-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathevaluate_model.py
474 lines (427 loc) · 21.3 KB
/
evaluate_model.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
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
import gymnasium as gym
import numpy as np
from sb3_contrib.common.maskable.policies import MaskableActorCriticPolicy
from sb3_contrib.common.wrappers import ActionMasker
from sb3_contrib.ppo_mask import MaskablePPO
from stable_baselines3 import PPO
from fullcostenv import FullCostEnv
from constants import *
from utils import *
import csv
import argparse
import sys
import random
import re
import os
# Inspect transition metadata to determine interaction phase type (setup or simulation)
def is_sim_type(transition):
for obj in transition["metaData"]:
if obj["type"] == "simulation":
return "Simulation"
return "Setup"
# Inspect transition metadata to determine ergonomic and monetary costs
def find_costs(transition):
costs = [0] * 5
for obj in transition["metaData"]:
if obj["type"] == "ergoHand":
costs[0] += obj["value"][1]
if obj["type"] == "ergoArm":
costs[1] += obj["value"][1]
if obj["type"] == "ergoShoulder":
costs[2] += obj["value"][1]
if obj["type"] == "ergoWholeBody":
costs[3] += obj["value"][1]
for obj in transition["cost"]:
if obj["category"] == "monetary":
costs[4] += obj["value"]
return costs
# Masking function, call the environment's masking function
def mask_fn(env: gym.Env) -> np.ndarray:
return env.valid_action_mask()
def run(arguments):
# Determine which input json file to use
f = FILENAME # default "cost_net.json"
if arguments.input_file is not None:
f = arguments.input_file
# Determine naming scheme for model output
outputFilename = f.replace(".json", "") + "-output.csv"
if arguments.output is not None:
outputFilename = arguments.output
if "/" not in f:
primitiveOutputFilename = "primitive-" + f.replace(".json", "") + "-output.csv"
else:
folder_path = f.split("/")
folders = "/".join(folder_path[:len(folder_path)-1])
primitiveOutputFilename = folders + "/primitive-" + folder_path[len(folder_path)-1].replace(".json", "") + "-output.csv"
if arguments.output is not None:
if "/" not in arguments.output:
primitiveOutputFilename = "primitive-" + arguments.output
else:
folder_path = arguments.output.split("/")
folders = "/".join(folder_path[:len(folder_path)-1])
primitiveOutputFilename = folders + "/primitive-" + folder_path[len(folder_path)-1].replace(".json", "") + "-output.csv"
# Load petrinet data from json (transitions, places)
[json_obj, weights, json_task, targets_obj, primitives_obj, json_agents] = LOAD_JOB_FILE(f)
# Setup evaluation environment
env = FullCostEnv(json_obj, weights, json_task, targets_obj, json_agents)
# Reset and mask environment
env.reset(0, {})
env = ActionMasker(env, mask_fn) # Wrap to enable masking
# Setup tracking
best_action_sequence = []
best_reward = -sys.maxsize - 1
# Load model for evaluation
model = None
model_files = []
if arguments.model is not None:
model_files = [arguments.model]
elif arguments.model_str is not None and arguments.model_dir is not None:
input_split = arguments.input_file.split("/")
escaped_pattern = re.escape(arguments.model_str).replace("FF", input_split[len(input_split)-1].replace(".json", "")).replace("XX", r"\d+")
pattern = re.compile(f"^{escaped_pattern}$")
model_files = ["/".join([arguments.model_dir, f]) for f in os.listdir(arguments.model_dir) if pattern.match(f)]
else:
print("ERROR: Unable to load model file(s).")
