-
Notifications
You must be signed in to change notification settings - Fork 0
/
ga_planner_dynamic.py
428 lines (363 loc) · 13.8 KB
/
ga_planner_dynamic.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
# In[]:
import random
import numpy as np
from math import pi
from operator import attrgetter
import matplotlib.pyplot as plt
import seaborn as sns
import multiprocessing as mp
from collision_checker import CollisionChecker
from geometry import World
from geometry import ObjectGenerator
from viewer import RealtimePlot
from prm import PRM
from cell_decomposition import CellDecomposition
import time
sns.set()
# np.random.seed(1)
class Individual(np.ndarray):
"""Container of a individual."""
fitness = None
def __new__(cls, a):
return np.asarray(a).view(cls)
class GeneticAlgorithm():
def __init__(self, world=None, NGEN=1000, n_ind=100, n_elite=10,
fitness_thresh=0.1, margin_on=False, verbose=True):
self.trajectory = [world.start]
self.history = []
self.pop = []
self.best_ind = []
self.gtot = 0
self.n_ind = n_ind # The number of individuals in a population.
self.n_elite = n_elite
self.NGEN = NGEN # The number of generation loop.
self.fitness_thresh = fitness_thresh
self.verbose = verbose
# --- Generate Collision Checker
self.world = world
if margin_on:
self.world_margin = world.mcopy(rate=1.05)
else:
self.world_margin = world
self.cc = CollisionChecker(world)
self.ccm = CollisionChecker(self.world_margin)
def create_pop_prm(self, m, n):
prm_list = [PRM(self.world_margin, 30, 3) for i in range(m)]
path_list = []
for prm in prm_list:
prm.single_query()
if prm.path_list != []:
prm.multi_query(n)
path_list += prm.path_list
return [Individual(path) for path in path_list]
def create_pop_cd(self):
cd = CellDecomposition(self.world_margin)
path_list = cd.main(self.n_ind, shortcut=True)
self.pop = [Individual(path) for path in path_list]
return self.pop
def set_fitness(self, eval_func):
"""Set fitnesses of each individual in a population."""
for i, fit in enumerate(map(eval_func, self.pop)):
self.pop[i].fitness = fit
def evalOneMax(self, gene):
"""Objective function."""
delta = [0.1]
score_list = []
for d in delta:
pts = self.smoothing(gene, d)
score = 0.0
for i in range(len(pts) - 1):
dist = np.linalg.norm(pts[i + 1] - pts[i])
score = score + dist
# if not cc.path_validation(pts[i], pts[i+1]):
# score = score + 10
if not self.ccm.path_validation(pts):
score = score + 10
score_list.append(score)
return min(score_list)
def eval_for_cx(self, gene):
"""Objective function."""
pts = gene.copy()
score = 0.0
for i in range(len(pts) - 1):
dist = np.linalg.norm(pts[i + 1] - pts[i])
score = score + dist
return score
def selTournament(self, tournsize):
"""Selection function."""
chosen = []
for i in range(self.n_ind - self.n_elite):
aspirants = [random.choice(self.pop) for j in range(tournsize)]
chosen.append(min(aspirants, key=attrgetter("fitness")))
return chosen
def selElite(self):
pop_sort = sorted(self.pop, key=attrgetter("fitness"))
elites = pop_sort[:self.n_elite]
return elites
def calc_intersection_point(self, A, B, C, D):
denominator = (B[0] - A[0]) * (C[1] - D[1]) - \
(B[1] - A[1]) * (C[0] - D[0])
# If two lines are parallel,
if abs(denominator) < 1e-6:
return None, None, None
AC = A - C
r = ((D[1] - C[1]) * AC[0] - (D[0] - C[0]) * AC[1]) / denominator
s = ((B[1] - A[1]) * AC[0] - (B[0] - A[0]) * AC[1]) / denominator
# If the intersection is out of the edges
if r < -1e-6 or r > 1.00001 or s < -1e-6 or s > 1.00001:
return None, r, s
# Endpoint and startpoint make the intersection.
if ((np.linalg.norm(r - 1.0) < 1e-6 and np.linalg.norm(s) < 1e-6) or
(np.linalg.norm(s - 1.0) < 1e-6 and np.linalg.norm(r) < 1e-6)):
return None, r, s
point_intersection = A + r * (B - A)
return point_intersection, r, s
def subcalc(self, queue, seed, offspring, elites, n):
np.random.seed(seed)
crossover = []
for i in range(10):
randint1 = np.random.randint(len(offspring))
randint2 = np.random.randint(len(elites))
child = self.cx(offspring[randint1], elites[randint2])
crossover.append(child)
queue.put(crossover)
def cx(self, ind1, ind2):
"""Crossover function for path planning."""
