def main():
filename = './data/hashcode.in'
print('Reading simulation...')
sim = Simulation(filename)
street_ids_with_cars = set()
for car in sim.cars:
street_ids_with_cars.update([s.id for s in car.path[:-1]])
for isect in sim.intersections:
isect.set_schedule(
[(s, 1) for s in isect.queues.keys() if s in street_ids_with_cars],
sim.duration)
time_beg = datetime.now()
score, arrivals = sim.simulate()
time_end = datetime.now()
print(time_end - time_beg)
print(f'score: {score}')
sim.write_schedule('./data/submission.csv')
def run(probability, samples=1, turning=False, fname=None):
# Run simulation for each radius
average_collisions = []
for i, sample in enumerate(range(samples)):
print '\tSim {} of {}'.format(i + 1, samples)
sim = Simulation(probability, turning=turning)
sim.run(animate=False, fname=fname)
average_collisions.append(sim.average_collisions())
return float(sum(average_collisions)) / len(average_collisions)
def run(netdef,tosave,modify,procs,thisProc,stims,param,repeats,sim_time,SaveSpikes,SaveVoltage,SaveConductance,SaveCurrent):
net = netdef()
if SaveVoltage:
net.recordVoltage()
repeats = int(repeats)
# Randomseed was 200 for most figures
# Changed to 200 for rat
# Changed to 200 for anurans
s = Simulation(net, randomseed=202,delay=25)
s.verbose = False
s.sim_time = sim_time
s.dt = 0.050
total = len(stims)*len(param)*repeats
spp = ceil(float(total)/procs)
start = thisProc*spp
end = (thisProc+1)*spp
count = 0
for a in param:
s.set_amplitude(net.sim_amp)
for d in stims*repeats:
if count >= start and count < end:
net = modify(net,a,d)
progress.update(count-start,spp,thisProc)
s.stim_dur = d
s.run()
key = [a,d]
net.savecells(tosave, key, spikes=SaveSpikes,voltage=SaveVoltage,conductance=SaveConductance,current=SaveCurrent)
count += 1
progress.update(spp,spp,thisProc)
r = [thisProc,net.savedparams,net.savedcells]
return r
def main():
# Read arguments
args = parser().parse_args()
seed(args.seed)
# Run once
if not args.run_all:
sim = Simulation(args.probability, turning=args.turning)
sim.run(animate=args.animate, fname=args.output_file)
print 'collisions = {}'.format(sim.average_collisions())
# Run all
else:
probabilities = np.linspace(0.04, 0.2, 17)
run_all(probabilities, turning=args.turning, fname=args.output_file)
def Main (args):
show_converge = True
if len(args) != 10:
print >>sys.stderr, "Usage: perturbation_constant.py setup stable_time mean_recovery end_time " + \
" begin_mean_perturb end_mean_perturn step_mean_perturb sampling_rate seed bootstrap"
else:
topo = open(args[0]).read()
stable = float(args[1])
mean_recovery = float(args[2])
end_time = float(args[3])
begin = float(args[4])
end = float(args[5])
step = float(args[6])
sampling_rate = float(args[7])
seed = int(args[8])
bootstrap = (args[9].lower() == "true")
print "Setting %s %f %f %f %f %f %f %f %d"%(args[0], stable, mean_recovery, end_time, begin, end,\
step, sampling_rate, seed)
topo_yaml = yaml.load(topo)
# If no fail links then just use links
links = topo_yaml['links']
if 'fail_links' in topo_yaml:
fail_links = topo_yaml['fail_links']
else:
fail_links = links
for mean in np.arange(begin, end, step):
Singleton.clear()
sim = Simulation()
sim.check_always = False
random.seed(seed)
print "mean_perturb %f"%(mean)
print "generating trace"
(end_time, new_trace) = TransformTrace(links, fail_links, mean, mean_recovery, stable, end_time, bootstrap)
print "done generating trace"
print "TRACE TRACE TRACE"
for t in new_trace:
print t
print "TRACE TRACE TRACE"
sim.Setup(topo, new_trace, False)
for time in np.arange(stable, end_time, sampling_rate):
sim.scheduleCheck(time)
## Measure latency less often
#for time in np.arange(stable, end_time, sampling_rate):
#for (ha, hb) in permutations(sim.hosts, 2):
#sim.scheduleSend(time, ha.name, ha.address, hb.address)
sim.Run()
sim.Report(show_converge)
sim.Clear()
def uploadData(setData):
try:
mySimulation = Simulation()
mySimulation.setUp(dataUploadStorage[session["user_id"]]["data"])
mySimulation.runNewSim()
except:
#print("Simulation failed.")
