forked from avalenzu/miconvex-planning
/
constructHopperModels.py
194 lines (148 loc) · 7.09 KB
/
constructHopperModels.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
from __future__ import division
import numpy as np
from uuid import uuid4
from pyomo.environ import *
from pyomo.opt import SolverFactory
from math import sqrt
from hopper import Hopper
from hopperUtil import *
desiredPrecision = 2
N = 25
tf = 2*1.6
legLength = 0.16
r0 = [0, legLength/2]
rf = [1.0, legLength]
v0 = [0, 0]
w0 = 0
hipOffset = {'front': {'x': 0.5, 'z': -0.25}, 'hind': {'x': -0.5, 'z': -0.25}}
matlab_hopper = eng.Hopper(legLength, hipOffset)
hop = Hopper(N, eng, matlab_hopper)
# hop.mdt_precision = int(ceil(-np.log2(desiredPrecision)))
hop.dtBounds = tuple((1/sqrt(legLength/9.81))*np.array([0.05, 0.2]))
hop.dtNom = 0.04*(1/sqrt(legLength/9.81))
hop.rotationMax = np.pi/8
hop.nOrientationSectors = 1 #int(floor(np.pi/8/desiredPrecision))
print 'hop.nOrientationSectors = %d' % hop.nOrientationSectors
hop.velocityMax = 3.
hop.positionMax = 1.5*rf[0]/legLength
hop.forceMax = 3.
hop.angularVelocityMax = 5.
addThreePlatfomWorld(hop, legLength, 0.3*legLength)
#addFlatWorld(hop, legLength)
hop.constructVisualizer()
m_nlp = hop.constructPyomoModel()
def normL2(m, var):
index = var.index_set()
return sum(var[i]**2 for i in index)
def normL1(m, var):
index = var.index_set()
slackName = '%sSlacks' % var.cname()
lbName = '%sLB' % slackName
ubName = '%sUB' % slackName
slackMax = max([max(np.abs(var[i].bounds)) for i in index])
setattr(m, slackName, Var(index, bounds=(0.0, slackMax)))
def _lbRule(m, *args):
return var[args] <= getattr(m, slackName)[args]
setattr(m, lbName, Constraint(index, rule=_lbRule))
def _ubRule(m, *args):
return var[args] >= -getattr(m, slackName)[args]
setattr(m, ubName, Constraint(index, rule=_ubRule))
return summation(getattr(m, slackName))
def normLInfinity(m, var):
index = var.index_set()
slackName = '%sSlack' % var.cname()
lbName = '%sLB' % slackName
ubName = '%sUB' % slackName
slackMax = max([max(np.abs(var[i].bounds)) for i in index])
setattr(m, slackName, Var(bounds=(0.0, slackMax)))
def _lbRule(m, *args):
return var[args] <= getattr(m, slackName)
setattr(m, lbName, Constraint(index, rule=_lbRule))
def _ubRule(m, *args):
return var[args] >= -getattr(m, slackName)
setattr(m, ubName, Constraint(index, rule=_ubRule))
return getattr(m, slackName)
def exprNormLInfinity(m, expr, slackMax):
slackName = 'slack_%s' % str(uuid4()).replace('-','')
lbName = '%sLB' % slackName
ubName = '%sUB' % slackName
setattr(m, slackName, Var(bounds=(0.0, slackMax)))
def _lbRule(m):
return expr <= getattr(m, slackName)
setattr(m, lbName, Constraint(rule=_lbRule))
def _ubRule(m):
return expr >= -getattr(m, slackName)
setattr(m, ubName, Constraint(rule=_ubRule))
return getattr(m, slackName)
#norm = normL1;
norm = normL2;
#norm = normLInfinity;
def objRule(m):
# return sum(m.beta[foot, bv, ti]**2 for foot in m.feet for bv in m.BV_INDEX for ti in m.t)
# + sum(m.pdd[foot, i, j]**2 for foot in m.feet for i in m.R2_INDEX for j in m.t)
#return sum(m.f[foot, i, j]**2 + m.pd[foot, i, j]**2 + m.pdd[foot, i, j]**2 for foot in m.feet for i in m.R2_INDEX for j in m.t) + sum(m.T[ti]**2 for ti in m.t)
#return sum(m.f[foot, i, j]**2 for foot in m.feet for i in m.R2_INDEX for j in m.t) + sum(m.hipTorque[foot, ti]**2 for foot in m.feet for ti in m.t)
#return sum(m.f[foot, i, j]**2 + m.pd[foot, i, j]**2 + m.pdd[foot, i, j]**2 for foot in m.feet for i in m.R2_INDEX for j in m.t) + sum(m.hipTorque[foot, ti]**2 for foot in m.feet for ti in m.t) + summation(m.dt)