exit(1)
for model_file in model_files:
print("Currently on: " + model_file)
model = MaskablePPO.load(model_file, weights_only=True)
# Reset environment and get initial observation
obs, _info = env.reset()
loop_count = 0
keep_looping = True
while keep_looping:
cummulative_reward = 0.0
iteration = 0
action_sequence = []
done = False
# Iterate until model finds the goal(s) or max timesteps are hit
while not done:
# Increase iteration count
iteration += 1
# Determine best action from current state
action, _states = model.predict(obs, action_masks=mask_fn(env))
# Step the model based on selected action
obs, rewards, done, shortcut, info = env.step(action)
# Update rewards
cummulative_reward += rewards
# Add selected action to sequence
action_sequence.append((env.transition_ids[action], action, rewards, info))
# If goal(s) are met or max timesteps are reached, mark as done and print
if iteration >= arguments.n_steps or done:
done = True
loop_count += 1
obs, _info = env.reset()
if len(best_action_sequence) == 0 or cummulative_reward > best_reward:
best_action_sequence = action_sequence.copy()
best_reward = cummulative_reward
if loop_count > arguments.n_samples:
keep_looping = False
action_sequence = best_action_sequence.copy()
print(len(action_sequence), best_reward)
currentTime = 0
# Write output to CSV
with open(outputFilename, "w+", newline='') as fh:
csv_writer = csv.writer(fh)
csv_writer.writerow(["Action",
"Type",
"Reward",
"Agent Assigned To Task",
"Start Time (s)",
"End Time (s)",
"From Standing Location",
"To Standing Location",
"From Hand Location",
"To Hand Location",
"Target",
"Weight",
"Force",
"Hand Cost",
"Arm Cost",
"Shoulder Cost",
"Whole Body Cost",
"Monetary Cost",
"Primitives",
"MVC",
"Vertical Hand Distance Traveled",
"Horizontal Hand Distance Traveled",
"Stand Distance Traveled",
"Reach Distance",
"Hand To Floor Distance",
"Is One Handed",
"Is Hand Work"])
for (transition_id, action, reward, info) in action_sequence:
transition = json_obj["transitions"][transition_id]
costs = find_costs(transition)
duration = transition["time"]
agents = ";".join(info["busyAgents"])
primtives = ""
mvcs = ""
isOneHanded = ""
isHandWork = ""
standDistanceTraveled = ""
horizontalHandDistanceTraveled = ""
verticalHandDistanceTraveled = ""
reachDistance = ""
distanceFromHandToFloor = ""
a = 0
fromHandLocation = ""
toHandLocation = ""
fromStandLocation = ""
toStandLocation = ""
targets = ""
weights = ""
forces = ""
# Iterate over transition metadata to extract information for CSV output
for m in transition["metaData"]:
if m["type"] == "primitiveAssignment":
try:
primitive = json_obj["nameLookup"][m["value"][1]]
primtives += primitive + ";"
if (primitive in ["Force", "Use", "Inspect", "Selection", "Hold", "Position"]):
some_prim_obj = primitives_obj[m["value"][1]]
targets += targets_obj[some_prim_obj["target"]]["name"] + ";"
weights += str(targets_obj[some_prim_obj["target"]]["weight"]) + ";"
except:
a = 1
elif m["type"] == "isOneHanded":
isOneHanded += "True" if m["value"][1] else "False" + ";"
elif m["type"] == "isHandWork":
isHandWork += "True" if m["value"][1] else "False" + ";"
elif m["type"] == "mVC":
mvcs += str(m["value"][1]) + ";"
elif m["type"] == "horizontalHandTravelDistance":
horizontalHandDistanceTraveled += str(m["value"][1]) + ";"
elif m["type"] == "verticalHandTravelDistance":
verticalHandDistanceTraveled += str(m["value"][1]) + ";"
elif m["type"] == "standTravelDistance":
standDistanceTraveled += str(m["value"][1]) + ";"
elif m["type"] == "reachDistance":
reachDistance += str(m["value"][1]) + ";"
elif m["type"] == "handDistanceToFloor":
distanceFromHandToFloor += str(m["value"][1]) + ";"
