# Define problem scaling
obj_scale = -1.0
constr_scale = np.ones(problem.ncons)
var_constr_scale = problem.unpack_constraints(constr_scale)
var_constr_scale['innovation'][:] = 1e3
dec_scale = np.ones(problem.ndec)
var_scale = problem.variables(dec_scale)
var_scale['isRp_tril'][:] = 1e-2
with problem.ipopt(dec_bounds, constr_bounds) as nlp:
nlp.add_str_option('linear_solver', 'ma57')
nlp.add_num_option('ma57_pre_alloc', 10.0)
nlp.add_num_option('tol', 1e-5)
nlp.add_int_option('max_iter', 10000)
nlp.set_scaling(obj_scale, dec_scale, constr_scale)
decopt, info = nlp.solve(dec0)
opt = problem.variables(decopt)
xopt = opt['x']
A = opt['A']
B = opt['B']
C = opt['C']
D = opt['D']
L = opt['L']
W = np.diag(opt['W_diag'])
ybias = opt['ybias']
isRp = symfem.tril_mat(ny, opt['isRp_tril'])
yopt = xopt @ C.T + u @ D.T + ybias
eopt = opt['e']
decopt, info = nlp.solve(dec0)
opt = problem.variables(decopt)
xopt = opt['x']
Ac = opt['Ac']
Bc = opt['Bc']
A = opt['A']
B = opt['B']
C = opt['C']
D = opt['D']
Ln = opt['Ln']
Kn = opt['Kn']
ybias = opt['ybias']
pred_orth = opt['pred_orth']
corr_orth = opt['corr_orth']
sRp = symfem.tril_mat(opt['sRp_tril'])
sPp = symfem.tril_mat(opt['sPp_tril'])
sPc = symfem.tril_mat(opt['sPc_tril'])
sQ = symfem.tril_mat(opt['sQ_tril'])
sQc = symfem.tril_mat(opt['sQc_tril'])
sR = symfem.tril_mat(opt['sR_tril'])
yopt = xopt @ C.T + u @ D.T + ybias
enopt = opt['en']
Pc = sPc @ sPc.T
Pp = sPp @ sPp.T
Rp = sRp @ sRp.T
Q = sQ @ sQ.T
Qc = sQc @ sQc.T
R = sR @ sR.T
# Define problem scaling
obj_scale = -1.0
constr_scale = np.ones(problem.ncons)
var_constr_scale = problem.unpack_constraints(constr_scale)
var_constr_scale['innovation'][:] = 1e2
dec_scale = np.ones(problem.ndec)
var_scale = problem.variables(dec_scale)
var_scale['sRp_tril'][:] = 1e2
var_scale['Ln'][:] = 1e2
with problem.ipopt(dec_bounds, constr_bounds) as nlp:
nlp.add_str_option('linear_solver', 'ma57')
nlp.add_num_option('ma57_pre_alloc', 25.0)
nlp.add_num_option('tol', 1e-6)
nlp.add_int_option('max_iter', 1000)
nlp.set_scaling(obj_scale, dec_scale, constr_scale)
decopt, info = nlp.solve(dec0)
opt = problem.variables(decopt)
xopt = opt['x']
A = opt['A']
B = opt['B']
C = opt['C']
D = opt['D']
Ln = opt['Ln']
ybias = opt['ybias']
sRp = symfem.tril_mat(opt['sRp_tril'])
yopt = xopt @ C.T + u @ D.T + ybias
enopt = opt['en']
opt = problem.variables(decopt)
xopt = opt['x']
Ac = opt['Ac']
Bc = opt['Bc']
Qc = opt['Qc']
A = opt['A']
B = opt['B']
C = opt['C']
D = opt['D']
L = opt['L']
Kn = opt['Kn']
ybias = opt['ybias']
pred_orth = opt['pred_orth']
corr_orth = opt['corr_orth']
sR = symfem.tril_mat(ny, opt['sR_tril'])
sRp = symfem.tril_mat(ny, opt['sRp_tril'])
isRp = symfem.tril_mat(ny, opt['isRp_tril'])
sPp = symfem.tril_mat(nx, opt['sPp_tril'])
sPc = symfem.tril_mat(nx, opt['sPc_tril'])
sQ = symfem.tril_mat(nx, opt['sQ_tril'])
yopt = xopt @ C.T + u @ D.T + ybias
eopt = opt['e']
Pc = sPc @ sPc.T
Pp = sPp @ sPp.T
Rp = sRp @ sRp.T
Q = sQ @ sQ.T
R = sR @ sR.T