pass
thesePaths = mySimulation.getAllPaths()
theseHeatMaps = mySimulation.getAllHeatData()
thisFileName = str(setData[0])
rows = str(setData[1])
cols = str(setData[2])
getLocations = parseLocations(rows, cols, setData[3])
getSubjectMap = parseSubjectMap(setData[4])
db = connectToDB()
cur = db.cursor(cursor_factory=psycopg2.extras.DictCursor)
cur.execute(
"INSERT INTO datasets (datasetname, userid, heatdata, vectordata, locationmap, subjectmap) VALUES (%s, %s, %s, %s, %s, %s)",
(thisFileName, 1, json.dumps(theseHeatMaps), json.dumps(thesePaths),
json.dumps(getLocations), json.dumps(getSubjectMap)))
db.commit()
cur.close()
db.close()
session["currentlyUploading"] = False
del dataUploadStorage[session["user_id"]]
session["uploadingFileName"] = ""
emit('finishedUploading')
def init(screen_size):
screen = init_screen(screen_size)
images = {}
images['star0'] = pygame.image.load('img/star0.png')
images['star0'].set_colorkey((0, 0, 0))
images['star1'] = pygame.image.load('img/star0.png')
images['star1'].set_colorkey((0, 0, 0))
images['star0yellow'] = colored_copy(images['star0'], (0, 255, 255, 0))
sim = Simulation()
sim.draw_stars(images, screen)
return screen, images, sim
def run_ratio(ratio, radii, tracking, coverage=0.1, T=120, dt=1, fname=None):
'''Runs the specified target-robot ratio for the specified environment radii'''
# Run simulation for each radius
observations = []
for i, R in enumerate(radii):
m_max, n_max = 10, 20
if ratio <= 1:
m = m_max
n = int(ratio * m_max)
else:
n = n_max
m = int(n_max / ratio)
print '\tSim {} of {}: n = {}, m = {}'.format(i + 1, len(radii), n, m)
sim = Simulation(m, n, T, dt, R, tracking=tracking)
sim.run(vis='', fname=fname)
observations.append(sim.average_observations(normalize=True))
return radii, observations
def quick_match(self, teams):
home = teams[0]
away = teams[1]
print('Selected teams: \n {} vs {}'.format(home.name, away.name))
input('press enter to continue')
result = Simulation.simulate(teams)
print(' {} vs {}\n {} - {}'.format(result[0].name, result[1].name,
result[0].result, result[1].result))
def main():
# Create first generation
pool = Pop(size_gen)
# Step through generations
for _ in range(num_gen):
# initialize data histogram for sim visualization
hist = Data()
terrain = Terrain(400, 4)
# create pybox2d dynamic bodies based on individual's gene
gen = [Biped(pool.population[i].name, pool.population[i].gene)
for i in range(size_gen)]
for k, biped in enumerate(gen):
# specify terrain and biped to race
race = Sim(terrain, biped)
if not hist.terrain:
hist.set_terrain(terrain)
# run simulation without visualization
score, bb, time = race.run(1e4)
# store sim data and fitness evaluation
pool.population[k].fitness = score
hist.timelines[biped.name] = race.history.timelines['timeline']
# resort gene pool
pool.population = list(
sorted(pool.population, key=lambda x: x.fitness))
# visualize top bipeds' simulations
shown = pool.population[-num_shown:]
timelines = [s.name for s in shown]
view.start()
view.run(hist, timelines, speed=3)
# evolve gene pool
pool.evolve()
def main():
# Read arguments
args = parser().parse_args()
seed(args.seed)
# Run once
if not args.run_all:
sim = Simulation(args.m,
args.n,
args.t,
args.dt,
args.r,
tracking=args.tracking)
sim.run(vis=args.visualization, fname=args.output_file)
print 'observations = {}'.format(
sim.average_observations(normalize=True))
# Run ratios
else:
ratios = [1 / 5., 1 / 2., 1, 4, 10]
run_ratios(ratios, tracking=args.tracking, fname=args.output_file)
def Main (args):
show_converge = True
if len(args) != 7:
print >>sys.stderr, "Usage: perturbation.py setup trace stable_time begin_mean_perturb end_mean_perturn step_mean_perturb seed"
else:
topo = open(args[0]).read()
trace = open(args[1]).readlines()
stable = float(args[2])
begin = float(args[3])
end = float(args[4])
step = float(args[5])
seed = int(args[6])
print "Setting %s %s %f %f %f %f %d"%(args[0], args[1], stable, begin, end, step, seed)
for mean in np.arange(begin, end, step):
Singleton.clear()
sim = Simulation()
sim.check_always = False
random.seed(seed)
print "mean_perturb %f"%(mean)
(end_time, new_trace) = TransformTrace(trace, mean, stable)
sim.Setup(topo, new_trace, True)
for time in np.arange(stable, end_time, 0.5 * mean):
sim.scheduleCheck(time)
# Measure latency less often
for time in np.arange(stable, end_time, mean):
for (ha, hb) in permutations(sim.hosts, 2):
sim.scheduleSend(time, ha.name, ha.address, hb.address)
sim.Run()
sim.Report(show_converge)
sim.Clear()
def _set_optimization_variables_initials(self, qinit, x0, uinit):
self.simulation = Simulation(self._discretization.system, \
