-
Notifications
You must be signed in to change notification settings - Fork 0
/
robot_control.py
446 lines (327 loc) · 17.1 KB
/
robot_control.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
# -*- coding: utf-8 -*-
"""
"""
from marco_nest_utils import utils
__author__ = 'marco'
import numpy as np
from scipy.integrate import odeint
from mpl_toolkits import mplot3d
from matplotlib import pyplot as plt
import time
from nest_multiscale.nest_multiscale import generate_poisson_trains, evaluate_fir, calculate_instantaneous_fr, set_poisson_fr
class cereb_control():
def __init__(self, nest_, Cereb_class, robot, desired_trajectory, desired_traj_vel=None,
cortical_input=None, control_tau=None):
self.nest_ = nest_
self.Cereb_class = Cereb_class
self.des_trajec = desired_trajectory # function of time
self.des_trajec_vel = desired_traj_vel
self.CTX_pop_list = self.get_CTX_pops()
self.starting_state = [des_tr(0) for des_tr in self.des_trajec] + [des_tr_v(0) for des_tr_v in self.des_trajec_vel]
self.T = None # will be set once linked to simulation handler
self.max_traj = None # will be updated later
self.min_traj = None
self.max_traj_vel = None
self.min_traj_vel = None
self.robot = robot
self.n_joints = len(robot.q)
assert self.n_joints == len(desired_trajectory), 'Should provide as many trajectory as robot joints'
self.u_R_dim = None # total number of moments
self.y_R_dim = self.u_R_dim # total number of error signals to be sent to IO
self.rng = np.random.default_rng(round(time.time() * 1000))
self.pop_list_to_robot = None
self.sd_list_R = None
self.cortical_input = cortical_input # the signal delivered from the cortex
self.control_tau = control_tau # the tau sequence which will generate the desired trajectory
def start(self, Sim_time, T_sample):
self.pop_list_to_robot = self.cereb_control_setup(Sim_time, T_sample)
self.sd_list_R = utils.attach_spikedetector(self.nest_, self.pop_list_to_robot)
self.cereb_control_buffers()
# robot
self.robot.set_robot_T(T_sample)
self.robot.n_joints = len(self.robot.q)
def before_loop(self):
self.robot.reset_state(self.starting_state)
self.tau_old = np.zeros((self.prev_steps, 2 * self.n_joints))
# io_old = np.zeros((prev_steps_io, 2 * self.n_joints))
# q_val = [0.] # visualizer buffer
def beginning_loop(self, trial_time, total_time):
# set future RBF input
if self.cortical_input is None:
self.generate_RBF_activity(trajectory_time=trial_time, simulation_time=total_time)
else:
self.generate_RBF_activity(trajectory_time=trial_time, simulation_time=total_time,
ctx_input=self.cortical_input[int(trial_time)] * 200.)
def ending_loop(self, trial_time, total_time):
# a) Update robot position
# use tau of (previous step!) to update robot position
tau_pos = self.tau[1::2] # to take odd elements: [1::2]
tau_neg = self.tau[0::2] # to take even elements: [0::2]
if self.n_joints == 1: k = 0.4
if self.n_joints == 2: k = 0.5
if self.control_tau is not None:
if self.cortical_input is None:
self.robot.update_state( # only apply the control sequence + tau pos and neg difference
(tau_pos - tau_neg) * k + self.control_tau(trial_time) * 1)
else:
self.robot.update_state( # apply the control sequence modulated by cortical input
(tau_pos - tau_neg) * k + # + tau pos and neg difference
self.control_tau(trial_time) * self.cortical_input[int(trial_time)] / 0.4102)
else:
self.robot.update_state((tau_pos - tau_neg) * k)
self.e_old = np.concatenate((self.e_old, np.zeros((1, self.n_joints))))
for k in range(self.n_joints):
# b) Inject robot error in IO
# joint position has just been updated to tt with tau(tt - T), so select tt!
e_new = self.get_error_value(trial_time + self.T, self.robot.q[k], k) # self.robot.q[j])
self.e_old[-1, k] = e_new
e = self.e_old[0, k]
if self.n_joints == 1: e = e * 5.
if self.n_joints == 2: e = e * 3.
if e > 0.:
if e > 7.: e = 7.
# set the future spike trains (in [tt + T, tt + 2T])
set_poisson_fr(self.nest_, e, [self.Cereb_class.CTX_pops['US_p'][k]], total_time + self.T,
self.T, self.rng)
elif e < 0.:
if e < -7.: e = -7.