elif m["type"] == "standing":
fromStandLocation = json_obj["nameLookup"][m["value"][0]]
toStandLocation = fromStandLocation
elif m["type"] == "hand":
fromHandLocation = json_obj["nameLookup"][m["value"][0]]
toHandLocation = fromHandLocation
elif m["type"] == "fromHandPOI":
fromHandLocation = json_obj["nameLookup"][m["value"][0]]
elif m["type"] == "toHandPOI":
toHandLocation = json_obj["nameLookup"][m["value"][0]]
elif m["type"] == "fromStandingPOI":
fromStandLocation = json_obj["nameLookup"][m["value"][0]]
elif m["type"] == "toStandingPOI":
toStandLocation = json_obj["nameLookup"][m["value"][0]]
elif m["type"] == "target":
targets += targets_obj[m["value"]]["name"] + ";"
weights += str(targets_obj[m["value"]]["weight"]) + ";"
elif m["type"] == "force":
forces += str(m["value"][1]) + ";"
# TODO: keep track of time, busy/working agents, who all is assigned to a task (multiple agent split)
csv_writer.writerow([env.transition_names[action],
is_sim_type(transition),
reward,
agents,
info["startTime"],
info["endTime"],
fromStandLocation,
toStandLocation,
fromHandLocation,
toHandLocation,
targets,
weights,
forces,
costs[0],
costs[1],
costs[2],
costs[3],
costs[4],
primtives,
mvcs,
verticalHandDistanceTraveled,
horizontalHandDistanceTraveled,
standDistanceTraveled,
reachDistance,
distanceFromHandToFloor,
isOneHanded,
isHandWork
])
currentTime = info["endTime"]
currentTime = 0
uuid = 0
# Write output to primitives CSV
with open(primitiveOutputFilename, "w+", newline='') as fh:
csv_writer = csv.writer(fh)
csv_writer.writerow(["Unique ID",
"Action",
"Primitive",
"Type",
"Reward",
"Agent Assigned To Task",
"Start Time (s)",
"End Time (s)",
"From Standing Location",
"To Standing Location",
"From Hand Location",
"To Hand Location",
"Target",
"Weight",
"Force",
"Hand Cost",
"Arm Cost",
"Shoulder Cost",
"Whole Body Cost",
"Monetary Cost",
"MVC",
"Vertical Hand Distance Traveled",
"Horizontal Hand Distance Traveled",
"Stand Distance Traveled",
"Reach Distance",
"Hand To Floor Distance",
"Is One Handed",
"Is Hand Work"])
for (transition_id, action, reward, info) in action_sequence:
uuid += 1
transition = json_obj["transitions"][transition_id]
costs = find_costs(transition)
duration = transition["time"]
agents = ";".join(info["busyAgents"])
primtives = ""
mvcs = ""
isOneHanded = ""
isHandWork = ""
standDistanceTraveled = ""
horizontalHandDistanceTraveled = ""
verticalHandDistanceTraveled = ""
reachDistance = ""
distanceFromHandToFloor = ""
a = 0
fromHandLocation = ""
toHandLocation = ""
fromStandLocation = ""
toStandLocation = ""
target = ""
weight = ""
force = ""
primitive_dictionary = {}
# iterate once to get all primitives
for m in transition["metaData"]:
if m["type"] == "primitiveAssignment":
try:
primitive_dictionary[m["value"][1]] = ["" for _ in range(28)]
primitive_dictionary[m["value"][1]][0] = uuid
primitive_dictionary[m["value"][1]][1] = env.transition_names[action]
primitive = json_obj["nameLookup"][m["value"][1]]
primitive_dictionary[m["value"][1]][2] = primitive
primitive_dictionary[m["value"][1]][3] = is_sim_type(transition)
primitive_dictionary[m["value"][1]][4] = reward
# ... left blank for agent (for now)
primitive_dictionary[m["value"][1]][5] = json_agents[m["value"][0]]["name"]
primitive_dictionary[m["value"][1]][6] = info["startTime"]
primitive_dictionary[m["value"][1]][7] = info["endTime"]
# ....
if (primitive in ["Force", "Use", "Inspect", "Selection", "Hold", "Position"]):
try:
some_prim_obj = primitives_obj[m["value"][1]]
primitive_dictionary[m["value"][1]][12] = targets_obj[some_prim_obj["target"]]["name"]
primitive_dictionary[m["value"][1]][13] = targets_obj[some_prim_obj["target"]]["weight"]
except:
a = 1
# ....
primitive_dictionary[m["value"][1]][15] = costs[0]
primitive_dictionary[m["value"][1]][16] = costs[1]
primitive_dictionary[m["value"][1]][17] = costs[2]
primitive_dictionary[m["value"][1]][18] = costs[3]
primitive_dictionary[m["value"][1]][19] = costs[4]
except:
a = 1
# Iterate over transition metadata to extract information for CSV output
for m in transition["metaData"]:
if m["type"] == "mVC":
primitive_dictionary[m["value"][0]][20] = str(m["value"][1])
elif m["type"] == "verticalHandTravelDistance":
primitive_dictionary[m["value"][0]][21] = str(m["value"][1])
elif m["type"] == "horizontalHandTravelDistance":
primitive_dictionary[m["value"][0]][22] = str(m["value"][1])
elif m["type"] == "standTravelDistance":
primitive_dictionary[m["value"][0]][23] = str(m["value"][1])
elif m["type"] == "reachDistance":
primitive_dictionary[m["value"][0]][24] = str(m["value"][1])
elif m["type"] == "handDistanceToFloor":
primitive_dictionary[m["value"][0]][25] = str(m["value"][1])
if m["type"] == "isOneHanded":
primitive_dictionary[m["value"][0]][26] = "True" if m["value"][1] else "False"
elif m["type"] == "isHandWork":
primitive_dictionary[m["value"][0]][27] = "True" if m["value"][1] else "False"
elif m["type"] == "standing":
fromStandLocation = json_obj["nameLookup"][m["value"][0]]
toStandLocation = fromStandLocation
elif m["type"] == "hand":
fromHandLocation = json_obj["nameLookup"][m["value"][0]]
toHandLocation = fromHandLocation
elif m["type"] == "fromHandPOI":
fromHandLocation = json_obj["nameLookup"][m["value"][0]]
elif m["type"] == "toHandPOI":
toHandLocation = json_obj["nameLookup"][m["value"][0]]
elif m["type"] == "fromStandingPOI":
fromStandLocation = json_obj["nameLookup"][m["value"][0]]
elif m["type"] == "toStandingPOI":
toStandLocation = json_obj["nameLookup"][m["value"][0]]
elif m["type"] == "target":
target = targets_obj[m["value"]]["name"]
weight = targets_obj[m["value"]]["weight"]
elif m["type"] == "force":
primitive_dictionary[m["value"][0]][14] = str(m["value"][1])
if len(primitive_dictionary.keys()) == 0:
primitive_dictionary[""] = [uuid,
env.transition_names[action],
"",
is_sim_type(transition),
reward,
agents,
info["startTime"],
info["endTime"],
fromStandLocation,
toStandLocation,
fromHandLocation,
toHandLocation,
target,
weight,
"",
"",
"",
"",
"",
"",
"",
"",
"",
"",
"",
"",
"",
"",
]
for key in primitive_dictionary:
primitive_dictionary[key][8] = fromStandLocation
primitive_dictionary[key][9] = toStandLocation
primitive_dictionary[key][10] = fromHandLocation
primitive_dictionary[key][11] = toHandLocation
if primitive_dictionary[key][12] == "":
primitive_dictionary[key][12] = target
primitive_dictionary[key][13] = weight
csv_writer.writerow(primitive_dictionary[key])
currentTime = info["endTime"]
if __name__ == "__main__":
parser = argparse.ArgumentParser()
parser.add_argument("--input-file", type=str, default=None, help="")
parser.add_argument("--model", type=str, default=None, help="")
parser.add_argument("--model-str", type=str, default=None, help="")
parser.add_argument("--model-dir", type=str, default=None, help="")
parser.add_argument("--output", type=str, default=None, help="")
parser.add_argument("--n-steps", type=int, default=MAX_TESTING_TIMESTEPS, help="")
parser.add_argument("--n-samples", type=int, default=0, help="")
args = parser.parse_args()
run(args)