forked from Rohit-Dhamija/MPC_for_differential_bots
-
Notifications
You must be signed in to change notification settings - Fork 0
/
mpc_multipleShooting_pathTracking_turtlebot3.py
executable file
·343 lines (228 loc) · 19.2 KB
/
mpc_multipleShooting_pathTracking_turtlebot3.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
#!/usr/bin/env python
import rospy
from geometry_msgs.msg import Twist
from nav_msgs.msg import Odometry
from nav_msgs.msg import Path
import casadi as ca
import numpy as np
import time
import math
from scipy.spatial import KDTree
pi = math.pi
inf = np.inf
t_start = time.time()
"""# variable parameters
"""
n_states = 3
n_controls = 2
N =100 # Prediction horizon(same as control horizon)
error_allowed = 0.1
U_ref = np.array([0.22,0], dtype ='f') # Reference velocity and reference omega
Q_x = 300 # gains to control error in x,y,theta during motion
Q_y = 300
Q_theta = 20
R1 = 250 # gains to control magnitude of V and omega
R2 = 80
error_allowed_in_g = 1e-100 # error in contraints (should be ~ 0)
"""# parameters that depend on simulator
"""
n_bound_var = n_states
x_bound_max = inf # enter x and y bounds when limited world like indoor environments
x_bound_min = -inf
y_bound_max = inf
y_bound_min = -inf
theta_bound_max = inf
theta_bound_min = -inf
v_max = 0.22
v_min = 0#-v_max # when we don't want to track the path on backward direction
omega_max = 2.84
omega_min = -omega_max
global x,y,theta,qx,qy,qz,qw,V,omega # (x,y,theta) will store the current position and orientation
# qx,qy,qz,qw will store the quaternions of the bot position
# V and omega will store the inputs to the bot(Speed and Angular Velocity)
global total_path_points
total_path_points = 0
global path
def odomfunc(odom):
global x,y,qx,qy,qz,qw,theta
x = odom.pose.pose.position.x
y = odom.pose.pose.position.y
qx = odom.pose.pose.orientation.x # quaternions of location
qy = odom.pose.pose.orientation.y
qz = odom.pose.pose.orientation.z
qw = odom.pose.pose.orientation.w
theta = math.atan2(2*(qx*qy+qw*qz),1-2*(qy*qy+qz*qz)) # finding yaw from quaternions
def pathfunc(Path):
global total_path_points,path
if total_path_points == 0:
total_path_points = len(Path.poses)
path = np.zeros((total_path_points,2))
for i in range(0,total_path_points):
path[i][0] = Path.poses[i].pose.position.x
path[i][1] = Path.poses[i].pose.position.y
def my_mainfunc():
rospy.init_node('mpc_multipleShooting_pointTracking_turtlebot3', anonymous=True)
rospy.Subscriber('/odom', Odometry , odomfunc)
rospy.Subscriber('/astroid_path', Path, pathfunc)
instance = rospy.Publisher('/cmd_vel', Twist, queue_size=10)
rate = rospy.Rate(10)
rate.sleep() #rate.sleep() to run odomfunc once
path_resolution = ca.norm_2(path[0,0:2] - path[1,0:2])
global delta_T #timestamp bw two predictions
delta_T = ( path_resolution / ((U_ref[0])) )/10
msg = Twist()
"""MPC"""
x_casadi =ca.SX.sym('x')
y_casadi = ca.SX.sym('y')
theta_casadi = ca.SX.sym('theta')
states =np.array([(x_casadi),(y_casadi),(theta_casadi)])
n_states = states.size
v_casadi =ca.SX.sym('v')
omega_casadi = ca.SX.sym('omega')
controls = np.array([v_casadi,omega_casadi])
n_controls = controls.size
rhs = np.array([v_casadi*ca.cos(theta_casadi),v_casadi*ca.sin(theta_casadi),omega_casadi])
f = ca.Function('f',[states,controls],[rhs]) # function to predict rhs using states and controls
U = ca.SX.sym('U', n_controls,N) # For storing predicted controls
