-
Notifications
You must be signed in to change notification settings - Fork 4
/
MPC.py
386 lines (338 loc) · 16.9 KB
/
MPC.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
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
import numpy as np
from message import Message
from data_loader import DataLoader
from group_shape_prediction import GroupShapePrediction
import general_helpers as gh
import grouping_helpers as gp_h
import MPC_helpers as mpc_h
import matplotlib.animation as animation
import matplotlib.pyplot as plt
from time import time
from copy import deepcopy
import sys
sys.path.append('sgan')
from scripts.inference import SGANInference as sgan
import rvo2
import pickle
def sim_step(robo_curr, robo_goal, dt, robo_max_v, t_horizon, ped_pos, ped_vel, ped_goals):
#RVOSimulator (float timeStep, float neighborDist, size_t maxNeighbors, float timeHorizon, float timeHorizonObst, float radius, float maxSpeed)
ped_max_spd = 1.75
sim = rvo2.PyRVOSimulator(dt, 2.5, 10, t_horizon, 2, 0.5, ped_max_spd)
robot = sim.addAgent((robo_curr[0], robo_curr[1]))
sim.setAgentMaxSpeed(robot, robo_max_v)
sim.setAgentPrefVelocity(robot, gh.get_pref_velocity(robo_curr, robo_goal, robo_max_v))
num_ped = len(ped_pos)
ped_list = []
for i in range(num_ped):
ped = sim.addAgent((ped_pos[i][0], ped_pos[i][1]))
ped_spd = np.linalg.norm(np.array(ped_vel[i]))
sim.setAgentVelocity(ped, (ped_vel[i][0], ped_vel[i][1]))
sim.setAgentMaxSpeed(ped, max(ped_spd, ped_max_spd))
sim.setAgentPrefVelocity(ped, gh.get_pref_velocity(ped_pos[i], ped_goals[i], ped_spd))
ped_list.append(ped)
sim.doStep()
ped_pos_new = np.array([sim.getAgentPosition(ped) for ped in ped_list])
ped_vel_new = np.array([sim.getAgentVelocity(ped) for ped in ped_list])
return ped_pos_new, ped_vel_new
def run_trial(dataset, dataset_idx, init_start_config, init_end_config, start_frame, time_limit,
args, mpc_w):
group_flag, laser_flag, pred_flag, react_flag, pred_method = args
# Meta debug flags
show_plot_flag = False
include_when_fail = False
# rollout params
num_rollouts = 12
time_steps = 8
dt = 0.1
tf_horizon = dt*time_steps
v = 1.75
time_horizon = np.arange(0, tf_horizon + dt, dt)
collision_radius = 0.5
final_thresh = 1.0
const_init = 0.35 # group shape size
if (dataset == "ucy") and (dataset_idx == 2):
const_init = 0.25
const_min = 0.05
num_eval = 1
if pred_method == "sgan":
dt_name, _ = gh.convert_dataset_id(dataset, dataset_idx)
path = "sgan/models/sgan-models/" + dt_name + "_" + str(time_steps) + "_model.pt"
model = sgan(path)
raw_msg = Message()
data = DataLoader(dataset, dataset_idx, 1 / dt)
raw_msg = data.update_message(raw_msg)
time_limit = time_limit + start_frame
# Metrics related
end_status = []
min_ped_distances = []
time_duration = []
path_lengths = []
path_irregularities = []
all_robo_paths = []
for test_num in range(num_eval):
#print('Test number #{}'.format(test_num+1))
time_out = True
min_ped_dist = 10000
msg = deepcopy(raw_msg)
const = const_init
if (group_flag) and (pred_flag) and (pred_method == "auto"):
_, dt_num = gh.convert_dataset_id(dataset, dataset_idx)
path = "checkpoints/model_fpsfix_" + str(dt_num) + ".pth"
gsp = GroupShapePrediction(msg, path)
start_config = init_start_config
end_config = init_end_config
robot_path = [start_config]
if react_flag:
ped_pos_current = []
ped_vel_current = []
ped_idxes_current = []
ped_idxes = []
ped_goals_current = []
for frame in range(start_frame, time_limit):
print([frame, time_limit], end='\r')
if (gh.at_goal(start_config, end_config, final_thresh) == True):
time_out = False
end_status.append(0) # 0 = success
min_ped_distances.append(min_ped_dist)
time_duration.append((frame - start_frame) * dt)
path_lengths.append(gh.estimate_path_length(robot_path))
path_irregularities.append(gh.estimate_path_irregularity(robot_path))
all_robo_paths.append(robot_path)
break
if frame >= msg.total_num_frames:
ped_pos_current = []
ped_vel_current = []
elif react_flag:
# remove any pedestrian who reaches goal
ped_pos_new = []
ped_vel_new = []
ped_idxes_new = []
ped_goals_new = []
for i in range(len(ped_pos_current)):
if (not gh.at_goal(ped_pos_current[i], ped_goals_current[i], final_thresh)):
ped_pos_new.append(ped_pos_current[i])
ped_vel_new.append(ped_vel_current[i])
ped_idxes_new.append(ped_idxes_current[i])
ped_goals_new.append(ped_goals_current[i])
# check for new pedestrians
new_idxes_current = msg.video_pedidx_matrix[frame]
for i, idx in enumerate(new_idxes_current):
if (not (idx in ped_idxes)):
ped_idxes.append(idx)
ped_pos_new.append(msg.video_position_matrix[frame][i])
ped_vel_new.append(msg.video_velocity_matrix[frame][i])
ped_idxes_new.append(idx)
ped_goals_new.append(msg.people_coords_complete[idx][-1])
ped_pos_current = np.array(ped_pos_new,dtype=np.float32)
ped_vel_current = np.array(ped_vel_new,dtype=np.float32)
ped_idxes_current = ped_idxes_new
ped_goals_current = np.array(ped_goals_new,dtype=np.float32)
# Modify msg
msg.video_position_matrix[frame] = ped_pos_current
msg.video_velocity_matrix[frame] = ped_vel_current
msg.video_pedidx_matrix[frame] = ped_idxes_current
for i, idx in enumerate(ped_idxes_current):
st_frame = msg.people_start_frame[idx]
end_frame = msg.people_end_frame[idx]
if (frame > end_frame):
msg.people_coords_complete[idx].append(ped_pos_current[i])
msg.people_velocity_complete[idx].append(ped_vel_current[i])
msg.people_end_frame[idx] = frame
else:
msg.people_coords_complete[idx][frame - st_frame] = ped_pos_current[i]
msg.people_velocity_complete[idx][frame - st_frame] = ped_vel_current[i]
else:
ped_pos_current = np.array(msg.video_position_matrix[frame],dtype=np.float32)
ped_vel_current = np.array(msg.video_velocity_matrix[frame],dtype=np.float32)
num_ppl = len(ped_pos_current)
rollouts = mpc_h.generate_rollouts(start_config, time_horizon, num_rollouts, v)
group_frames_collection = None
if(num_ppl == 0):
has_ped = False
obstacle_points = []
obstacle_vels = []
ped_pos = []
ped_pos_current = []
ped_pos_alt = []
ped_vel_alt = []
groups_vertices = []
groups_vertices_current = []
else:
min_ped_dist_curr = gh.get_min_ped_dist(ped_pos_current, start_config)
if (min_ped_dist_curr < min_ped_dist):
min_ped_dist = min_ped_dist_curr
if (min_ped_dist < collision_radius):
# print("Collision took place!!!")
time_out = False
end_status.append(1) # 1 = collision
if include_when_fail:
min_ped_distances.append(min_ped_dist)
time_duration.append((frame - start_frame) * dt)
path_lengths.append(gh.estimate_path_length(robot_path))
path_irregularities.append(gh.estimate_path_irregularity(robot_path))
all_robo_paths.append(robot_path)
break
ped_pos_alt = ped_pos_current
ped_vel_alt = ped_vel_current
if group_flag:
# Shrink group space if already inside group space
const = const_init
if laser_flag:
ped_pos_alt, ped_vel_alt = gh.ped_to_scans(start_config,
ped_pos_current,
ped_vel_current)
groups_frame_current = gh.get_frame_groups(msg,
ped_pos_alt,
ped_vel_alt,
laser_flag,
const)
while mpc_h.check_inside_groups(msg, start_config, groups_frame_current):
const = max(const - 0.1, const_min)
groups_frame_current = gh.get_frame_groups(msg,
ped_pos_alt,
ped_vel_alt,
laser_flag,
const)
if const == const_min:
break
groups_vertices_current = gh.frame_to_vertices(msg, groups_frame_current)
group_frames_collection = [groups_frame_current]
if(pred_flag):
groups_vertices = [groups_vertices_current]
if(pred_method=="auto"):
inputs_pos, inputs_vel = gh.create_model_input(msg,frame,dt=dt,
length=time_steps)
if (laser_flag):
inputs_pos, inputs_vel = gh.ped_series_to_scans(start_config,
inputs_pos,
inputs_vel)
groups_frames_predicted = gsp.laser_predict(inputs_pos,
inputs_vel,
const)
else:
groups_frames_predicted = gsp.predict(inputs_pos,
inputs_vel,
const)
for pred_frame in groups_frames_predicted:
group_frames_collection.append(pred_frame)
groups_vertices.append(gh.frame_to_vertices(msg, pred_frame))
else:
raise Exception('Undefined prediction method!')
