""" Definition of the RailEnv environment and related level-generation functions. Generator functions are functions that take width, height and num_resets as arguments and return a GridTransitionMap object. """ # TODO: _ this is a global method --> utils or remove later from enum import IntEnum import msgpack import numpy as np from flatland.core.env import Environment from flatland.envs.agent_utils import EnvAgentStatic, EnvAgent from flatland.envs.env_utils import get_new_position from flatland.envs.generators import random_rail_generator from flatland.envs.observations import TreeObsForRailEnv class RailEnvActions(IntEnum): DO_NOTHING = 0 MOVE_LEFT = 1 MOVE_FORWARD = 2 MOVE_RIGHT = 3 STOP_MOVING = 4 class RailEnv(Environment): """ RailEnv environment class. RailEnv is an environment inspired by a (simplified version of) a rail network, in which agents (trains) have to navigate to their target locations in the shortest time possible, while at the same time cooperating to avoid bottlenecks. The valid actions in the environment are: 0: do nothing 1: turn left and move to the next cell; if the agent was not moving, movement is started 2: move to the next cell in front of the agent; if the agent was not moving, movement is started 3: turn right and move to the next cell; if the agent was not moving, movement is started 4: stop moving Moving forward in a dead-end cell makes the agent turn 180 degrees and step to the cell it came from. The actions of the agents are executed in order of their handle to prevent deadlocks and to allow them to learn relative priorities. TODO: WRITE ABOUT THE REWARD FUNCTION, and possibly allow for alpha and beta to be passed as parameters to __init__(). """ def __init__(self, width, height, rail_generator=random_rail_generator(), number_of_agents=1, obs_builder_object=TreeObsForRailEnv(max_depth=2), prediction_builder_object=None ): """ Environment init. Parameters ------- rail_generator : function The rail_generator function is a function that takes the width, height and agents handles of a rail environment, along with the number of times the env has been reset, and returns a GridTransitionMap object and a list of starting positions, targets, and initial orientations for agent handle. Implemented functions are: random_rail_generator : generate a random rail of given size rail_from_GridTransitionMap_generator(rail_map) : generate a rail from a GridTransitionMap object rail_from_manual_specifications_generator(rail_spec) : generate a rail from a rail specifications array TODO: generate_rail_from_saved_list or from list of ndarray bitmaps --- width : int The width of the rail map. Potentially in the future, a range of widths to sample from. height : int The height of the rail map. Potentially in the future, a range of heights to sample from. number_of_agents : int Number of agents to spawn on the map. Potentially in the future, a range of number of agents to sample from. obs_builder_object: ObservationBuilder object ObservationBuilder-derived object that takes builds observation vectors for each agent. """ self.rail_generator = rail_generator self.rail = None self.width = width self.height = height self.obs_builder = obs_builder_object self.obs_builder._set_env(self) self.prediction_builder = prediction_builder_object if self.prediction_builder: self.prediction_builder._set_env(self) self.action_space = [1] self.observation_space = self.obs_builder.observation_space # updated on resets? self.actions = [0] * number_of_agents self.rewards = [0] * number_of_agents self.done = False self.dones = dict.fromkeys(list(range(number_of_agents)) + ["__all__"], False) self.obs_dict = {} self.rewards_dict = {} self.dev_obs_dict = {} self.agents = [None] * number_of_agents # live agents self.agents_static = [None] * number_of_agents # static agent information self.num_resets = 0 self.reset() self.num_resets = 0 # yes, set it to zero again! self.valid_positions = None # no more agent_handles def get_agent_handles(self): return range(self.get_num_agents()) def get_num_agents(self, static=True): if static: return len(self.agents_static) else: return len(self.agents) def add_agent_static(self, agent_static): """ Add static