def solve(self, example):
'''
Solve problem
Args:
example: example object
Returns:
Results structure
'''
problem = example.cvxpy_problem
if 'verbose' in self._settings:
verbose = self._settings["verbose"]
else:
verbose = False
try:
obj_val = problem.solve(solver=cvxpy.ECOS, verbose=verbose)
except cvxpy.SolverError:
if 'verbose' in self._settings: # if verbose is null, suppress it
if self._settings['verbose']:
print("Error in ECOS solution\n")
return Results(s.SOLVER_ERROR, None, None, None,
None, None)
status = self.STATUS_MAP.get(problem.status, s.SOLVER_ERROR)
# Obtain time and number of iterations
run_time = problem.solver_stats.solve_time \
+ problem.solver_stats.setup_time
niter = problem.solver_stats.num_iters
# Get primal, dual solution
x, y = example.revert_cvxpy_solution()
# Validate status
if not is_qp_solution_optimal(example.qp_problem, x, y,
high_accuracy=self._settings.get('high_accuracy')):
status = s.SOLVER_ERROR
# Validate execution time
if 'time_limit' in self._settings:
if run_time > self._settings['time_limit']:
status = s.TIME_LIMIT
return Results(status,
obj_val,
x,
y,
run_time,
niter)
def solve(self, example):
'''
Solve problem
Args:
problem: problem structure with QP matrices
Returns:
Results structure
'''
problem = example.qp_problem
settings = self._settings.copy()
high_accuracy = settings.pop('high_accuracy', None)
# Setup OSQP
m = osqp.OSQP()
m.setup(problem['P'], problem['q'], problem['A'], problem['l'],
problem['u'],
**settings)
# Solve
results = m.solve()
status = self.STATUS_MAP.get(results.info.status_val, s.SOLVER_ERROR)
if status in s.SOLUTION_PRESENT:
if not is_qp_solution_optimal(problem,
results.x,
results.y,
high_accuracy=high_accuracy):
status = s.SOLVER_ERROR
# Verify solver time
if settings.get('time_limit') is not None:
if results.info.run_time > settings.get('time_limit'):
status = s.TIME_LIMIT
return_results = Results(status,
results.info.obj_val,
results.x,
results.y,
results.info.run_time,
results.info.iter)
return_results.status_polish = results.info.status_polish
return_results.setup_time = results.info.setup_time
return_results.solve_time = results.info.solve_time
return_results.update_time = results.info.update_time
return_results.rho_updates = results.info.rho_updates
return return_results
def solve(self, example):
'''
Solve problem
Args:
example: example object
Returns:
Results structure
'''
p = example.qp_problem
if p['A'] is not None:
# Convert Matrices in CSR format
p['A'] = p['A'].tocsr()
if p['P'] is not None:
# Convert P matrix to COO format
p['P'] = p['P'].tocoo()
# Get problem dimensions
n = p['n']
m = p['m']
# Adjust infinity values in bounds
u = np.copy(p['u'])
l = np.copy(p['l'])
for i in range(m):
if u[i] >= 1e20:
u[i] = grb.GRB.INFINITY
if l[i] <= -1e20:
l[i] = -grb.GRB.INFINITY
# Create a new model
model = grb.Model("qp")
# Add variables
for i in range(n):
model.addVar(ub=grb.GRB.INFINITY, lb=-grb.GRB.INFINITY)
model.update()
x = model.getVars()
# Add inequality constraints: iterate over the rows of A
# adding each row into the model
if p['A'] is not None:
# Ignore newer interface for now
# if hasattr(model, '_v811_addMConstrs'):
# # New interface, all constraints at once
# # Get senses
# senses = np.repeat(grb.GRB.EQUAL, m)
# double_idx = np.array([]) # Array of double constrants for ranges
# for i in range(m):
# # if (np.abs(l[i] - u[i]) < 1e-08):
# # # equal case already handled!
# if (l[i] == -grb.GRB.INFINITY) & (u[i] != grb.GRB.INFINITY):
# senses[i] = grb.GRB.LESS_EQUAL
# elif (l[i] != -grb.GRB.INFINITY) & (u[i] == grb.GRB.INFINITY):
# # Construct sense
#
# else:
# Old interface
for i in range(m):
start = p['A'].indptr[i]
end = p['A'].indptr[i + 1]
variables = [x[j] for j in p['A'].indices[start:end]] # nnz
coeff = p['A'].data[start:end]
expr = grb.LinExpr(coeff, variables)
if (np.abs(l[i] - u[i]) < 1e-08):
model.addConstr(expr, grb.GRB.EQUAL, u[i])
elif (l[i] == -grb.GRB.INFINITY) & (u[i] == grb.GRB.INFINITY):
# Dummy constraint that is always satisfied.
# Gurobi crashes if both constraints in addRange function
# are infinite.
model.addConstr(0. * expr, grb.GRB.LESS_EQUAL, 10.)