else:
groups_vertices = groups_vertices_current
obstacle_points = groups_vertices
obstacle_vels = []
else:
# Shrink personal space if already inside personal space
const = const_init
while mpc_h.check_inside_PS(start_config, ped_pos_current, ped_vel_current,
const):
const = max(const - 0.1, const_min)
if const == const_min:
break
groups_vertices=None
groups_vertices_current=None
if(pred_flag):
if(pred_method=="sgan"): # PREDICTION METHOD
input_to_sgan, _ = gh.create_model_input(msg,frame,dt=dt,
length=time_steps)
ped_pos_predicted = model.evaluate(input_to_sgan)
ped_pos = gh.combine_current_and_predicted_pos(ped_pos_current,
ped_pos_predicted)
elif(pred_method=="linear"):
ped_pos = gh.propagate_peds(msg,frame,dt,tf_horizon)
obstacle_points = ped_pos
ped_vel_predicted = (ped_pos[:, 1:, :] - ped_pos[:, :-1, :]) / dt
obstacle_vels = np.concatenate((np.expand_dims(ped_vel_current, 1),
ped_vel_predicted), axis=1)
else:
ped_pos = ped_pos_current
obstacle_points = ped_pos_current
obstacle_vels = ped_vel_current
has_ped = True
costs = mpc_h.evaluate_rollouts(msg,
start_config,
rollouts,
obstacle_points,
obstacle_vels,
group_frames_collection,
collision_radius,
end_config,
mpc_w,
const,
groups=group_flag,
pred=pred_flag,
has_obstacles=has_ped)
lowest_cost_ind = np.argmin(costs)
start_config = rollouts[lowest_cost_ind][1]
robot_path.append(start_config)
if react_flag:
ped_pos_current, ped_vel_current = sim_step(start_config, end_config,
dt, v, tf_horizon,
ped_pos_current, ped_vel_current,
ped_goals_current)
if (time_out == True):
end_status.append(2) # 2 = timeout
if include_when_fail:
min_ped_distances.append(min_ped_dist)
time_duration.append((frame - start_frame) * dt)
path_lengths.append(gh.estimate_path_length(robot_path))
path_irregularities.append(gh.estimate_path_irregularity(robot_path))
all_robo_paths.append(robot_path)
success = 0
for st in end_status:
if st == 0:
success += 1
success_rate = success / len(end_status) * 100
if success == 0:
return (0, 0, 0, 0, 0, all_robo_paths)
else:
min_ped_distance = np.min(np.array(min_ped_distances))
avg_duration = np.mean(np.array(time_duration))
avg_path_length = np.mean(path_lengths)
avg_path_irregularity = np.mean(path_irregularities)
return (success_rate, min_ped_distance, avg_duration,
avg_path_length, avg_path_irregularity, all_robo_paths)
if __name__ == "__main__":
args = gh.get_flags()
group_flag, laser_flag, pred_flag, react_flag, pred_method = args
pfile_name = "test_cases/all.txt"
with open(pfile_name, "rb") as fp:
cases = pickle.load(fp)
cases = [cases[0]]
if group_flag:
exp_name = "group_" + pred_method
else:
exp_name = "ped_" + pred_method
if laser_flag:
exp_name += "_laser"
if react_flag:
w = 0.3
exp_name += "_react"
else:
w = 0.65
fname = "results/" + exp_name + ".txt"
num_cases = len(cases)
all_results = []
for i, case in enumerate(cases):
print([i, num_cases])
dataset = case[0]
dataset_idx = case[1]
init_start_config = case[2]
init_end_config = case[3]
start_frame = case[4]
time_limit = case[5]
start_time = time()
metrics = run_trial(dataset,
dataset_idx,
init_start_config,
init_end_config,
start_frame,
time_limit,
args,
w)
end_time = time()
print("Time taken for the past trial: ", end_time - start_time)
all_results.append((case, metrics))
with open(fname, "wb") as f:
pickle.dump(all_results, f)