# set the future spike trains (in [tt + T, tt + 2T])
set_poisson_fr(self.nest_, -e, [self.Cereb_class.CTX_pops['US_n'][k]], total_time + self.T,
self.T, self.rng)
self.e_old = self.e_old[1:, :]
# update tau value for next step
new_tau = calculate_instantaneous_fr(self.nest_, self.sd_list_R, self.pop_list_to_robot,
time=total_time, T_sample=self.T)
# new_tau, t_prev = calculate_instantaneous_burst_activity(self.nest_, self.sd_list_R, self.pop_list_to_robot,
# time=actual_sim_time, t_prev=t_prev)
self.tau, self.tau_old = evaluate_fir(self.tau_old, new_tau.reshape((1, self.u_R_dim)), kernel=self.kernel_robot)
if self.tau_sol is None:
self.tau_sol = np.array([self.tau])
else:
self.tau_sol = np.concatenate((self.tau_sol, [self.tau]), axis=0)
# print(self.tau_sol)
# update io
# new_io = calculate_instantaneous_fr(self.nest_, sd_list_io, list_io,
# time=actual_sim_time, T_sample=self.T)
# io, io_old = evaluate_fir(io_old, new_io.reshape((1, 2 * self.n_joints)), kernel=kernel)
# if io_sol is None:
# io_sol = np.array([io])
# else:
# io_sol = np.concatenate((io_sol, [io]), axis=0)
# def after_loop(self):
# self.io_sol = io_sol
# self.tau_sol = tau_sol
def cereb_control_setup(self, Sim_time, T_sample):
# set proper sampling time in cereb control
self.set_T_and_max_func(T_sample, Sim_time)
# divide dcn in pos tau and neg tau
pop_list_to_robot = self.Cereb_class.get_dcn_indexes(self.n_joints)
self.u_R_dim = len(pop_list_to_robot) # total number of moments
self.y_R_dim = self.u_R_dim # total number of error signals to be sent to IO
return pop_list_to_robot # , pf_pc_conn, w0
def cereb_control_buffers(self):
tau0 = np.array([0.] * self.u_R_dim)
self.tau_sol = None # will contain all solutions over time
self.kernel_robot = half_gaussian(sigma=600., width=300., sampling_time=self.T)
self.tau = tau0
self.prev_steps = len(self.kernel_robot)
self.tau_old = np.zeros((self.prev_steps, 2 * self.n_joints))
# sd_list_io, list_io = self.attach_io_spike_det()
# io0 = np.array([0.] * 2)
# io_sol = None # will contain all solutions over time
# io = io0
# prev_steps_io = len(kernel)
# io_old = np.zeros((prev_steps, 2 * self.n_joints))
self.robot.reset_state(self.starting_state)
self.e_old = np.zeros((int(140 / self.T), self.n_joints))
def generate_RBF_activity(self, trajectory_time, simulation_time, ctx_input=80.):
"""
:param time: actual time, in ms
:param CTX_pops: cortex populations, projecting to glomeruli
:param ctx_input: signal amplitude
:return:
"""
sd = 3. / len(self.des_trajec)
input_mf_dim = len(self.CTX_pop_list[0]) # dimension of every mf input (j1pos, j1vel, j2pos, ...)
glom_grid = np.linspace(1, input_mf_dim, input_mf_dim, endpoint=True)
if self.des_trajec_vel is None:
for k, des_tr in enumerate(self.des_trajec):
min_val = self.min_traj[k]
max_val = self.max_traj[k]
mu_x = 1 + (input_mf_dim - 1) * (des_tr(trajectory_time) - min_val) / (max_val - min_val)
# microm
fr = gaussian_1D(glom_grid, mu_x, sd) * ctx_input
spike_times = generate_poisson_trains(self.CTX_pop_list[k], fr, self.T, simulation_time, self.rng)
generator_params = [{"spike_times": s_t, "spike_weights": [1.] * len(s_t)} for s_t in spike_times]
self.nest_.SetStatus(self.CTX_pop_list[k], generator_params)
elif self.des_trajec_vel is not None:
for k, des_tr, des_tr_vel in zip(range(len(self.des_trajec)), self.des_trajec, self.des_trajec_vel):
if ctx_input < 0.:
ctx_input = 0.