self._pdata, qinit)
self.simulation.run_system_simulation(x0, \
self._discretization.time_points, uinit, print_status = False)
xinit = self.simulation.simulation_results
repretitions_xinit = \
self._discretization.optimization_variables["X"][:,:-1].shape[1] / \
self._discretization.number_of_intervals
Xinit = ci.repmat(xinit[:, :-1], repretitions_xinit, 1)
Xinit = ci.horzcat([ \
Xinit.reshape((self._discretization.system.nx, \
Xinit.numel() / self._discretization.system.nx)),
xinit[:, -1],
])
uinit = inputchecks.check_controls_data(uinit, \
self._discretization.system.nu, \
self._discretization.number_of_intervals)
Uinit = uinit
qinit = inputchecks.check_constant_controls_data(qinit, \
self._discretization.system.nq)
Qinit = qinit
self._optimization_variables_initials = ci.veccat([ \
Uinit,
Qinit,
Xinit,
])
def next_game(self):
if self.fixtures != []:
for game in self.fixtures[0]:
home = Club(game[0])
away = Club(game[1])
if home.name == self.club.name or away.name == self.club.name:
print(game)
input('press enter to continue')
teams = [home, away]
result = Simulation.simulate(teams)
self.competition._update_standings(result)
self.played.append([(result[0].name, result[1].name),
(result[0].result, result[1].result)])
self.fixtures.pop(0)
else:
print('End of the season')
self.menu()
def sim_process(sim_queue, client_queue):
sim = Simulation('../model/Cirship.fmu')
print('Simulation ready')
id = 0
watches = {
'real': set(),
'bool': set(),
}
while True:
while not client_queue.empty():
msg = client_queue.get_nowait()
if msg['type'] == 'set':
ref = msg['ref']
value = msg['value']
if isinstance(value, bool):
sim.set_bool(ref, value)
else:
sim.set_real(ref, value)
elif msg['type'] == 'watch':
for ref in msg.get('real', []):
watches['real'].add(ref)
for ref in msg.get('bool', []):
watches['bool'].add(ref)
t = sim.update()
id += 1
reals = {ref: sim.get_real(ref) for ref in watches['real']}
bools = {ref: sim.get_bool(ref) for ref in watches['bool']}
refs = {'id': id, 'time': t, **reals, **bools}
sim_queue.put(refs)
time.sleep(0.05)
def _set_optimization_variables_initials(self, qinit, x0, uinit):
self.simulation = Simulation(self._discretization.system, \
self._pdata, qinit)
self.simulation.run_system_simulation(x0, \
self._discretization.time_points, uinit, print_status = False)
xinit = self.simulation.simulation_results
repretitions_xinit = \
self._discretization.optimization_variables["X"][:,:-1].shape[1] / \
self._discretization.number_of_intervals
Xinit = ci.repmat(xinit[:, :-1], repretitions_xinit, 1)
Xinit = ci.horzcat([ \
Xinit.reshape((self._discretization.system.nx, \
Xinit.size() / self._discretization.system.nx)),
xinit[:, -1],
])
uinit = inputchecks.check_controls_data(uinit, \
self._discretization.system.nu, \
self._discretization.number_of_intervals)
Uinit = uinit
qinit = inputchecks.check_constant_controls_data(qinit, \
self._discretization.system.nq)
Qinit = qinit
self._optimization_variables_initials = ci.veccat([ \
Uinit,
Qinit,
Xinit,
])
def simulate(ind):
return Simulation().simulate(ind)
class DoE(DoEProblem):
'''The class :class:`casiopeia.doe.DoE` is used to set up
Design-of-Experiments-problems for systems defined with the
:class:`casiopeia.system.System` class.
The aim of the experimental design optimization is to identify a set of
controls that can be used for the generation of measurement data which
allows for a better estimation of the unknown parameters of a system.
To achieve this, an information function on the covariance matrix of the
estimated parameters is minimized. The values of the estimated parameters,
though they are mostly an initial
guess for their values, are not changed during the optimization.
Optimum experimental design and parameter estimation methods can be used
interchangeably until a desired accuracy of the parameters has been
achieved.
'''
@property
def optimized_controls(self):
return self.design_results["x"][ \
:(self._discretization.number_of_intervals * \
self._discretization.system.nu)]
def _discretize_system(self, system, time_points, discretization_method, \
**kwargs):
if system.nx == 0:
self._discretization = NoDiscretization(system, time_points)
elif system.nx != 0:
if discretization_method == "collocation":
self._discretization = ODECollocation( \
system, time_points, **kwargs)
elif discretization_method == "multiple_shooting":
self._discretization = ODEMultipleShooting( \
system, time_points, **kwargs)
else:
raise NotImplementedError('''
Unknown discretization method: {0}.
Possible values are "collocation" and "multiple_shooting".
