rail_env.py 15.1 KB
Newer Older
1
2
3
4
5
6
"""
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.
"""
hagrid67's avatar
hagrid67 committed
7
# TODO:  _ this is a global method --> utils or remove later
8

9
10
from enum import IntEnum

maljx's avatar
maljx committed
11
import msgpack
12
import numpy as np
13
14

from flatland.core.env import Environment
15
from flatland.envs.agent_utils import EnvAgentStatic, EnvAgent
16
17
18
19
from flatland.envs.env_utils import get_new_position
from flatland.envs.generators import random_rail_generator
from flatland.envs.observations import TreeObsForRailEnv

20

spiglerg's avatar
spiglerg committed
21
22
23
24
25
26
27
class RailEnvActions(IntEnum):
    DO_NOTHING = 0
    MOVE_LEFT = 1
    MOVE_FORWARD = 2
    MOVE_RIGHT = 3
    STOP_MOVING = 4

u214892's avatar
u214892 committed
28

29
30
31
32
33
34
35
36
37
38
39
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
spiglerg's avatar
spiglerg committed
40
41
42
43
        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
44
45
46
47
48
49
50
51
52
53
54
55
56
57

    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,
spiglerg's avatar
spiglerg committed
58
                 rail_generator=random_rail_generator(),
59
                 number_of_agents=1,
u214892's avatar
u214892 committed
60
61
                 obs_builder_object=TreeObsForRailEnv(max_depth=2),
                 ):
62
63
64
65
66
67
        """
        Environment init.

        Parameters
        -------
        rail_generator : function
68
69
70
71
            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.
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
            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)

101
        self.action_space = [1]
spiglerg's avatar
spiglerg committed
102
        self.observation_space = self.obs_builder.observation_space  # updated on resets?
103

104
105
        self.actions = [0] * number_of_agents
        self.rewards = [0] * number_of_agents
106
107
        self.done = False

108
109
        self.dones = dict.fromkeys(list(range(number_of_agents)) + ["__all__"], False)

110
111
        self.obs_dict = {}
        self.rewards_dict = {}
Erik Nygren's avatar
Erik Nygren committed
112
        self.dev_obs_dict = {}
113

114
115
        self.agents = [None] * number_of_agents  # live agents
        self.agents_static = [None] * number_of_agents  # static agent information
116
117
        self.num_resets = 0
        self.reset()
118
        self.num_resets = 0  # yes, set it to zero again!
119

120
121
        self.valid_positions = None

122
    # no more agent_handles
123
    def get_agent_handles(self):
124
125
126
127
128
129
130
        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)
131

hagrid67's avatar
hagrid67 committed
132
133
134
135
136
137
138
    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

139
140
    def restart_agents(self):
        """ Reset the agents to their starting positions defined in agents_static
hagrid67's avatar
hagrid67 committed
141
        """
142
143
144
145
146
147
        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)
hagrid67's avatar
hagrid67 committed
148
        """
149
        tRailAgents = self.rail_generator(self.width, self.height, self.get_num_agents(), self.num_resets)
hagrid67's avatar
hagrid67 committed
150

151
        if regen_rail or self.rail is None:
hagrid67's avatar
hagrid67 committed
152
            self.rail = tRailAgents[0]
153

hagrid67's avatar
hagrid67 committed
154
        if replace_agents:
hagrid67's avatar
hagrid67 committed
155
156
            self.agents_static = EnvAgentStatic.from_lists(*tRailAgents[1:4])

157
        self.restart_agents()
hagrid67's avatar
hagrid67 committed
158

159
160
        self.num_resets += 1

u214892's avatar
u214892 committed
161
        # TODO perhaps dones should be part of each agent.
162
        self.dones = dict.fromkeys(list(range(self.get_num_agents())) + ["__all__"], False)
163

164
165
        # Reset the state of the observation builder with the new environment
        self.obs_builder.reset()
spiglerg's avatar
spiglerg committed
166
        self.observation_space = self.obs_builder.observation_space  # <-- change on reset?
167
168
169
170

        # Return the new observation vectors for each agent
        return self._get_observations()

spiglerg's avatar
spiglerg committed
171
172
173
    def step(self, action_dict_):
        action_dict = action_dict_.copy()

174
175
176
        alpha = 1.0
        beta = 1.0

177
        invalid_action_penalty = 0  # previously -2; GIACOMO: we decided that invalid actions will carry no penalty
178
179
        step_penalty = -1 * alpha
        global_reward = 1 * beta
180
181
        stop_penalty = 0  # penalty for stopping a moving agent
        start_penalty = 0  # penalty for starting a stopped agent
182
183

        # Reset the step rewards
184
        self.rewards_dict = dict()
185
186
        for iAgent in range(self.get_num_agents()):
            self.rewards_dict[iAgent] = 0
187
188

        if self.dones["__all__"]:
189
            self.rewards_dict = [r + global_reward for r in self.rewards_dict]
190
191
            return self._get_observations(), self.rewards_dict, self.dones, {}

192
193
194
        # for i in range(len(self.agents_handles)):
        for iAgent in range(self.get_num_agents()):
            agent = self.agents[iAgent]
Erik Nygren's avatar
Erik Nygren committed
195