else:
model.addRange(expr, lower=l[i], upper=u[i])
# Define objective
if p['P'] is not None:
if hasattr(model, '_v811_setMObjective'):
# New interface for gurobi > v8.1.1
model._v811_setMObjective(0.5 * p['P'], p['q'])
else:
# Old interface
obj = grb.QuadExpr() # Set quadratic part
if p['P'].count_nonzero(
): # If there are any nonzero elms in P
for i in range(p['P'].nnz):
obj.add(.5 * p['P'].data[i] * x[p['P'].row[i]] *
x[p['P'].col[i]])
obj.add(grb.LinExpr(p['q'], x)) # Add linear part
model.setObjective(obj) # Set objective
# Set parameters
if 'verbose' in self._settings: # if verbose is null, suppress it
if self._settings['verbose'] == 0:
model.setParam("OutputFlag", 0)
else:
# If verbose not specified, suppress it as well
model.setParam("OutputFlag", 0)
for param, value in self._settings.items(): # Set other parameters
if (param != "verbose") and (param != "time_limit") \
and (param != "high_accuracy"):
model.setParam(param, value)
# Update model
model.update()
# Solve problem
try:
model.optimize()
except: # Error in the solution
if self._settings['verbose']:
print("Error in GUROBI solution\n")
run_time = model.Runtime
return Results(s.SOLVER_ERROR, None, None, None, run_time, None)
# Get status
status = self.STATUS_MAP.get(model.Status, s.SOLVER_ERROR)
# Get computation time
run_time = model.Runtime
# Total Number of iterations
niter = model.BarIterCount
if status in s.SOLUTION_PRESENT:
# Get objective value
objval = model.objVal
# Get solution
x = np.array([x[i].X for i in range(n)])
# Get dual variables (Gurobi uses swapped signs (-1))
constrs = model.getConstrs()
y = -np.array([constrs[i].Pi for i in range(m)])
if not is_qp_solution_optimal(
p, x, y,
high_accuracy=self._settings.get('high_accuracy')):
status = s.SOLVER_ERROR
# Validate execution time (do not trust commercial solvers)
if 'time_limit' in self._settings:
if run_time > self._settings['time_limit']:
status = s.TIME_LIMIT
return Results(status, objval, x, y, run_time, niter)
else:
return Results(status, None, None, None, run_time, niter)
def solve(self, example):
'''
Solve problem
Args:
example: example object
Returns:
Results structure
'''
p = example.qp_problem
n, m = p['n'], p['m']
if p['P'] is not None:
P = np.ascontiguousarray(p['P'].todense())
if 'A_nobounds' in p:
A = np.ascontiguousarray(p['A_nobounds'].todense())
m = A.shape[0]
elif p['A'] is not None:
A = np.ascontiguousarray(p['A'].todense())
# Define contiguous array vectors
q = np.ascontiguousarray(p['q'])
if 'l_nobounds' in p:
l = np.ascontiguousarray(p['l_nobounds'])
else:
l = np.ascontiguousarray(p['l'])
if 'u_nobounds' in p:
u = np.ascontiguousarray(p['u_nobounds'])
else:
u = np.ascontiguousarray(p['u'])
if 'lx' in p:
lx = np.ascontiguousarray(p['lx'])
else:
lx = np.ascontiguousarray(-np.inf * np.ones(n))
if 'ux' in p:
ux = np.ascontiguousarray(p['ux'])
else:
ux = np.ascontiguousarray(np.inf * np.ones(n))
# Redirect output if verbose is False
if 'verbose' in self._settings:
if self._settings['verbose'] is False:
with stdout_redirected():
qpoases_m = qpoases.PyQProblem(n, m)
else:
qpoases_m = qpoases.PyQProblem(n, m)
else:
# Suppress output also if we do not specify verbose
with stdout_redirected():
qpoases_m = qpoases.PyQProblem(n, m)
options = qpoases.PyOptions()
for param, value in self._settings.items():
if param == 'verbose':
if value is False:
options.printLevel = qpoases.PyPrintLevel.NONE
elif param == 'time_limit':
qpoases_cpu_time = np.array([value])
elif param == 'nWSR':
qpoases_nWSR = np.array([value])
else:
exec("options.%s = %s" % (param, value))
qpoases_m.setOptions(options)
if 'time_limit' not in self._settings:
# Set default to max 10 seconds in runtime
qpoases_cpu_time = np.array([10.])