'''.format(str(discretization_method)))
def _set_parameter_guess(self, pdata):
self._pdata = inputchecks.check_parameter_data(pdata, \
self._discretization.system.np)
def _apply_parameters_to_equality_constraints(self):
optimization_variables_for_equality_constraints = ci.veccat([ \
self._discretization.optimization_variables["U"],
self._discretization.optimization_variables["Q"],
self._discretization.optimization_variables["X"],
self._discretization.optimization_variables["EPS_U"],
self._discretization.optimization_variables["P"],
])
optimization_variables_parameters_applied = ci.veccat([ \
self._discretization.optimization_variables["U"],
self._discretization.optimization_variables["Q"],
self._discretization.optimization_variables["X"],
ci.mx(*self._discretization.optimization_variables["EPS_U"].shape),
self._pdata,
])
equality_constraints_fcn = ci.mx_function( \
"equality_constraints_fcn", \
[optimization_variables_for_equality_constraints], \
[self._discretization.equality_constraints])
[self._equality_constraints_parameters_applied] = \
equality_constraints_fcn([optimization_variables_parameters_applied])
def _apply_parameters_to_discretization(self):
self._apply_parameters_to_equality_constraints()
def _set_optimization_variables(self):
self._optimization_variables = ci.veccat([ \
self._discretization.optimization_variables["U"],
self._discretization.optimization_variables["Q"],
self._discretization.optimization_variables["X"],
])
def _set_optimization_variables_initials(self, qinit, x0, uinit):
self.simulation = Simulation(self._discretization.system, \
self._pdata, qinit)
self.simulation.run_system_simulation(x0, \
self._discretization.time_points, uinit, print_status = False)
xinit = self.simulation.simulation_results
repretitions_xinit = \
self._discretization.optimization_variables["X"][:,:-1].shape[1] / \
self._discretization.number_of_intervals
Xinit = ci.repmat(xinit[:, :-1], repretitions_xinit, 1)
Xinit = ci.horzcat([ \
Xinit.reshape((self._discretization.system.nx, \
Xinit.size() / self._discretization.system.nx)),
xinit[:, -1],
])
uinit = inputchecks.check_controls_data(uinit, \
self._discretization.system.nu, \
self._discretization.number_of_intervals)
Uinit = uinit
qinit = inputchecks.check_constant_controls_data(qinit, \
self._discretization.system.nq)
Qinit = qinit
self._optimization_variables_initials = ci.veccat([ \
Uinit,
Qinit,
Xinit,
])
def _set_optimization_variables_lower_bounds(self, umin, qmin, xmin, x0):
umin_user_provided = umin
umin = inputchecks.check_controls_data(umin, \
self._discretization.system.nu, 1)
if umin_user_provided is None:
umin = -np.inf * np.ones(umin.shape)
Umin = ci.repmat(umin, 1, \
self._discretization.optimization_variables["U"].shape[1])
qmin_user_provided = qmin
qmin = inputchecks.check_constant_controls_data(qmin, \
self._discretization.system.nq)
if qmin_user_provided is None:
qmin = -np.inf * np.ones(qmin.shape)
Qmin = qmin
xmin_user_provided = xmin
xmin = inputchecks.check_states_data(xmin, \
self._discretization.system.nx, 0)
if xmin_user_provided is None:
xmin = -np.inf * np.ones(xmin.shape)
Xmin = ci.repmat(xmin, 1, \
self._discretization.optimization_variables["X"].shape[1])
Xmin[:,0] = x0
self._optimization_variables_lower_bounds = ci.veccat([ \
Umin,
Qmin,
Xmin,
])
def _set_optimization_variables_upper_bounds(self, umax, qmax, xmax, x0):
umax_user_provided = umax
umax = inputchecks.check_controls_data(umax, \
self._discretization.system.nu, 1)
if umax_user_provided is None:
umax = np.inf * np.ones(umax.shape)
Umax = ci.repmat(umax, 1, \
self._discretization.optimization_variables["U"].shape[1])
qmax_user_provided = qmax
qmax = inputchecks.check_constant_controls_data(qmax, \
self._discretization.system.nq)
if qmax_user_provided is None:
qmax = -np.inf * np.ones(qmax.shape)
Qmax = qmax
xmax_user_provided = xmax
xmax = inputchecks.check_states_data(xmax, \
self._discretization.system.nx, 0)
if xmax_user_provided is None:
xmax = np.inf * np.ones(xmax.shape)
Xmax = ci.repmat(xmax, 1, \
self._discretization.optimization_variables["X"].shape[1])
Xmax[:,0] = x0
self._optimization_variables_upper_bounds = ci.veccat([ \
Umax,
Xmax,
])
def _set_measurement_data(self):
# The DOE problem does not depend on actual measurement values,
# the measurement deviations are only needed to set up the objective;
# therefore, dummy-values for the measurements can be used
# (see issue #7 for further information)
measurement_data = np.zeros((self._discretization.system.nphi, \
self._discretization.number_of_intervals + 1))
self._measurement_data_vectorized = ci.vec(measurement_data)
def _set_weightings(self, wv, weps_u):
input_error_weightings = \
inputchecks.check_input_error_weightings(weps_u, \
self._discretization.system.neps_u,
self._discretization.number_of_intervals)
measurement_weightings = \
inputchecks.check_measurement_weightings(wv, \
self._discretization.system.nphi, \
self._discretization.number_of_intervals + 1)
self._weightings_vectorized = ci.veccat([ \
input_error_weightings,
measurement_weightings,
])
def _set_measurement_deviations(self):
self._measurement_deviations = ci.vertcat([ \
ci.vec(self._discretization.measurements) - \
self._measurement_data_vectorized + \
ci.vec(self._discretization.optimization_variables["V"])
])
def _setup_constraints(self):
self._constraints = ci.vertcat([ \
self._measurement_deviations,
self._discretization.equality_constraints,
])
def _set_cov_matrix_derivative_directions(self):