# --- If the ind1 and the ind2 is the same path, return ind1 and exit.
if len(ind1) == len(ind2) and all((ind1 == ind2).flatten()):
return ind1
# --- Initialize
best = []
id1 = [0]
id2 = [0]
tmp1 = ind1.copy()
tmp2 = ind2.copy()
j = 0
# --- Search for the intersection
for i1 in range(len(ind1) - 1):
for i2 in range(len(ind2) - 1):
# Calculate an intersection between line AB and line CD.
pt, r, s = self.calc_intersection_point(
ind1[i1], ind1[i1 + 1], ind2[i2], ind2[i2 + 1])
# If intersection is found,
if pt is not None:
if np.linalg.norm(r - 1.0) > 1e-6 and \
np.linalg.norm(r) > 1e-6:
# Add the intersection to the point lists.
tmp1 = np.insert(tmp1, i1 + j + 1, pt, axis=0)
tmp2 = np.insert(tmp2, i2 + j + 1, pt, axis=0)
# Revise the intersection lists.
id1.append(i1 + j + 1)
id2.append(i2 + j + 1)
# j: Num. of the intersection points.
j = j + 1
# Add the last point of the path to the intersection lists.
id1 = id1 + [len(ind1) + j + 1]
id2 = id2 + [len(ind2) + j + 1]
# --- Select the best path based on the path length.
for i in range(len(id1) - 1):
if (self.eval_for_cx(tmp1[id1[i]:id1[i + 1] + 1])
< self.eval_for_cx(tmp2[id2[i]: id2[i + 1] + 1])):
best = best + list(tmp1[id1[i]: id1[i + 1]])
else:
best = best + list(tmp2[id2[i]: id2[i + 1]])
return Individual(best)
def node_reduction(self, path):
path = np.array(path)
new_path = [path[0]]
for i in range(1, len(path) - 1):
u = path[i + 1] - path[i]
umag = np.linalg.norm(u)
if umag > 0.1:
new_path.append(path[i])
elif not self.ccm.line_validation(new_path[-1], path[i + 1]):
new_path.append(path[i])
new_path.append(path[-1])
path = new_path.copy()
new_path = [path[0]]
for i in range(1, len(path) - 1):
u = path[i + 1] - path[i]
v = path[i - 1] - path[i]
umag = np.linalg.norm(u)
vmag = np.linalg.norm(v)
if umag != 0 and vmag != 0:
cos = np.dot(u, v) / (umag * vmag)
if abs(cos) < 0.99:
new_path.append(path[i])
elif not self.ccm.line_validation(new_path[-1], path[i + 1]):
new_path.append(path[i])
new_path.append(path[-1])
new_path = np.array(new_path)
return new_path
def mut_normal(self, ind, indpb, maxiter):
"""Mutation function."""
mut = ind.copy()
for i in range(1, len(ind) - 1):
if random.random() < indpb:
var = 0.5
for j in range(maxiter):
mut[i] = ind[i] + np.random.normal(0.0, var, 2)
var = var * 0.5
if self.ccm.path_validation(
[mut[i - 1], mut[i], mut[i + 1]]):
break
else:
mut[i] = ind[i]
return Individual(mut)
def smoothing(self, pts, delta=0.1):
# resampled_path = post_process.resampling(pts, delta)
# bezier_path = post_process.bezier(resampled_path, 50)
# return bezier_path
return pts
def main(self, duration):
'''
Main Routine for Genetic Algorithm
'''
ini_time = time.time()
# --- Evaluate the initial population
self.set_fitness(self.evalOneMax)
self.best_ind = min(self.pop, key=attrgetter("fitness"))
self.history.append([self.gtot, self.best_ind.fitness])
# self.trajectory.append([world.start[0], world.start[1]])
np.vstack((self.trajectory, self.world.start))
# if verbose:
# rp.plot(self.pop, self.best_ind, self.trajectory, self.history)
# --- Generation loop starts.
print('\n[Genetic Algorithm]')
print("Generation loop start.")
print("Generation: 0. Best fitness: {}\n"
.format(str(self.best_ind.fitness)))
for g in range(self.NGEN):
'''
STEP1 : Selection.