X =ca.SX.sym('X', n_states, N+1) # For storing predicted states
P = ca.SX.sym('P',1, n_states + n_states*(N) + n_controls*(N) ) # For storing odometry, next N path points and next N referance controls
obj = 0
g = []
Q = ca.diagcat(Q_x, Q_y,Q_theta)
R = ca.diagcat(R1, R2)
for i in range(0,N):
cost_pred_st = ca.mtimes( ca.mtimes( (X[0:n_states,i] - P[n_states*(i+1) :n_states*(i+1) + n_states ].reshape((n_states,1)) ).T , Q ) , (X[0:n_states,i] - P[n_states*(i+1) :n_states*(i+1) + n_states ].reshape((n_states,1)) ) ) + ca.mtimes( ca.mtimes( ( (U[0:n_controls,i]) - P[n_states*(N+1)+n_controls*(i):n_states*(N+1)+n_controls*(i) + n_controls].reshape((n_controls,1)) ).T , R ) , U[0:n_controls,i] - P[n_states*(N+1)+n_controls*(i):n_states*(N+1)+n_controls*(i) + n_controls].reshape((n_controls,1)) )
obj = obj + cost_pred_st
pred_st = np.zeros((n_states,1))
for i in range(0,N+1): # adding contraints so the predictions are in sync with vehicle model
if i == 0:
g = ca.vertcat( g,( X[0:n_states,i] - P[0:n_states].reshape((n_states,1)) ) )
else:
#f_value = f(X[0:n_states,i-1],U[0:n_controls,i-1]) # euler method not used
#pred_st = X[0:n_states,i-1] + delta_T*f_value
K1 = f(X[0:n_states,i-1],U[0:n_controls,i-1]) # Runge Kutta method of order 4
K2 = f(X[0:n_states,i-1] + np.multiply(K1,delta_T/2),U[0:n_controls,i-1])
K3 = f(X[0:n_states,i-1] + np.multiply(K2,delta_T/2),U[0:n_controls,i-1])
K4 = f(X[0:n_states,i-1] + np.multiply(K3,delta_T),U[0:n_controls,i-1])
pred_st = X[0:n_states,i-1] + (delta_T/6)*(K1+2*K2+2*K3+K4) # predicted state
g = ca.vertcat( g,(X[0:n_states,i] - pred_st[0:n_states].reshape((n_states,1)) ) )
OPT_variables = X.reshape((n_states*(N+1),1))
OPT_variables = ca.vertcat( OPT_variables, U.reshape((n_controls*N,1)) )
nlp_prob ={
'f':obj,
'x':OPT_variables,
'g':g,
'p':P
}
opts = {
'ipopt':
{
'max_iter': 100,
'print_level': 0,
'acceptable_tol': 1e-8,
'acceptable_obj_change_tol': 1e-6
},
'print_time': 0
}
solver = ca.nlpsol('solver', 'ipopt', nlp_prob, opts)
lbg = ca.DM.zeros(((n_states)*(N+1),1)) # bounds on g
ubg = ca.DM.zeros(((n_states)*(N+1),1))
lbg[0:(n_states)*(N+1)] = - error_allowed_in_g
ubg[0:(n_states)*(N+1)] = error_allowed_in_g
lbx = ca.DM.zeros((n_states*(N+1) + n_controls*N,1)) # bounds on X
ubx = ca.DM.zeros((n_states*(N+1) + n_controls*N,1))
lbx[0:n_bound_var*(N+1):3] = x_bound_min
ubx[0:n_bound_var*(N+1):3] = x_bound_max
lbx[1:n_bound_var*(N+1):3] = y_bound_min
ubx[1:n_bound_var*(N+1):3] = y_bound_max
lbx[2:n_bound_var*(N+1):3] = theta_bound_min
ubx[2:n_bound_var*(N+1):3] = theta_bound_max
lbx[n_bound_var*(N+1):(n_bound_var*(N+1)+n_controls*N):2] = v_min
ubx[(n_bound_var*(N+1)):(n_bound_var*(N+1)+n_controls*N):2] = v_max
lbx[(n_bound_var*(N+1)+1):(n_bound_var*(N+1)+n_controls*N):2] = omega_min
ubx[(n_bound_var*(N+1)+1):(n_bound_var*(N+1)+n_controls*N):2] = omega_max
X_init = np.array([x,y,theta], dtype = 'f')
X_target = np.array([ path[total_path_points-1][0], path[total_path_points-1][1], 0 ] , dtype = 'f')
P = X_init
close_index = KDTree(path).query(P[0:n_states-1])[1]
for i in range(0,N):
P = ca.vertcat(P,path[close_index+i,0:2])
P = ca.vertcat(P, math.atan((path[close_index+i+1][1] - path[close_index+i][1])/(path[close_index+i+1][0] - path[close_index+i][0])) )
for i in range(0,N):
P = ca.vertcat(P, U_ref[0])
P = ca.vertcat(P, U_ref[1])