# These correspond to the optimization variables of the parameter
# estimation problem; the evaluation of the covariance matrix, though,
# does not depend on the actual values of V, EPS_E and EPS_U, and with
# this, the DoE problem does not
self._cov_matrix_derivative_directions = ci.veccat([ \
self._discretization.optimization_variables["P"],
self._discretization.optimization_variables["X"],
self._discretization.optimization_variables["EPS_U"],
self._discretization.optimization_variables["V"],
])
def _setup_gauss_newton_lagrangian_hessian(self):
gauss_newton_lagrangian_hessian_diag = ci.vertcat([ \
ci.mx(self._cov_matrix_derivative_directions.shape[0] - \
self._weightings_vectorized.shape[0], 1), \
self._weightings_vectorized])
self._gauss_newton_lagrangian_hessian = ci.diag( \
gauss_newton_lagrangian_hessian_diag)
def _setup_covariance_matrix_for_evaluation(self):
covariance_matrix_free_variables = ci.veccat([ \
self._discretization.optimization_variables["P"],
self._discretization.optimization_variables["U"],
self._discretization.optimization_variables["Q"],
self._discretization.optimization_variables["X"],
self._discretization.optimization_variables["EPS_U"],
])
self._covariance_matrix_fcn = ci.mx_function("covariance_matrix_fcn", \
[covariance_matrix_free_variables], \
[self._covariance_matrix.covariance_matrix])
def _apply_parameters_to_objective(self):
# As mentioned above, the objective does not depend on the actual
# values of V, but on the values of P and EPS_E and EPS_U, while
# P is fed from pdata, and EPS_E, EPS_u are supposed to be 0
objective_free_variables = ci.veccat([ \
self._discretization.optimization_variables["P"],
self._discretization.optimization_variables["U"],
self._discretization.optimization_variables["Q"],
self._discretization.optimization_variables["X"],
self._discretization.optimization_variables["EPS_U"],
])
objective_free_variables_parameters_applied = ci.veccat([ \
self._pdata,
self._discretization.optimization_variables["U"],
self._discretization.optimization_variables["Q"],
self._discretization.optimization_variables["X"],
ci.mx(*self._discretization.optimization_variables["EPS_U"].shape),
])
objective_fcn = ci.mx_function("objective_fcn", \
[objective_free_variables], [self._objective_parameters_free])
[self._objective] = objective_fcn( \
[objective_free_variables_parameters_applied])
def __init__(self, system, time_points, \
uinit = None, umin = None, umax = None, \
qinit = None, qmin = None, qmax = None, \
pdata = None, x0 = None, \
xmin = None, xmax = None, \
wv = None, weps_u = None, \
discretization_method = "collocation", \
optimality_criterion = "A", **kwargs):
r'''
:raises: AttributeError, NotImplementedError
:param system: system considered for parameter estimation, specified
using the :class:`casiopeia.system.System` class
:type system: casiopeia.system.System
:param time_points: time points :math:`t_\text{N} \in \mathbb{R}^\text{N}`
used to discretize the continuous time problem. Controls
will be applied at the first :math:`N-1` time points,
while measurements take place at all :math:`N` time points.
:type time_points: numpy.ndarray, casadi.DMatrix, list
:param umin: optional, lower bounds of the time-varying controls
:math:`u_\text{min} \in \mathbb{R}^{\text{n}_\text{u}}`;
if not values are given, :math:`-\infty` will be used
:type umin: numpy.ndarray, casadi.DMatrix
:param umax: optional, upper bounds of the time-vaying controls
:math:`u_\text{max} \in \mathbb{R}^{\text{n}_\text{u}}`;
if not values are given, :math:`\infty` will be used
:type umax: numpy.ndarray, casadi.DMatrix
:param uinit: optional, initial guess for the values of the time-varying controls
:math:`u_\text{N} \in \mathbb{R}^{\text{n}_\text{u} \times \text{N}-1}`
that (might) change at the switching time points;
if no values are given, 0 will be used; note that a poorly
or wrongly chosen initial guess can cause the optimization
to fail, and note that the
the second dimension of :math:`u_N` is :math:`N-1` and not
:math:`N`, since there is no control value applied at the
last time point
:type uinit: numpy.ndarray, casadi.DMatrix
:param qmin: optional, lower bounds of the time-constant controls
:math:`q_\text{min} \in \mathbb{R}^{\text{n}_\text{q}}`;
if not values are given, :math:`-\infty` will be used
:type qmin: numpy.ndarray, casadi.DMatrix
:param qmax: optional, upper bounds of the time-constant controls
:math:`q_\text{max} \in \mathbb{R}^{\text{n}_\text{q}}`;
if not values are given, :math:`\infty` will be used
:type qmax: numpy.ndarray, casadi.DMatrix
:param qinit: optional, initial guess for the optimal values of the
time-constant controls
:math:`q_\text{init} \in \mathbb{R}^{\text{n}_\text{q}}`;
if not values are given, 0 will be used; note that a poorly
or wrongly chosen initial guess can cause the optimization
to fail
:type qinit: numpy.ndarray, casadi.DMatrix
:param pdata: values of the time-constant parameters
:math:`p \in \mathbb{R}^{\text{n}_\text{p}}`
:type pdata: numpy.ndarray, casadi.DMatrix
:param x0: state values :math:`x_0 \in \mathbb{R}^{\text{n}_\text{x}}`
at the first time point :math:`t_0`
:type x0: numpy.ndarray, casadi.DMatrix, list
:param xmin: optional, lower bounds of the states
:math:`x_\text{min} \in \mathbb{R}^{\text{n}_\text{x}}`;
if no value is given, :math:`-\infty` will be used
:type xmin: numpy.ndarray, casadi.DMatrix
:param xmax: optional, lower bounds of the states
:math:`x_\text{max} \in \mathbb{R}^{\text{n}_\text{x}}`;