#return sum(m.f[foot, i, j]**2 + m.pdd[foot, i, j]**2 for foot in m.feet for i in m.R2_INDEX for j in m.t) + sum(m.hipTorque[foot, ti]**2 for foot in m.feet for ti in m.t) + summation(m.dt)
footRegionChanges = 0.0
for t in m.t:
if t != m.t[-1]:
for region in m.REGION_INDEX:
if hop.regions[region]['mu'] != 0.:
for foot in m.feet:
current_indicator = getattr(m, 'footRegionConstraints[%d,%s,%d]indicator_var' % (region, foot, t));
next_indicator = getattr(m, 'footRegionConstraints[%d,%s,%d]indicator_var' % (region, foot, t+1));
footRegionChanges += exprNormLInfinity(m, next_indicator - current_indicator, 1.0)
#return footRegionChanges + norm(m, m.pd) + norm(m, m.f) + norm(m, m.hipTorque)
return 1e1*footRegionChanges + norm(m, m.pdd) + norm(m, m.beta) + norm(m, m.hipTorque)
m_nlp.Obj = Objective(rule=objRule, sense=minimize)
m_nlp.rx0 = Constraint(expr=m_nlp.r['x',m_nlp.t[1]] == r0[0]/legLength)
#m_nlp.rz0 = Constraint(expr=m_nlp.r['z',m_nlp.t[1]] <= 1)
m_nlp.th0 = Constraint(expr=m_nlp.th[m_nlp.t[1]] == 0)
m_nlp.vx0 = Constraint(expr=m_nlp.v['x',m_nlp.t[1]] == 0)
m_nlp.vz0 = Constraint(expr=m_nlp.v['z',m_nlp.t[1]] == 0)
m_nlp.w0 = Constraint(expr=m_nlp.w[m_nlp.t[1]] == 0)
m_nlp.Fx0 = Constraint(expr=m_nlp.F['x', m_nlp.t[1]] == 0)
m_nlp.Fz0 = Constraint(expr=m_nlp.F['z', m_nlp.t[1]] == 0)
m_nlp.T0 = Constraint(expr=m_nlp.T[m_nlp.t[1]] == 0)
m_nlp.rxf = Constraint(expr=m_nlp.r['x',m_nlp.t[-1]] >= rf[0]/legLength)
#m_nlp.rzf = Constraint(expr=m_nlp.r['z',m_nlp.t[-1]] == m_nlp.r['z', m_nlp.t[1]])
m_nlp.thf = Constraint(expr=m_nlp.th[m_nlp.t[-1]] == 0)
m_nlp.vxf = Constraint(expr=m_nlp.v['x',m_nlp.t[-1]] == m_nlp.v['x',m_nlp.t[1]])
m_nlp.vzf = Constraint(expr=m_nlp.v['z',m_nlp.t[-1]] == 0)
m_nlp.wf = Constraint(expr=m_nlp.w[m_nlp.t[-1]] == 0)
m_nlp.Fxf = Constraint(expr=m_nlp.F['x', m_nlp.t[-1]] == 0)
m_nlp.Fzf = Constraint(expr=m_nlp.F['z', m_nlp.t[-1]] == 0)
m_nlp.Tf = Constraint(expr=m_nlp.T[m_nlp.t[-1]] == 0)
def _maxVerticalVelocityRule(m, t):
return m.v['z', t] <= 0.5
m_nlp.maxVerticalVelocityConstraint = Constraint(m_nlp.t, rule=_maxVerticalVelocityRule)
def _periodicFootPosition(m, foot, xz):
return m.p[foot, xz, m.t[1]] == m.p[foot, xz, m.t[-1]]
#m_nlp.periodicFootPosition = Constraint(m_nlp.feet, m_nlp.R2_INDEX, rule=_periodicFootPosition)
#m = constructMDTModel(m_nlp, desiredPrecision)
m = constructRelaxedModel(m_nlp)
#for z_data in m.z.values():
#z_data._component().branchPriority = 1
m_nlp_orig = m_nlp.clone()
#m.dt.fix()
#def _momentRule(m, t):
#return m.T[t] == sum(m.footRelativeToCOM[foot,'x',t]*m.f[foot,'z',t] - m.footRelativeToCOM[foot,'z',t]*m.f[foot, 'x',t] for foot in m.feet)
#m_nlp.momentAbountCOM = Constraint(m_nlp.t, rule=_momentRule)
def _hipTorqueRule(m, foot, t):
return m.hipTorque[foot, t] == m.p[foot,'x',t]*m.f[foot,'z',t] - m.p[foot,'z',t]*m.f[foot, 'x',t]
#m_nlp.hipTorqueConstraint = Constraint(m_nlp.feet, m_nlp.t, rule=_hipTorqueRule)
m_nlp.pwSin.deactivate()
m_nlp.pwCos.deactivate()
def _cos(m, t):
return m.cth[t] == cos(m.th[t])
m_nlp.Cos = Constraint(m_nlp.t, rule=_cos)
def _sin(m, t):
return m.sth[t] == sin(m.th[t])
m_nlp.Sin = Constraint(m_nlp.t, rule=_sin)
opt_nlp = SolverFactory('ipopt')
opt_minlp = constructCouenneSolver()
#opt = constructGurobiSolver(mipgap=0.8, MIPFocus=1, TimeLimit=90., Threads=11)
opt = constructGurobiSolver(TimeLimit=480., Threads=11)
#opt = constructGurobiSolver(TimeLimit=50., Threads=11)
hop.constructVisualizer()