-
Notifications
You must be signed in to change notification settings - Fork 21
/
cityflow_agent.py
310 lines (253 loc) · 15.3 KB
/
cityflow_agent.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
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
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
# -*- coding: utf-8 -*-
'''
Interacting with traffic_light_dqn.py and map_computor.py
1) retriving values from sumo_computor.py
2) update state
3) controling logic
'''
from agent import State
from sys import platform
import sys
import os
import map_computor
import numpy as np
import shutil
import json
import collections
from agent import State
import cityflow as engine
class Vehicles:
initial_speed = 5.0
def __init__(self):
# add what ever you need to maintain
self.id = None
self.speed = None
self.wait_time = None
self.stop_count = None
self.enter_time = None
self.has_read = False
self.first_stop_time = -1
self.entering = True
# --- my modification ---
self.recount_waiting_time = 0
class CityFlowAgent:
class ParaSet:
def __init__(self, dic_paras):
for key, value in dic_paras.items():
setattr(self, key, value)
if hasattr(self, "REWARDS_INFO_DICT"):
self.LIST_TRUE_REWARD = []
list_reward = list(self.REWARDS_INFO_DICT.keys())
list_reward.sort()
for reward_name in list_reward:
if self.REWARDS_INFO_DICT[reward_name][0]:
self.LIST_TRUE_REWARD.append(reward_name)
def __init__(self, path_set):
self.path_set = path_set
self.para_set = self.load_conf(os.path.join(self.path_set.PATH_TO_CONF, self.path_set.CITYFLOW_AGENT_CONF))
self.eng = map_computor.reset(self.para_set)
start_phase = map_computor.start_sumo(self.eng)
self.global_dic_vehicles = collections.OrderedDict()
self.global_dic_location_vehicles = collections.OrderedDict()
self.global_dic_this_node_vehicles = collections.OrderedDict()
self.global_dic_waiting_time_vehicles = collections.OrderedDict()
self.global_list_node_lane_vehicle = collections.OrderedDict()
for node_id_1 in map_computor.get_node_id_list():
self.global_dic_location_vehicles[node_id_1] = collections.OrderedDict()
self.global_dic_this_node_vehicles[node_id_1] = collections.OrderedDict()
for lane in map_computor.global_listLanes[node_id_1]:
self.global_dic_waiting_time_vehicles[lane] = collections.OrderedDict()
for lane in map_computor.global_all_lanes_each_node[node_id_1]:
self.global_list_node_lane_vehicle[lane] = []
self.current_phase = collections.OrderedDict()
self.current_phase_duration = collections.OrderedDict()
for node_id_index in range(len(map_computor.get_node_id_list())):
node_id = map_computor.get_node_id_list()[node_id_index]
self.current_phase[node_id] = start_phase[node_id_index]
self.current_phase_duration[node_id] = 0
self.state = collections.OrderedDict() # it's global!!!
self.all_vehicles_location_enter_time_dict = collections.OrderedDict()
self.all_vehicles_this_node_enter_time_dict = collections.OrderedDict()
for node_id_1 in map_computor.get_node_id_list():
self.all_vehicles_location_enter_time_dict[node_id_1] = collections.OrderedDict()
self.all_vehicles_this_node_enter_time_dict[node_id_1] = collections.OrderedDict()
self.update_state()
self.update_vehicles()
self.update_vehicles_location()
self.update_vehicle_arrive_leave_time()
self.f_log_rewards = os.path.join(self.path_set.PATH_TO_OUTPUT, "log_rewards.txt")
self.f_log_rewards_control = os.path.join(self.path_set.PATH_TO_OUTPUT, "control_log_rewards.txt")
if not os.path.exists(self.f_log_rewards):
f = open(self.f_log_rewards, 'w')
list_reward_keys = np.sort(self.para_set.LIST_TRUE_REWARD)
head_str = "node_id, current_time, action, " + ', '.join(list_reward_keys) + '\n'
f.write(head_str)
f.close()
if not os.path.exists(self.f_log_rewards_control):
f = open(self.f_log_rewards_control, 'w')
head_str = "node_id, current_time, local_travel_time, average_local_travel_time" + '\n'
f.write(head_str)
f.close()
def end_sumo(self, episode, current_time, file_name_travel_time, episode_time):
map_computor.end_sumo(self.eng, episode, current_time, file_name_travel_time, episode_time)
def end_sumo_test(self, episode, current_time, file_name_travel_time, episode_time):