min_val = self.min_traj[k]
max_val = self.max_traj[k]
mu_x_pos = 1 + (input_mf_dim-1) * (des_tr(trajectory_time) - min_val) / (max_val - min_val) # microm
fr_pos = gaussian_1D(glom_grid, mu_x_pos, sd) * ctx_input
min_val = self.min_traj_vel[k]
max_val = self.max_traj_vel[k]
mu_x_vel = 1 + (input_mf_dim-1) * (des_tr_vel(trajectory_time) - min_val) / (max_val - min_val) # microm
fr_vel = gaussian_1D(glom_grid, mu_x_vel, sd) * ctx_input
spike_times_pos = generate_poisson_trains(self.CTX_pop_list[2*k], fr_pos, self.T, simulation_time, self.rng)
generator_params = [{"spike_times": s_t, "spike_weights": [1.] * len(s_t)} for s_t in spike_times_pos]
self.nest_.SetStatus(self.CTX_pop_list[2*k], generator_params)
spike_times_vel = generate_poisson_trains(self.CTX_pop_list[2*k+1], fr_vel, self.T, simulation_time, self.rng)
generator_params = [{"spike_times": s_t, "spike_weights": [1.] * len(s_t)} for s_t in spike_times_vel]
self.nest_.SetStatus(self.CTX_pop_list[2*k+1], generator_params)
def get_CTX_pops(self):
'''
Separate cortical population in subpopulation according to the order:
J1_pos, J1_vel, J2_pos, J2_vel, ...
(if vel is not defined: J1_pos, J2_pos, ...)
:return: List of cortical populations according
'''
n_joints = len(self.des_trajec)
CTX_len = len(self.Cereb_class.CTX_pops['CTX'])
if self.des_trajec_vel is None:
sub_list_len = int(CTX_len/n_joints)
pop_list = [self.Cereb_class.CTX_pops['CTX'][j*sub_list_len:(j+1)*sub_list_len] for j in range(n_joints)]
elif self.des_trajec_vel is not None:
sub_list_len = int(CTX_len/(n_joints * 2))
pop_list = [self.Cereb_class.CTX_pops['CTX'][j*sub_list_len:(j+1)*sub_list_len] for j in range(n_joints * 2)]
return pop_list
def set_T_and_max_func(self, t_sample_, sim_time_):
self.T = t_sample_
t_grid = np.linspace(0, (sim_time_ - sim_time_ % t_sample_), int(sim_time_ / t_sample_), endpoint=False)
self.max_traj = [max([des_tr(t) for t in t_grid]) for des_tr in self.des_trajec]
self.min_traj = [min([des_tr(t) for t in t_grid]) for des_tr in self.des_trajec]
if self.des_trajec_vel is not None:
self.max_traj_vel = [max([des_tr(t) for t in t_grid]) for des_tr in self.des_trajec_vel]
self.min_traj_vel = [min([des_tr(t) for t in t_grid]) for des_tr in self.des_trajec_vel]
def get_error_value(self, time, robot_state, joint):
error = self.des_trajec[joint](time) - robot_state
return error
def attach_io_spike_det(self):
io_list = []
sub_pop_IO_len = int(len(self.Cereb_class.Cereb_pops['IO']) / self.n_joints)
for j in range(self.n_joints):
io_neg = list(self.Cereb_class.Cereb_pops['IO'][
2 * j * sub_pop_IO_len // 2:(2 * j + 1) * sub_pop_IO_len // 2])
io_pos = list(self.Cereb_class.Cereb_pops['IO'][
(2 * j + 1) * sub_pop_IO_len // 2:(2 * j + 2) * sub_pop_IO_len // 2])
io_list += [io_neg] + [io_pos]
return utils.attach_spikedetector(self.nest_, io_list), io_list
class rbot():
def __init__(self):
q0 = np.zeros((1, 1))
qd0 = np.zeros((1, 1))
g = 9.81
l2 = 1. # m
m2 = 1. # kg
self.int_t = None # to be set by call in simulation_handler
self.q = q0
self.qd = qd0
self.joint_pos = None
self.joint_vel = None
self.rhs = self.robot_rhs(m2, l2, g)
def set_robot_T(self, t_sample):
sub_interv = 10 # integrate in 1 T period, sampling every T/sub_interv
self.int_t = np.linspace(0, t_sample / 1000., sub_interv + 1)
def robot_rhs(self, m, l, g):
I = m * l ** 2 / 3
def rhs(x, t, u):
qd = x[1]
qdd = (u[0] - g * m * l / 2 * np.sin(x[0])) / I
return np.array([qd, qdd])
return rhs