196
            if iAgent not in action_dict:  # no action has been supplied for this agent
spiglerg's avatar
spiglerg committed
197
198
                if agent.moving:
                    # Keep moving
199
                    # Change MOVE_FORWARD to DO_NOTHING
200
                    action_dict[iAgent] = RailEnvActions.DO_NOTHING
spiglerg's avatar
spiglerg committed
201
202
                else:
                    action_dict[iAgent] = RailEnvActions.DO_NOTHING
203

204
            if self.dones[iAgent]:  # this agent has already completed...
205
                continue
206
            action = action_dict[iAgent]
207

spiglerg's avatar
spiglerg committed
208
            if action < 0 or action > len(RailEnvActions):
209
                print('ERROR: illegal action=', action,
210
                      'for agent with index=', iAgent)
211
212
                return

spiglerg's avatar
spiglerg committed
213
214
215
216
217
218
            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
219
220
                self.rewards_dict[iAgent] += stop_penalty

221
            if not agent.moving and not (action == RailEnvActions.DO_NOTHING or action == RailEnvActions.STOP_MOVING):
222
                # Only allow agent to start moving by pressing forward.
spiglerg's avatar
spiglerg committed
223
                agent.moving = True
224
                self.rewards_dict[iAgent] += start_penalty
spiglerg's avatar
spiglerg committed
225
226

            if action != RailEnvActions.DO_NOTHING and action != RailEnvActions.STOP_MOVING:
u214892's avatar
u214892 committed
227
228
                cell_isFree, new_cell_isValid, new_direction, new_position, transition_isValid = \
                    self._check_action_on_agent(action, agent)
229
                if all([new_cell_isValid, transition_isValid, cell_isFree]):
hagrid67's avatar
hagrid67 committed
230
                    agent.old_direction = agent.direction
231
232
                    agent.old_position = agent.position
                    agent.position = new_position
hagrid67's avatar
hagrid67 committed
233
                    agent.direction = new_direction
234
                else:
spiglerg's avatar
spiglerg committed
235
                    # Logic: if the chosen action is invalid,
236
                    # and it was LEFT or RIGHT, and the agent was moving, then keep moving FORWARD.
237
                    if (action == RailEnvActions.MOVE_LEFT or action == RailEnvActions.MOVE_RIGHT) and agent.moving:
spiglerg's avatar
spiglerg committed
238
239
                        cell_isFree, new_cell_isValid, new_direction, new_position, transition_isValid = \
                            self._check_action_on_agent(RailEnvActions.MOVE_FORWARD, agent)
240
241
242
243
244
245
246
247
248
249
250
251
252

                        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
253

254
            if np.equal(agent.position, agent.target).all():
255
                self.dones[iAgent] = True
256
            else:
257
                self.rewards_dict[iAgent] += step_penalty
258
259

        # Check for end of episode + add global reward to all rewards!
260
        if np.all([np.array_equal(agent2.position, agent2.target) for agent2 in self.agents]):
261
            self.dones["__all__"] = True
262
            self.rewards_dict = [0 * r + global_reward for r in self.rewards_dict]
263
264
265

        # Reset the step actions (in case some agent doesn't 'register_action'
        # on the next step)
266
        self.actions = [0] * self.get_num_agents()
267
268
        return self._get_observations(), self.rewards_dict, self.dones, {}

u214892's avatar
u214892 committed
269
270
271
272
273
274
    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?
spiglerg's avatar
spiglerg committed
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
        # 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

hagrid67's avatar
hagrid67 committed
295
296
297
298
299
300
    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
spiglerg's avatar
spiglerg committed
301
        if action == RailEnvActions.MOVE_LEFT:
hagrid67's avatar
hagrid67 committed
302
303
304
305
            new_direction = agent.direction - 1
            if num_transitions <= 1:
                transition_isValid = False

spiglerg's avatar
spiglerg committed
306
        elif action == RailEnvActions.MOVE_RIGHT:
hagrid67's avatar
hagrid67 committed
307
308
309
310
311
312
            new_direction = agent.direction + 1
            if num_transitions <= 1:
                transition_isValid = False

        new_direction %= 4

spiglerg's avatar
spiglerg committed
313
        if action == RailEnvActions.MOVE_FORWARD:
hagrid67's avatar
hagrid67 committed
314
315
316
317
318
319
320
321
            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

322
    def _get_observations(self):
323
        self.obs_dict = self.obs_builder.get_many(list(range(self.get_num_agents())))
324
        return self.obs_dict
325

maljx's avatar
maljx committed
326
327
328
329
    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]
330
331
332
333
334

        msgpack.packb(grid_data)
        msgpack.packb(agent_data)
        msgpack.packb(agent_static_data)

maljx's avatar
maljx committed
335
336
337
        msg_data = {
            "grid": grid_data,
            "agents_static": agent_static_data,
338
            "agents": agent_data}
maljx's avatar
maljx committed
339
340
341
342
343
        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 = {
344
            "agents": agent_data}
maljx's avatar
maljx committed
345
346
347
348
349
        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"])
spiglerg's avatar
fix?    
spiglerg committed
350
351
        # 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"]]
maljx's avatar
maljx committed
352
353
354
        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
355
356
        self.rail.height = self.height
        self.rail.width = self.width
maljx's avatar
maljx committed
357
358
359
360
361
362
363
364
365
366
        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)