map_computor.end_sumo_test(self.eng, episode, current_time, file_name_travel_time, episode_time)
def load_conf(self, conf_file):
dic_paras = json.load(open(conf_file, "r"))
return self.ParaSet(dic_paras)
def get_observation(self):
return self.state
def get_current_time(self):
return map_computor.get_current_time(self.eng)
def get_current_phase(self):
return self.current_phase
def take_action(self, joint_action, p_indicator, warm_up):
rewards_detail_dict_list =collections.OrderedDict()
for node_id in joint_action.keys():
current_phase_number = self.get_current_phase()[node_id]
if self.current_phase[node_id] <= 2: # 0, 2: straight
if (self.current_phase_duration[node_id] < self.para_set.MIN_PHASE_TIME):
joint_action[node_id] = 0
elif self.current_phase_duration[node_id] >= self.para_set.MAX_PHASE_TIME_STRAIGHT:
joint_action[node_id] = 1
else: # 1, 3: left
if (self.current_phase_duration[node_id] < self.para_set.MIN_PHASE_TIME):
joint_action[node_id] = 0
elif self.current_phase_duration[node_id] >= self.para_set.MAX_PHASE_TIME_LEFT:
joint_action[node_id] = 1
rewards_detail_dict_list[node_id] = []
for i in range(self.para_set.MIN_ACTION_TIME):
joint_action_in_second = collections.OrderedDict()
for node_id in joint_action.keys():
joint_action_in_second[node_id] = 0
current_phase_number = self.get_current_phase()[node_id]
if joint_action[node_id] == 1 and i == 0:
joint_action_in_second[node_id] = 1
self.current_phase, self.current_phase_duration, self.global_dic_vehicles, self.global_dic_location_vehicles, self.global_dic_this_node_vehicles, self.all_vehicles_location_enter_time_dict, self.all_vehicles_this_node_enter_time_dict, self.global_dic_waiting_time_vehicles, self.global_list_node_lane_vehicle\
= map_computor.run(eng=self.eng,
joint_action=joint_action_in_second,
current_phase=self.get_current_phase(),
current_phase_duration=self.current_phase_duration,
global_vehicle_dict=self.global_dic_vehicles,
global_dic_location_vehicles=self.global_dic_location_vehicles,
global_dic_this_node_vehicles=self.global_dic_this_node_vehicles,
all_vehicles_location_enter_time_dict=self.all_vehicles_location_enter_time_dict,
all_vehicles_this_node_enter_time_dict=self.all_vehicles_this_node_enter_time_dict,
global_dic_waiting_time_vehicles=self.global_dic_waiting_time_vehicles,
global_list_node_lane_vehicle=self.global_list_node_lane_vehicle,
rewards_info_dict=self.para_set.REWARDS_INFO_DICT,
true_reward=self.para_set.LIST_TRUE_REWARD,
f_log_rewards=self.f_log_rewards,
rewards_detail_dict_list=rewards_detail_dict_list,
reward_indicator=p_indicator,
warm_up=warm_up) # run 1s SUMO
#reward, reward_detail_dict = self.cal_reward(action)
global_reward = self.cal_reward_from_list(rewards_detail_dict_list) # each node: reward of step0 + ... + reward of step5
#self.update_vehicles()
self.update_state()
return global_reward, joint_action
def get_control_reward(self):
rewards_detail_dict_list = collections.OrderedDict()
rewards_detail_this_node_dict_list = collections.OrderedDict()
for node_id_1 in map_computor.get_node_id_list():
rewards_detail_dict_list[node_id_1] = []
rewards_detail_this_node_dict_list[node_id_1] = []
map_computor.run_control(eng=self.eng,
rewards_info_dict=self.para_set.REWARDS_CONTROL_INFO_DICT,
rewards_this_node_info_dict=self.para_set.REWARDS_CONTROL_AUX_INFO_DICT,
f_log_rewards_control=self.f_log_rewards_control,
rewards_detail_dict_list=rewards_detail_dict_list,
rewards_detail_this_node_dict_list=rewards_detail_this_node_dict_list,
all_vehicles_location_enter_time_dict=self.all_vehicles_location_enter_time_dict,
all_vehicles_this_node_enter_time_dict=self.all_vehicles_this_node_enter_time_dict)
self.clear_local_travel_time()
global_reward_aux = self.cal_reward_from_list(rewards_detail_dict_list)
global_reward_main = self.cal_reward_from_list(rewards_detail_this_node_dict_list)
return global_reward_aux, global_reward_main
def take_action_pre_train(self, phase_time_now):
rewards_detail_dict_list = collections.OrderedDict()
joint_action = collections.OrderedDict()
for i in range(4):
for j in range(4):
node_id = "node%d%d" % (i, j)
current_phase_number = self.get_current_phase()[node_id]