info for a single agent. Returns the index of the new agent. """ self.agents_static.append(agent_static) return len(self.agents_static) - 1 def restart_agents(self): """ Reset the agents to their starting positions defined in agents_static """ self.agents = EnvAgent.list_from_static(self.agents_static) def reset(self, regen_rail=True, replace_agents=True): """ if regen_rail then regenerate the rails. if replace_agents then regenerate the agents static. Relies on the rail_generator returning agent_static lists (pos, dir, target) """ tRailAgents = self.rail_generator(self.width, self.height, self.get_num_agents(), self.num_resets) if regen_rail or self.rail is None: self.rail = tRailAgents[0] if replace_agents: self.agents_static = EnvAgentStatic.from_lists(*tRailAgents[1:4]) self.restart_agents() self.num_resets += 1 # TODO perhaps dones should be part of each agent. self.dones = dict.fromkeys(list(range(self.get_num_agents())) + ["__all__"], False) # Reset the state of the observation builder with the new environment self.obs_builder.reset() self.observation_space = self.obs_builder.observation_space # <-- change on reset? # Return the new observation vectors for each agent return self._get_observations() def step(self, action_dict_): action_dict = action_dict_.copy() alpha = 1.0 beta = 1.0 invalid_action_penalty = 0 # previously -2; GIACOMO: we decided that invalid actions will carry no penalty step_penalty = -1 * alpha global_reward = 1 * beta stop_penalty = 0 # penalty for stopping a moving agent start_penalty = 0 # penalty for starting a stopped agent # Reset the step rewards self.rewards_dict = dict() for iAgent in range(self.get_num_agents()): self.rewards_dict[iAgent] = 0 if self.dones["__all__"]: self.rewards_dict = [r + global_reward for r in self.rewards_dict] return self._get_observations(), self.rewards_dict, self.dones, {} # for i in range(len(self.agents_handles)): for iAgent in range(self.get_num_agents()): agent = self.agents[iAgent] if iAgent not in action_dict: # no action has been supplied for this agent if agent.moving: # Keep moving # Change MOVE_FORWARD to DO_NOTHING action_dict[iAgent] = RailEnvActions.DO_NOTHING else: action_dict[iAgent] = RailEnvActions.DO_NOTHING if self.dones[iAgent]: # this agent has already completed... continue action = action_dict[iAgent] if action < 0 or action > len(RailEnvActions): print('ERROR: illegal action=', action, 'for agent with index=', iAgent) return if action == RailEnvActions.DO_NOTHING and agent.moving: # Keep moving action = RailEnvActions.MOVE_FORWARD if action == RailEnvActions.STOP_MOVING and agent.moving: agent.moving = False self.rewards_dict[iAgent] += stop_penalty if not agent.moving and not (action == RailEnvActions.DO_NOTHING or action == RailEnvActions.STOP_MOVING): # Only allow agent to start moving by pressing forward. agent.moving = True self.rewards_dict[iAgent] += start_penalty if action != RailEnvActions.DO_NOTHING and action != RailEnvActions.STOP_MOVING: cell_isFree, new_cell_isValid, new_direction, new_position, transition_isValid = \ self._check_action_on_agent(action, agent) if all([new_cell_isValid, transition_isValid, cell_isFree]): agent.old_direction = agent.direction agent.old_position = agent.position agent.position = new_position agent.direction = new_direction else: # Logic: if the chosen action is invalid, # and it was LEFT or RIGHT, and the agent was moving, then keep moving FORWARD. if (action == RailEnvActions.MOVE_LEFT or action == RailEnvActions.MOVE_RIGHT) and agent.moving: cell_isFree, new_cell_isValid, new_direction, new_position, transition_isValid = \ self._check_action_on_agent(RailEnvActions.MOVE_FORWARD, agent) if all([new_cell_isValid, transition_isValid, cell_isFree]): agent.old_direction = agent.direction agent.old_position = agent.position agent.position = new_position agent.direction = new_direction else: # the action was not valid, add penalty self.rewards_dict[iAgent] += invalid_action_penalty else: # the action was not valid, add penalty self.rewards_dict[iAgent] += invalid_action_penalty if np.equal(agent.position, agent.target).all(): self.dones[iAgent] = True else: self.rewards_dict[iAgent] += step_penalty # Check