if no value is given, :math:`\infty` will be used
:type xmax: numpy.ndarray, casadi.DMatrix
:param wv: weightings for the measurements
:math:`w_\text{v} \in \mathbb{R}^{\text{n}_\text{y} \times \text{N}}`
:type wv: numpy.ndarray, casadi.DMatrix
:param weps_u: weightings for the input errors
:math:`w_{\epsilon_\text{u}} \in \mathbb{R}^{\text{n}_{\epsilon_\text{u}}}`
(only necessary
if input errors are used within ``system``)
:type weps_u: numpy.ndarray, casadi.DMatrix
:param discretization_method: optional, the method that shall be used for
discretization of the continuous time
problem w. r. t. the time points given
in :math:`t_\text{N}`; possible values are
"collocation" (default) and
"multiple_shooting"
:type discretization_method: str
:param optimality_criterion: optional, the information function
:math:`I_\text{X}(\cdot)` to be used on the
covariance matrix, possible values are
`A` (default) and `D`, while
.. math ::
\begin{aligned}
I_\text{A}(\Sigma_\text{p}) & = \frac{1}{n_\text{p}} \text{Tr}(\Sigma_\text{p}),\\
I_\text{D}(\Sigma_\text{p}) & = \begin{vmatrix} \Sigma_\text{p} \end{vmatrix} ^{\frac{1}{n_\text{p}}},
\end{aligned}
for further information see e. g. [#f1]_
:type optimality_criterion: str
Depending on the discretization method specified in
`discretization_method`, the following parameters can be used
for further specification:
:param collocation_scheme: optional, scheme used for setting up the
collocation polynomials,
possible values are `radau` (default)
and `legendre`
:type collocation_scheme: str
:param number_of_collocation_points: optional, order of collocation
polynomials
:math:`d \in \mathbb{Z}` (default
values is 3)
:type number_of_collocation_points: int
:param integrator: optional, integrator to be used with multiple shooting.
See the CasADi documentation for a list of
all available integrators. As a default, `cvodes`
is used.
:type integrator: str
:param integrator_options: optional, options to be passed to the CasADi
integrator used with multiple shooting
(see the CasADi documentation for a list of
all possible options)
:type integrator_options: dict
You do not need to specify initial guesses for the estimated states,
since these are obtained with a system simulation using the initial
states and the provided initial guesses for the controls.
The resulting optimization problem has the following form:
.. math::
\begin{aligned}
\text{arg}\,\underset{u, q, x}{\text{min}} & & I(\Sigma_{\text{p}}(x, u, q; p)) &\\
\text{subject to:} & & g(x, u, q; p) & = 0\\
& & u_\text{min} \leq u_\text{k} & \leq u_\text{max} \hspace{1cm} k = 1, \dots, N-1\\
& & x_\text{min} \leq x_\text{k} & \leq x_\text{max} \hspace{1cm} k = 1, \dots, N\\
& & x_1 \leq x(t_1) & \leq x_1
\end{aligned}
where :math:`\Sigma_p = \text{Cov}(p)` and :math:`g(\cdot)` contains the
discretized system dynamics
according to the specified discretization method. If the system is
non-dynamic, it only contains the user-provided equality constraints.
.. rubric:: References
.. [#f1] |linkf1|_
.. _linkf1: http://ginger.iwr.uni-heidelberg.de/vplan/images/5/54/Koerkel2002.pdf
.. |linkf1| replace:: *Körkel, Stefan: Numerische Methoden für Optimale Versuchsplanungsprobleme bei nichtlinearen DAE-Modellen, PhD Thesis, Heidelberg university, 2002, pages 74/75.*
'''
intro()
self._discretize_system( \
system, time_points, discretization_method, **kwargs)
self._set_parameter_guess(pdata)
self._apply_parameters_to_discretization()
self._set_optimization_variables()
self._set_optimization_variables_initials(qinit, x0, uinit)
self._set_optimization_variables_lower_bounds(umin, qmin, xmin, x0)
self._set_optimization_variables_upper_bounds(umax, qmax, xmax, x0)
self._set_measurement_data()
self._set_weightings(wv, weps_u)
self._set_measurement_deviations()
self._set_cov_matrix_derivative_directions()
self._setup_constraints()
self._setup_gauss_newton_lagrangian_hessian()
self._setup_covariance_matrix()
self._setup_covariance_matrix_for_evaluation()
self._set_optimiality_criterion(optimality_criterion)
self._setup_objective()
self._apply_parameters_to_objective()
self._setup_nlp()
def _print_experimental_properties(self, covariance_matrix):
np.set_printoptions(linewidth = 200, \
formatter={'float': lambda x: format(x, ' 10.8e')})
print("\nParameters p_i:")
for k, pk in enumerate(self._pdata):
print(" p_{0:<3} = {1} +/- {2}".format( \
k, pk, ci.sqrt(abs(ci.diag(covariance_matrix)[k]))))
print("\nCovariance matrix for this setup:")
print(np.atleast_2d(covariance_matrix))
def _compute_initial_covariance_matrix(self):
covariance_matrix_initial_input = ci.veccat([ \
self._pdata,
self._optimization_variables_initials,
np.zeros(self._discretization.optimization_variables["EPS_U"].shape)
])
self._covariance_matrix_initial = self._covariance_matrix_fcn( \
[covariance_matrix_initial_input])[0]
def _compute_optimized_covariance_matrix(self):
covariance_matrix_optimized_input = ci.veccat([ \
self._pdata,
self.design_results["x"],
np.zeros(self._discretization.optimization_variables["EPS_U"].shape)
])
self._covariance_matrix_optimized = self._covariance_matrix_fcn( \
[covariance_matrix_optimized_input])[0]
def plot_confidence_ellipsoids(self, properties = "initial"):
r'''
:param properties: Set whether the experimental properties for the
initial setup ("initial", default), the optimized setup
("optimized") or for both setups ("all") shall be
plotted. In the later case, both ellipsoids for one
pair of parameters will be displayed within one plot.