if (self.current_phase_duration < phase_time_now[current_phase_number]):
joint_action[node_id] = 0
else:
joint_action[node_id] = 1
rewards_detail_dict_list[node_id] = []
for i in range(self.para_set.MIN_ACTION_TIME):
joint_action_in_second = collections.OrderedDict()
for node_id in joint_action.keys():
joint_action_in_second[node_id] = 0
current_phase_number = self.get_current_phase()[node_id]
if joint_action[node_id] == 1 and i == 0:
joint_action_in_second[node_id] = 1
self.current_phase, self.current_phase_duration, self.global_dic_vehicles = map_computor.run(joint_action=joint_action_in_second,
current_phase=self.get_current_phase(),
current_phase_duration=self.current_phase_duration,
global_vehicle_dict=self.global_dic_vehicles,
rewards_info_dict=self.para_set.REWARDS_INFO_DICT,
f_log_rewards=self.f_log_rewards,
rewards_detail_dict_list=rewards_detail_dict_list) # run 1s SUMO
global_reward = self.cal_reward_from_list(rewards_detail_dict_list)
#self.update_vehicles()
self.update_state()
return global_reward, joint_action
def update_vehicles(self):
self.global_dic_vehicles = map_computor.update_vehicles_state(self.eng, self.global_dic_vehicles)
def update_vehicles_location(self):
self.global_dic_location_vehicles, self.global_dic_this_node_vehicles, self.all_vehicles_location_enter_time_dict, self.all_vehicles_this_node_enter_time_dict, self.global_dic_waiting_time_vehicles \
= map_computor.update_vehicles_location(self.eng, self.global_dic_location_vehicles, self.global_dic_this_node_vehicles, self.all_vehicles_location_enter_time_dict, self.all_vehicles_this_node_enter_time_dict, self.global_dic_waiting_time_vehicles)
def update_vehicle_arrive_leave_time(self):
self.global_list_node_lane_vehicle = map_computor.update_dic_lane_vehicle_arrive_leave_time(self.eng, self.global_list_node_lane_vehicle)
def clear_local_travel_time(self):
self.all_vehicles_location_enter_time_dict = collections.OrderedDict()
self.all_vehicles_this_node_enter_time_dict = collections.OrderedDict()
for node_id_1 in map_computor.get_node_id_list():
self.all_vehicles_location_enter_time_dict[node_id_1] = collections.OrderedDict()
self.all_vehicles_this_node_enter_time_dict[node_id_1] = collections.OrderedDict()
def update_state(self):
status_trackers = map_computor.status_calculator(self.eng, self.global_dic_vehicles)
DIC_PHASE_MAP = {
1: 2,
2: 3,
3: 4,
4: 1
}
for node_id, status_tracker in status_trackers.items():
self.state[node_id] = State(
queue_length=None,
num_of_vehicles=np.reshape(np.array(status_tracker[1]), newshape=(1,) + State.D_NUM_OF_VEHICLES),
waiting_time=None,
map_feature=None,
cur_phase=np.eye(State.D_CUR_PHASE[0], dtype=np.int16)[[self.current_phase[node_id]]], # one_hot
next_phase=np.eye(State.D_NEXT_PHASE[0], dtype=np.int16)[[DIC_PHASE_MAP[self.current_phase[node_id]]]], # one_hot
time_this_phase=None,
if_terminal=None
)
'''
self.state[node_id] = State(
queue_length=np.reshape(np.array(status_tracker[0]), newshape=(1,) + State.D_QUEUE_LENGTH),
num_of_vehicles=np.reshape(np.array(status_tracker[1]), newshape=(1,) + State.D_NUM_OF_VEHICLES),
waiting_time=np.reshape(np.array(status_tracker[2]), newshape=(1,) + State.D_WAITING_TIME),
map_feature=None,
cur_phase=np.eye(State.D_CUR_PHASE[0], dtype=np.int16)[[self.current_phase[node_id]]], # one_hot
next_phase=np.eye(State.D_NEXT_PHASE[0], dtype=np.int16)[[(self.current_phase[node_id] + 1) % (len(map_computor.get_node_phases(node_id)))]], # one_hot
time_this_phase=np.reshape(np.array([self.current_phase_duration[node_id]]), newshape=(1,) + State.D_TIME_THIS_PHASE),
if_terminal=False
)
'''
# it looks useless
def cal_reward(self, action):
# get directly from sumo
reward, reward_detail_dict = map_computor.get_rewards_from_sumo(self.global_dic_vehicles, action, self.para_set.REWARDS_INFO_DICT)
return reward*(1-0.8), reward_detail_dict
def cal_reward_from_list(self, global_reward_detail_dict_list):
global_reward = collections.OrderedDict()
for node_id, reward_detail_dict_list in global_reward_detail_dict_list.items():
reward = map_computor.get_rewards_from_dict_list(reward_detail_dict_list)
global_reward[node_id] = reward*(1-0.8)
return global_reward
if __name__ == '__main__':
pass