def update_state(self, tau):
state = np.concatenate((self.q, self.qd))
sol = odeint(self.rhs, state, self.int_t, args=(tau,))
self.q = np.array([sol[-1, 0]])
self.qd = np.array([sol[-1, 1]])
if self.joint_pos is None:
self.joint_pos = np.array([self.q])
else:
self.joint_pos = np.concatenate((self.joint_pos, [self.q]), axis=0)
if self.joint_vel is None:
self.joint_vel = np.array([self.qd])
else:
self.joint_vel = np.concatenate((self.joint_vel, [self.qd]), axis=0)
def reset_state(self, starting_state_):
self.q = np.array([starting_state_[0]])
self.qd = np.array([starting_state_[1]])
# self.joint_pos = np.concatenate((self.joint_pos, self.q * np.ones(1)), axis=0)
class rrbot():
def __init__(self):
q0 = np.zeros(2)
qd0 = np.zeros(2)
g = 0. # 9.81
l1 = 1. # m
l2 = 1. # m
m1 = 1. # kg
m2 = 1. # kg
self.int_t = None # to be set by call in simulation_handler
self.q = q0
self.qd = qd0
self.joint_pos = None
self.joint_vel = None
self.rhs = self.robot_rhs(m1, m2, l1, l2, g)
def set_robot_T(self, t_sample):
sub_interv = 10 # integrate in 1 T period, sampling every T/sub_interv
self.int_t = np.linspace(0, t_sample / 1000., sub_interv + 1)
def robot_rhs(self, m1, m2, l1, l2, g):
M = lambda q: np.array([[(m1 + m2) * l1 ** 2 + m2 * l2 ** 2 + 2 * m2 * l1 * l2 * np.cos(q[1]), m2 * l2 ** 2 + m2 * l1 * l2 * np.cos(q[1])],
[m2 * l2 ** 2 + m2 * l1 * l2 * np.cos(q[1]), m2 * l2 ** 2]])
C = lambda q, qd: np.array([[-m2 * l1 * l2 * (2 * qd[0] * qd[1] + qd[0] ** 2) * np.sin(q[1])], # change sign according to notation
[-m2 * l1 * l2 * qd[0] * qd[1] * np.sin(q[1])]])
G = lambda q, qd : np.array([[-(m1 + m2) * g * l1 * np.sin(q[0]) - m2 * g * l2 * np.sin(q[0] + q[1])],
[-m2 * g * l2 * np.sin(q[0] + q[1])]])
def rhs(x, t, u):
q0 = x[0:2].T
qd0 = x[2:4].T
qd = qd0
qdd = np.linalg.solve(M(q0), u - (G(q0, qd0).T)[0] - (C(q0, qd0).T)[0])
return np.concatenate((qd, qdd))
return rhs
def update_state(self, tau):
state = np.concatenate((self.q, self.qd))
sol = odeint(self.rhs, state, self.int_t, args=(tau,))
self.q = sol[-1, 0:2]
self.qd = sol[-1, 2:4]
if self.joint_pos is None:
self.joint_pos = np.array([self.q])
else:
self.joint_pos = np.concatenate((self.joint_pos, [self.q]), axis=0)
if self.joint_vel is None:
self.joint_vel = np.array([self.qd])
else:
self.joint_vel = np.concatenate((self.joint_vel, [self.qd]), axis=0)
def reset_state(self, starting_state_):
self.q = np.array(starting_state_[0:2])
self.qd = np.array(starting_state_[2:4])
# self.joint_pos = np.concatenate((self.joint_pos, self.q * np.ones(1)), axis=0)
def half_gaussian(sigma, width, sampling_time):
s = sigma/sampling_time
w = int(width/sampling_time)
time_p = w - np.linspace(0, w, w+1) # create kernel linspace
# create a half gaussian kernel:
kernel = 1. / (np.sqrt(2. * np.pi) * s) * np.exp(-np.power((time_p - 0.) / s, 2.) / 2)
kernel = kernel/kernel.sum() # normalize
# vsl.simple_plot(-time_p, kernel) # visualize
return kernel
def gaussian_2D(x, mu, sigma):
g = np.exp(-np.power(np.linalg.norm(x - mu, axis=1) / sigma, 2) / 2) / np.exp(0)
return np.round(g, 3)
def gaussian_1D(x, mu, sigma):
g = np.exp(-np.power((np.array(x) - mu) / sigma, 2) / 2) / np.exp(0)
return np.round(g, 3)
def plot_3D(x, z, fr, tr_time):
if tr_time % 200 == 0:
fig = plt.figure()
ax = plt.axes(projection='3d')
ax.scatter(x, z, fr, cmap='viridis', edgecolor='none')
ax.set_title('Surface plot')
ax.set_xlim3d(0, 400)
ax.set_ylim3d(0, 400)
ax.set_zlim3d(0, 40)
ax.set_xlabel('x')
ax.set_ylabel('z')
plt.show()