if 'nWSR' not in self._settings:
# Set default to max 1000000 working set recalculations
qpoases_nWSR = np.array([1000000])
# Solve problem
status = qpoases_m.init(P, q, A, lx, ux, l, u,
qpoases_nWSR, qpoases_cpu_time)
# Check status
status = self.STATUS_MAP.get(status, s.SOLVER_ERROR)
# run_time
run_time = qpoases_cpu_time[0]
# number of iterations
niter = qpoases_nWSR[0]
if status in s.SOLUTION_PRESENT:
x = np.zeros(n)
y_temp = np.zeros(n + m)
obj_val = qpoases_m.getObjVal()
qpoases_m.getPrimalSolution(x)
qpoases_m.getDualSolution(y_temp)
# Change sign
y = -y_temp[n:]
if 'A_nobounds' in p:
# If explicit bounds provided, reconstruct dual variable
# Bounds on all the variables
y_bounds = -y_temp[:n]
# Limit to only the actual bounds
y_bounds = y_bounds[p['bounds_idx']]
y = np.concatenate((y, y_bounds))
if not is_qp_solution_optimal(p, x, y):
status = s.SOLVER_ERROR
return Results(status, obj_val,
x, y,
run_time, niter)
else:
return Results(status, None, None, None,
run_time, niter)
def solve(self, example):
'''
Solve problem
Args:
example: example object
Returns:
Results structure
'''
p = example.qp_problem
# Get problem dimensions
n = p['P'].shape[0]
m = p['A'].shape[0]
'''
Load problem
'''
# Create environment
env = mosek.Env()
# Create optimization task
task = env.Task()
if 'verbose' in self._settings: # if verbose is null, suppress it
if self._settings['verbose']:
# Define a stream printer to grab output from MOSEK
def streamprinter(text):
import sys
sys.stdout.write(text)
sys.stdout.flush()
env.set_Stream(mosek.streamtype.log, streamprinter)
task.set_Stream(mosek.streamtype.log, streamprinter)
# Load problem into task object
# Append 'm' empty constraints.
# The constraints will initially have no bounds.
task.appendcons(m)
# Append 'n' variables.
# The variables will initially be fixed at zero (x=0).
task.appendvars(n)
# Add linear cost by iterating over all variables
for j in range(n):
task.putcj(j, p['q'][j])
task.putvarbound(j, mosek.boundkey.fr, -np.inf, np.inf)
# Add constraints
if p['A'] is not None:
row_A, col_A, el_A = spa.find(p['A'])
task.putaijlist(row_A, col_A, el_A)
for j in range(m):
# Get bounds and keys
u_temp = p['u'][j] if p['u'][j] < 1e20 else np.inf
l_temp = p['l'][j] if p['l'][j] > -1e20 else -np.inf
# Divide 5 cases
if (np.abs(l_temp - u_temp) < 1e-08):
bound_key = mosek.boundkey.fx
elif l_temp == -np.inf and u_temp == np.inf:
bound_key = mosek.boundkey.fr
elif l_temp != -np.inf and u_temp == np.inf:
bound_key = mosek.boundkey.lo
elif l_temp != -np.inf and u_temp != np.inf:
bound_key = mosek.boundkey.ra
elif l_temp == -np.inf and u_temp != np.inf:
bound_key = mosek.boundkey.up
# Add bound
task.putconbound(j, bound_key, l_temp, u_temp)
# Add quadratic cost
if p['P'].count_nonzero(): # If there are any nonzero elms in P
P = spa.tril(p['P'], format='coo')
task.putqobj(P.row, P.col, P.data)
# Set problem minimization
task.putobjsense(mosek.objsense.minimize)
'''
Set parameters
'''
for param, value in self._settings.items():
if param == 'verbose':
if value is False:
self._handle_str_param(task, 'MSK_IPAR_LOG'.strip(), 0)
elif param != 'time_limit':
if isinstance(param, str):
self._handle_str_param(task, param.strip(), value)
else:
self._handle_enum_param(task, param, value)
'''
Solve problem
'''
try:
# Optimization and check termination code
task.optimize()
except:
if self._settings['verbose']:
print("Error in MOSEK solution\n")
return Results(s.SOLVER_ERROR, None, None, None,
None, None)
if 'verbose' in self._settings: # if verbose is null, suppress it
if self._settings['verbose']:
task.solutionsummary(mosek.streamtype.msg)
'''
Extract results
'''
# Get solution type and status
soltype, solsta = self.choose_solution(task)
# Map status using statusmap
status = self.STATUS_MAP.get(solsta, s.SOLVER_ERROR)
# Get statistics
cputime = task.getdouinf(mosek.dinfitem.optimizer_time) + \
task.getdouinf(mosek.dinfitem.presolve_time)
total_iter = task.getintinf(mosek.iinfitem.intpnt_iter)
if status in s.SOLUTION_PRESENT:
# get primal variables values
x = np.zeros(task.getnumvar())
task.getxx(soltype, x)
# get obj value
objval = task.getprimalobj(soltype)
# get dual
y = np.zeros(task.getnumcon())
task.gety(soltype, y)
# it appears signs are inverted
y = -y
if not is_qp_solution_optimal(p, x, y):
status = s.SOLVER_ERROR
# Validate execution time
if 'time_limit' in self._settings:
if cputime > self._settings['time_limit']:
status = s.TIME_LIMIT
return Results(status, objval, x, y,
cputime, total_iter)
else:
return Results(status, None, None, None,
cputime, None)