:type properties: str
Plot confidence ellipsoids for all parameter pairs.
Since the number of plots is possibly big, all plots will be saved
within a folder *confidence_ellipsoids_scriptname* in you current
working directory rather than being displayed directly.
'''
self._plot_confidence_ellipsoids(pdata = self._pdata, \
properties = properties)
'--motion-noise',
dest='m_noise',
required=False,
type=float,
help='The amount of motion noise to add to the particle filter.')
parser.add_argument(
'--ignore-regions',
dest='ignore_regions',
action='store_true',
help='Set this flag to disable using the region probabilities.')
args = parser.parse_args()
config = get_pf_config(args.config_file)
building_map = BuildingMap(args.map_data)
# If this is a feed generator run, start the user simulator.
if args.make_feed:
simulation = Simulation(building_map, args.feed)
w = DisplayWindow(building_map, args.map_image, sim=simulation)
w.start_make_feed()
# Otherwise, run the particle filter on an existing feed.
else:
display_on = False if args.no_disp else True
c_noise = args.c_noise if args.c_noise else 0
m_noise = args.m_noise if args.m_noise else 0
feed_processor = FeedProcessor(args.feed,
args.loop_feed,
c_noise,
m_noise,
ignore_regions=args.ignore_regions)
pf = ParticleFilter(config, building_map, feed_processor)
w = DisplayWindow(building_map, args.map_image, pf, display=display_on)
w.start_particle_filter()
from sim import Simulation
import network
import neuron
import netshow as ns
import progress
# Create our network
networks = {}
networks["C"] = network.DTN_Coincidence()
networks["AC"] = network.DTN_AntiCoincidence()
# Initialize the simulation with the network
s = {}
for net in networks:
s[net] = Simulation(networks[net])
s[net].verbose = False
s[net].sim_time = 100
s[net].dt = 0.025
# Run the simulations
stims = [i for i in range(1,31,1)]
repeats = 15
param = [(i*2)+10 for i in range(11)]
total = len(stims)*len(param)*repeats
print "Running %d simulations..." % total
count = 0
for a in param:
number_of_calculated_models = 0
total_number_of_models = total_number_of_models * 60
for n in number_of_samples:
number_of_samples_text = str(n)
for seed in seeds:
seed_text = str(seed)
for length in sample_length_values:
sample_length_text = str(length)
rand = random.Random()
for wd, mc in window_device.items():
w = wd[0]
d = wd[1]
sample_gen = SampleGen(mc, rand)
simulator = Simulation(sample_gen)
# time used to generate all samples
rand.seed(seed)
sample_generation_time = 0
for i in range(n):
start = time.time()
sample_gen.sample(length)
end = time.time()
sample_generation_time += end - start
rand.seed(seed)
# evaluate the MC given the current combination of parameters
start = time.time()
# performance indicator 1 - prediction accuracy
accuracy = simulator.runSimulation(mc, n, length, is_dpm)
end = time.time()
# performance indicator 2 - execution time
from sim import Simulation sirkus = Simulation() sirkus.testY()
def sim(nt, dri):
ip = InputParameters(nt, dri)
s = Simulation(ip)
s.run()
#! /usr/bin/env python3
from sim import Simulation
if __name__ == "__main__":
sim = Simulation(ants=5, x=100, y=100, num_rivers=20)
sim.run()
class DoE(DoEProblem):
'''The class :class:`casiopeia.doe.DoE` is used to set up
Design-of-Experiments-problems for systems defined with the
:class:`casiopeia.system.System` class.
The aim of the experimental design optimization is to identify a set of
controls that can be used for the generation of measurement data which
allows for a better estimation of the unknown parameters of a system.
To achieve this, an information function on the covariance matrix of the
estimated parameters is minimized. The values of the estimated parameters,
though they are mostly an initial
guess for their values, are not changed during the optimization.
Optimum experimental design and parameter estimation methods can be used
interchangeably until a desired accuracy of the parameters has been
achieved.
