy as np from typing import Any, Dict, List, Tuple, Union import gym from gym import error, spaces from mlagents_envs.base_env import ActionTuple, BaseEnv from mlagents_envs.base_env import DecisionSteps, TerminalSteps from mlagents_envs import logging_util class UnityGymException(error.Error): """ Any error related to the gym wrapper of ml-agents. """ pass logger = logging_util.get_logger(__name__) logging_util.set_log_level(logging_util.INFO) GymStepResult = Tuple[np.ndarray, float, bool, Dict] class UnityToGymWrapper(gym.Env): """ Provides Gym wrapper for Unity Learning Environments. """ def __init__( self, unity_env: BaseEnv, uint8_visual: bool = False, flatten_branched: bool = False, allow_multiple_obs: bool = False, ): """ Environment initialization :param unity_env: The Unity BaseEnv to be wrapped in the gym. Will be closed when the UnityToGymWrapper closes. :param uint8_visual: Return visual observations as uint8 (0-255) matrices instead of float (0.0-1.0). :param flatten_branched: If True, turn branched discrete action spaces into a Discrete space rather than MultiDiscrete. :param allow_multiple_obs: If True, return a list of np.ndarrays as observations with the first elements containing the visual observations and the last element containing the array of vector observations. If False, returns a single np.ndarray containing either only a single visual observation or the array of vector observations. """ self._env = unity_env # Take a single step so that the brain information will be sent over if not self._env.behavior_specs: self._env.step() self.visual_obs = None # Save the step result from the last time all Agents requested decisions. self._previous_decision_step: DecisionSteps = None self._flattener = None # Hidden flag used by Atari environments to determine if the game is over self.game_over = False self._allow_multiple_obs = allow_multiple_obs # Check brain configuration if len(self._env.behavior_specs) != 1: raise UnityGymException( "There can only be one behavior in a UnityEnvironment " "if it is wrapped in a gym." ) self.name = list(self._env.behavior_specs.keys())[0] self.group_spec = self._env.behavior_specs[self.name] if self._get_n_vis_obs() == 0 and self._get_vec_obs_size() == 0: raise UnityGymException( "There are no observations provided by the environment." ) if not self._get_n_vis_obs() >= 1 and uint8_visual: logger.warning( "uint8_visual was set to true, but visual observations are not in use. " "This setting will not have any effect." ) else: self.uint8_visual = uint8_visual if ( self._get_n_vis_obs() + self._get_vec_obs_size() >= 2 and not self._allow_multiple_obs ): logger.warning( "The environment contains multiple observations. " "You must define allow_multiple_obs=True to receive them all. " "Otherwise, only the first visual observation (or vector observation if" "there are no visual observations) will be provided in the observation." ) # Check for number of agents in scene. self._env.reset() decision_steps, _ = self._env.get_steps(self.name) self._check_agents(len(decision_steps)) self._previous_decision_step = decision_steps # Set action spaces if self.group_spec.action_spec.is_discrete(): self.action_size = self.group_spec.action_spec.discrete_size branches = self.group_spec.action_spec.discrete_branches if self.group_spec.action_spec.discrete_size == 1: self._action_space = spaces.Discrete(branches[0]) else: if flatten_branched: self._flattener = ActionFlattener(branches) self._action_space = self._flattener.action_space else: self._action_space = spaces.MultiDiscrete(branches) elif self.group_spec.action_spec.is_continuous(): if flatten_branched: logger.warning( "The environment has a non-discrete action space. It will " "not be flattened." ) self.action_size = self.group_spec.action_spec.continuous_size high = np.array([1] * self.group_spec.action_spec.continuous_size) self._action_space = spaces.Box(-high, high, dtype=np.float32) else: raise UnityGymException( "The gym wrapper does not provide explicit support for both discrete " "and continuous actions." ) # Set observations space list_spaces: List[gym.Space] = [] shapes = self._get_vis_obs_shape() for shape in shapes: if uint8_visual: list_spaces.append(spaces.Box(0, 255, dtype=np.uint8, shape=shape)) else: list_spaces.append(spaces.Box(0, 1, dtype=np.float32, shape=shape)) if self._get_vec_obs_size() > 0: # vector observation is last high = np.array([np.inf] * self._get_vec_obs_size()) list_spaces.append(spaces.Box(-high, high, dtype=np.float32)) if self._allow_multiple_obs: self._observation_space = spaces.Tuple(list_spaces) else: self._observation_space = list_spaces[0] # only return the first one def reset(self) -> Union[List[np.ndarray], np.ndarray]: """Resets the state of the environment and returns an initial observation. Returns: observation (object/list): the initial observation of the space. """ self._env.reset() decision_step, _ = self._env.get_steps(self.name) n_agents = len(decision_step) self._check_agents(n_agents) self.game_over = False res: GymStepResult = self._single_step(decision_step) return res[0] def step(self, action: List[Any]) -> GymStepResult: """Run one timestep of the environment's dynamics. When end of episode is reached, you are responsible for calling `reset()` to reset this environment's state. Accepts an action and returns