'''
t0 = time.time()
# Elite selection
elites = self.selElite()
# Tournament Selection
offspring = self.selTournament(tournsize=3)
'''
STEP2 : Mutation.
'''
t1 = time.time()
mutant = []
for ind in offspring:
if np.random.rand() < 0.5:
tmp = self.mut_normal(ind, indpb=0.3, maxiter=3)
mutant.append(tmp)
else:
mutant.append(ind)
offspring = mutant.copy()
# mutant = self.create_pop_cd(10)
mutant = self.create_pop_prm(2, 5)
'''
Step3 : Crossover.
'''
t2 = time.time()
proc = 8
n_cross = 10
queue = mp.Queue()
ps = [
mp.Process(target=self.subcalc, args=(
queue, i, offspring, elites, n_cross))
for i in np.random.randint(100, size=proc)
]
for p in ps:
p.start()
crossover = []
for i in range(proc):
crossover += queue.get()
'''
STEP4: Update next generation.
'''
t3 = time.time()
n_offspring = self.n_ind - \
(len(elites) + len(crossover) + len(mutant))
self.pop = (list(elites) + list(crossover) +
list(offspring[0:n_offspring])) + list(mutant)
# --- Delete the redundant points on the path.
self.pop = [Individual(self.node_reduction(path))
for path in self.pop]
self.set_fitness(self.evalOneMax)
'''
Output
'''
t4 = time.time()
# --- Print best fitness in the population.
self.best_ind = min(self.pop, key=attrgetter("fitness"))
print("Generation: {: > 2}".format(g))
print("Best fitness: {: .3f}".format(self.best_ind.fitness))
print("Time: {0:.3f}, {1:.3f}, {2:.3f}, {3:.3f}, Total: {4:.3f} \n"
.format(t1 - t0, t2 - t1, t3 - t2, t4 - t3, t4 - t0))
# Fitness transition
self.history.append([self.gtot, self.best_ind.fitness])
'''
# STEP5: Termination
'''
if time.time() - ini_time > duration:
# --- Visualization
if self.verbose:
rp.plot(self.pop, self.best_ind,
self.trajectory, self.history)
break
if self.best_ind.fitness <= self.fitness_thresh:
print('The best fitness reaches the threshold value.')
break
if g >= 5:
diff = np.abs(self.history[-5][1] - self.history[-1][1])
if diff < 1e-2:
print('The best fitness does not change for 10 steps.')
break
self.gtot += 1
if __name__ == '__main__':
'''
World setting
'''
# World
world = World()
world.generate_frame([-pi, pi], [-pi, pi])
world.type = 'cartesian'
# Objects in the cartesian space
og = ObjectGenerator(world)
og.generate_object_sample1()
# --- Set start/goal point
world.start = np.array([-pi / 2, -pi / 2]) * 1.9
world.goal = np.array([pi / 2, pi / 2]) * 1.9
'''
Functions
'''
NGEN = 1000
n_ind = 100
n_elite = 10
fitness_thresh = 0.1
verbose = True
ga = GeneticAlgorithm(world, NGEN, n_ind, n_elite, fitness_thresh, verbose)
rp = RealtimePlot(world, 100, dt=0.01)
# --- Initial Population
# initial_pop = create_pop_cd(100)
initial_pop = ga.create_pop_prm(10, 10)
ga.pop = initial_pop.copy()
for i in range(100):
ga.main(1.0)
# Update objects.
if i % 10 < 5:
vel = 0.05
else:
vel = -0.05
dl = np.array([[vel, 0.0], [-vel, 0.0]])
world.update_objects(dl)
ga.world_margin.update_objects(dl)
# Update current robot position.
dl = 0.1
world.update_start(ga.best_ind, dl)
ga.world_margin.update_start(ga.best_ind, dl)
for i in range(len(ga.pop)):
ga.pop[i][0] = world.start
# Create trajectory
ga.trajectory = np.vstack((ga.trajectory, world.start))
if ga.best_ind.fitness <= ga.fitness_thresh:
print('Goal.')
break
if verbose:
rp.fig.clf()
plt.show()