for end of episode + add global reward to all rewards! if np.all([np.array_equal(agent2.position, agent2.target) for agent2 in self.agents]): self.dones["__all__"] = True self.rewards_dict = [0 * r + global_reward for r in self.rewards_dict] # Reset the step actions (in case some agent doesn't 'register_action' # on the next step) self.actions = [0] * self.get_num_agents() return self._get_observations(), self.rewards_dict, self.dones, {} def _check_action_on_agent(self, action, agent): # compute number of possible transitions in the current # cell used to check for invalid actions new_direction, transition_isValid = self.check_action(agent, action) new_position = get_new_position(agent.position, new_direction) # Is it a legal move? # 1) transition allows the new_direction in the cell, # 2) the new cell is not empty (case 0), # 3) the cell is free, i.e., no agent is currently in that cell new_cell_isValid = ( np.array_equal( # Check the new position is still in the grid new_position, np.clip(new_position, [0, 0], [self.height - 1, self.width - 1])) and # check the new position has some transitions (ie is not an empty cell) self.rail.get_transitions(new_position) > 0) # If transition validity hasn't been checked yet. if transition_isValid is None: transition_isValid = self.rail.get_transition( (*agent.position, agent.direction), new_direction) # Check the new position is not the same as any of the existing agent positions # (including itself, for simplicity, since it is moving) cell_isFree = not np.any( np.equal(new_position, [agent2.position for agent2 in self.agents]).all(1)) return cell_isFree, new_cell_isValid, new_direction, new_position, transition_isValid def predict(self): if not self.prediction_builder: return {} return self.prediction_builder.get() def check_action(self, agent, action): transition_isValid = None possible_transitions = self.rail.get_transitions((*agent.position, agent.direction)) num_transitions = np.count_nonzero(possible_transitions) new_direction = agent.direction if action == RailEnvActions.MOVE_LEFT: new_direction = agent.direction - 1 if num_transitions <= 1: transition_isValid = False elif action == RailEnvActions.MOVE_RIGHT: new_direction = agent.direction + 1 if num_transitions <= 1: transition_isValid = False new_direction %= 4 if action == RailEnvActions.MOVE_FORWARD: if num_transitions == 1: # - dead-end, straight line or curved line; # new_direction will be the only valid transition # - take only available transition new_direction = np.argmax(possible_transitions) transition_isValid = True return new_direction, transition_isValid def _get_observations(self): self.obs_dict = self.obs_builder.get_many(list(range(self.get_num_agents()))) return self.obs_dict def _get_predictions(self): if not self.prediction_builder: return {} return {} def get_full_state_msg(self): grid_data = self.rail.grid.tolist() agent_static_data = [agent.to_list() for agent in self.agents_static] agent_data = [agent.to_list() for agent in self.agents] msgpack.packb(grid_data) msgpack.packb(agent_data) msgpack.packb(agent_static_data) msg_data = { "grid": grid_data, "agents_static": agent_static_data, "agents": agent_data} return msgpack.packb(msg_data, use_bin_type=True) def get_agent_state_msg(self): agent_data = [agent.to_list() for agent in self.agents] msg_data = { "agents": agent_data} return msgpack.packb(msg_data, use_bin_type=True) def set_full_state_msg(self, msg_data): data = msgpack.unpackb(msg_data, use_list=False) self.rail.grid = np.array(data[b"grid"]) # agents are always reset as not moving self.agents_static = [EnvAgentStatic(d[0], d[1], d[2], moving=False) for d in data[b"agents_static"]] self.agents = [EnvAgent(d[0], d[1], d[2], d[3], d[4]) for d in data[b"agents"]] # setup with loaded data self.height, self.width = self.rail.grid.shape self.rail.height = self.height self.rail.width = self.width self.dones = dict.fromkeys(list(range(self.get_num_agents())) + ["__all__"], False) def save(self, filename): with open(filename, "wb") as file_out: file_out.write(self.get_full_state_msg()) def load(self, filename): with open(filename, "rb") as file_in: load_data = file_in.read() self.set_full_state_msg(load_data)