a tuple (observation, reward, done, info). Args: action (object/list): an action provided by the environment Returns: observation (object/list): agent's observation of the current environment reward (float/list) : amount of reward returned after previous action done (boolean/list): whether the episode has ended. info (dict): contains auxiliary diagnostic information. """ if self._flattener is not None: # Translate action into list action = self._flattener.lookup_action(action) action = np.array(action).reshape((1, self.action_size)) action_tuple = ActionTuple() if self.group_spec.action_spec.is_continuous(): action_tuple.add_continuous(action) else: action_tuple.add_discrete(action) self._env.set_actions(self.name, action_tuple) self._env.step() decision_step, terminal_step = self._env.get_steps(self.name) self._check_agents(max(len(decision_step), len(terminal_step))) if len(terminal_step) != 0: # The agent is done self.game_over = True return self._single_step(terminal_step) else: return self._single_step(decision_step) def _single_step(self, info: Union[DecisionSteps, TerminalSteps]) -> GymStepResult: if self._allow_multiple_obs: visual_obs = self._get_vis_obs_list(info) visual_obs_list = [] for obs in visual_obs: visual_obs_list.append(self._preprocess_single(obs[0])) default_observation = visual_obs_list if self._get_vec_obs_size() >= 1: default_observation.append(self._get_vector_obs(info)[0, :]) else: if self._get_n_vis_obs() >= 1: visual_obs = self._get_vis_obs_list(info) default_observation = self._preprocess_single(visual_obs[0][0]) else: default_observation = self._get_vector_obs(info)[0, :] if self._get_n_vis_obs() >= 1: visual_obs = self._get_vis_obs_list(info) self.visual_obs = self._preprocess_single(visual_obs[0][0]) done = isinstance(info, TerminalSteps) return (default_observation, info.reward[0], done, {"step": info}) def _preprocess_single(self, single_visual_obs: np.ndarray) -> np.ndarray: if self.uint8_visual: return (255.0 * single_visual_obs).astype(np.uint8) else: return single_visual_obs def _get_n_vis_obs(self) -> int: result = 0 for sen_spec in self.group_spec.sensor_specs: if len(sen_spec.shape) == 3: result += 1 return result def _get_vis_obs_shape(self) -> List[Tuple]: result: List[Tuple] = [] for sen_spec in self.group_spec.sensor_specs: if len(sen_spec.shape) == 3: result.append(sen_spec.shape) return result def _get_vis_obs_list( self, step_result: Union[DecisionSteps, TerminalSteps] ) -> List[np.ndarray]: result: List[np.ndarray] = [] for obs in step_result.obs: if len(obs.shape) == 4: result.append(obs) return result def _get_vector_obs( self, step_result: Union[DecisionSteps, TerminalSteps] ) -> np.ndarray: result: List[np.ndarray] = [] for obs in step_result.obs: if len(obs.shape) == 2: result.append(obs) return np.concatenate(result, axis=1) def _get_vec_obs_size(self) -> int: result = 0 for sen_spec in self.group_spec.sensor_specs: if len(sen_spec.shape) == 1: result += sen_spec.shape[0] return result def render(self, mode="rgb_array"): return self.visual_obs def close(self) -> None: """Override _close in your subclass to perform any necessary cleanup. Environments will automatically close() themselves when garbage collected or when the program exits. """ self._env.close() def seed(self, seed: Any = None) -> None: """Sets the seed for this env's random number generator(s). Currently not implemented. """ logger.warning("Could not seed environment %s", self.name) return @staticmethod def _check_agents(n_agents: int) -> None: if n_agents > 1: raise UnityGymException( f"There can only be one Agent in the environment but {n_agents} were detected." ) @property def metadata(self): return {"render.modes": ["rgb_array"]} @property def reward_range(self) -> Tuple[float, float]: return -float("inf"), float("inf") @property def spec(self): return None @property def action_space(self): return self._action_space @property def observation_space(self): return self._observation_space class ActionFlattener: """ Flattens branched discrete action spaces into single-branch discrete action spaces. """ def __init__(self, branched_action_space): """ Initialize the flattener. :param branched_action_space: A List containing the sizes of each branch of the action space, e.g. [2,3,3] for three branches with size 2, 3, and 3 respectively. """ self._action_shape = branched_action_space self.action_lookup = self._create_lookup(self._action_shape) self.action_space = spaces.Discrete(len(self.action_lookup)) @classmethod def _create_lookup(self, branched_action_space): """ Creates a Dict that maps discrete actions (scalars) to branched actions (lists). Each key in the Dict maps to one unique set of branched actions, and each value contains the List of branched actions. """ possible_vals = [range(_num) for _num in branched_action_space] all_actions = [list(_action) for _action in itertools.product(*possible_vals)] # Dict should be faster than List for large action spaces action_lookup = { _scalar: _action for (_scalar, _action) in enumerate(all_actions) } return action_lookup def lookup_action(self, action): """ Convert a scalar discrete action into a unique set of branched actions. :param: action: A scalar value representing one of the discrete actions. :return: The List containing the branched actions. """ return self.action_lookup[action]