initial_X = ca.DM.zeros((n_states*(N+1))) #all initial predicted states are X_init
initial_X[0:n_states*(N+1):3] = X_init[0]
initial_X[1:n_states*(N+1):3] = X_init[1]
initial_X[2:n_states*(N+1):3] = X_init[2]
initial_con = ca.DM.zeros((n_controls*N,1)) #initial search value of control matrix
n_iter = 0
while ( ca.norm_2( P[0:n_states-1].reshape((n_states-1,1)) - X_target[0:n_states-1] ) > error_allowed ) :
n_iter += 1
args = {
'lbx':lbx,
'lbg':lbg,
'ubx':ubx,
'ubg':ubg,
'p':P,
'x0':ca.vertcat(initial_X,initial_con),
}
sol = solver(
x0=args['x0'],
lbx=args['lbx'],
ubx=args['ubx'],
lbg=args['lbg'],
ubg=args['ubg'],
p=args['p']
)
X_U_sol = sol['x']
V = (X_U_sol[n_states*(N+1)].full())[0][0]
omega = (X_U_sol[n_states*(N+1)+1].full())[0][0]
#omega_left_wheel = (V - omega*robot_dia)/wheel_rad # differential drive kinematics (when global x cross y faces upward)
#omega_right_wheel = (V + omega*robot_dia)/wheel_rad
#omega_left_wheel = (V + omega*robot_dia)/wheel_rad # differential drive kinematics (when global x cross y faces downward)
#omega_right_wheel = (V - omega*robot_dia)/wheel_rad
msg.linear.x = V #linear.x and linear.y are velocities in local coordinates of bot
msg.linear.y = 0 # linear.y always zero, linear.x is the speed of a diff. bot
msg.linear.z = 0
msg.angular.x = 0
msg.angular.y = 0
msg.angular.z = omega
instance.publish(msg)
P[0:n_states] = [x,y,theta]
close_index = KDTree(path).query(np.array([x,y]))[1]
if N+(close_index-1) < total_path_points : # Updating P for next N path points and next N reference controls
P[n_states:n_states*(N+1):n_states] = path[close_index:N+close_index,0]
P[n_states+1:n_states*(N+1):n_states] = path[close_index:N+close_index,1]
for i in range(0,N):
P[n_states*(i+1+1)-1] = math.atan( (path[i+close_index+1][1] - path[i+close_index][1])/(path[i+close_index+1][0] - path[i+close_index][0] + 1e-9) )
P[n_states*(N+1):n_states*(N+1)+n_controls*(N-1)]= P[n_states*(N+1)+n_controls:n_states*(N+1)+n_controls*(N)]
P[n_states*(N+1)+n_controls*(N-1):n_states*(N+1)+n_controls*(N)] = U_ref
else:
print (" The end point in inside horizon, slowing down")
P[n_states:n_states*(N)] = P[n_states*2:n_states*(N+1)]
P[n_states*(N):n_states*(N+1)-1] = path[(total_path_points-1),0:2]
P[n_states*(N+1)-1] = math.atan( (path[total_path_points-1][1] - path[total_path_points-1-1][1])/(path[total_path_points-1][0] - path[total_path_points-1-1][0]) )
P[n_states*(N+1):n_states*(N+1)+n_controls*(N-1)]= P[n_states*(N+1)+n_controls:n_states*(N+1)+n_controls*(N)]
P[n_states*(N+1)+n_controls*(N-1):n_states*(N+1)+n_controls*(N)] = np.array([0,0], dtype ='f') # we need to stop the bot at end, hence referance controls 0 at end
for i in range(0,N*n_states): #initial search value of state for next iteration should be the predicted one for that iteration
initial_X[i] = X_U_sol[i+n_states]
for i in range(0,(N-1)*n_controls): #initial search value of control for next iteration should be the predicted one for that iteration
initial_con[i] = X_U_sol[n_states*(N+1)+i+n_controls]
rate.sleep()
print ("PATH TRACKED")
print ("Total MPC iterations = " , n_iter)
t_end = time.time()
print ("Total Time taken = " , t_end - t_start)
msg.linear.x = 0 # stopping the bot
msg.linear.y = 0
msg.linear.z = 0
msg.angular.x = 0
msg.angular.y = 0
msg.angular.z = 0
instance.publish(msg)
if __name__ == '__main__':
try:
my_mainfunc()
except rospy.ROSInterruptException:
pass