'''
@property
def optimized_controls(self):
return self.design_results["x"][ \
:(self._discretization.number_of_intervals * \
self._discretization.system.nu)]
def _discretize_system(self, system, time_points, discretization_method, \
**kwargs):
if system.nx == 0:
self._discretization = NoDiscretization(system, time_points)
elif system.nx != 0:
if discretization_method == "collocation":
self._discretization = ODECollocation( \
system, time_points, **kwargs)
elif discretization_method == "multiple_shooting":
self._discretization = ODEMultipleShooting( \
system, time_points, **kwargs)
else:
raise NotImplementedError('''
Unknown discretization method: {0}.
Possible values are "collocation" and "multiple_shooting".
'''.format(str(discretization_method)))
def _set_parameter_guess(self, pdata):
self._pdata = inputchecks.check_parameter_data(pdata, \
self._discretization.system.np)
def _apply_parameters_to_equality_constraints(self):
optimization_variables_for_equality_constraints = ci.veccat([ \
self._discretization.optimization_variables["U"],
self._discretization.optimization_variables["Q"],
self._discretization.optimization_variables["X"],
self._discretization.optimization_variables["EPS_U"],
self._discretization.optimization_variables["P"],
])
optimization_variables_parameters_applied = ci.veccat([ \
self._discretization.optimization_variables["U"],
self._discretization.optimization_variables["Q"],
self._discretization.optimization_variables["X"],
ci.mx(*self._discretization.optimization_variables["EPS_U"].shape),
self._pdata,
])
equality_constraints_fcn = ci.mx_function( \
"equality_constraints_fcn", \
[optimization_variables_for_equality_constraints], \
[self._discretization.equality_constraints])
self._equality_constraints_parameters_applied = \
equality_constraints_fcn(optimization_variables_parameters_applied)
def _apply_parameters_to_discretization(self):
self._apply_parameters_to_equality_constraints()
def _set_optimization_variables(self):
self._optimization_variables = ci.veccat([ \
self._discretization.optimization_variables["U"],
self._discretization.optimization_variables["Q"],
self._discretization.optimization_variables["X"],
])
def _set_optimization_variables_initials(self, qinit, x0, uinit):
self.simulation = Simulation(self._discretization.system, \
self._pdata, qinit)
self.simulation.run_system_simulation(x0, \
self._discretization.time_points, uinit, print_status = False)
xinit = self.simulation.simulation_results
repretitions_xinit = \
self._discretization.optimization_variables["X"][:,:-1].shape[1] / \
self._discretization.number_of_intervals
Xinit = ci.repmat(xinit[:, :-1], repretitions_xinit, 1)
Xinit = ci.horzcat([ \
Xinit.reshape((self._discretization.system.nx, \
Xinit.numel() / self._discretization.system.nx)),
xinit[:, -1],
])
uinit = inputchecks.check_controls_data(uinit, \
self._discretization.system.nu, \
self._discretization.number_of_intervals)
Uinit = uinit
qinit = inputchecks.check_constant_controls_data(qinit, \
self._discretization.system.nq)
Qinit = qinit
self._optimization_variables_initials = ci.veccat([ \
Uinit,
Qinit,
Xinit,
])
def _set_optimization_variables_lower_bounds(self, umin, qmin, xmin, x0):
umin_user_provided = umin
umin = inputchecks.check_controls_data(umin, \
self._discretization.system.nu, 1)
if umin_user_provided is None:
umin = -np.inf * np.ones(umin.shape)
Umin = ci.repmat(umin, 1, \
self._discretization.optimization_variables["U"].shape[1])
qmin_user_provided = qmin
qmin = inputchecks.check_constant_controls_data(qmin, \
self._discretization.system.nq)
if qmin_user_provided is None:
qmin = -np.inf * np.ones(qmin.shape)
Qmin = qmin
xmin_user_provided = xmin
xmin = inputchecks.check_states_data(xmin, \
self._discretization.system.nx, 0)
if xmin_user_provided is None:
xmin = -np.inf * np.ones(xmin.shape)
Xmin = ci.repmat(xmin, 1, \
self._discretization.optimization_variables["X"].shape[1])
Xmin[:,0] = x0
self._optimization_variables_lower_bounds = ci.veccat([ \
Umin,
Qmin,
Xmin,
])
def _set_optimization_variables_upper_bounds(self, umax, qmax, xmax, x0):
umax_user_provided = umax
umax = inputchecks.check_controls_data(umax, \
self._discretization.system.nu, 1)
if umax_user_provided is None:
umax = np.inf * np.ones(umax.shape)
Umax = ci.repmat(umax, 1, \
self._discretization.optimization_variables["U"].shape[1])
qmax_user_provided = qmax
qmax = inputchecks.check_constant_controls_data(qmax, \
self._discretization.system.nq)
if qmax_user_provided is None:
qmax = -np.inf * np.ones(qmax.shape)
Qmax = qmax
xmax_user_provided = xmax
xmax = inputchecks.check_states_data(xmax, \
self._discretization.system.nx, 0)
if xmax_user_provided is None:
xmax = np.inf * np.ones(xmax.shape)
Xmax = ci.repmat(xmax, 1, \
self._discretization.optimization_variables["X"].shape[1])
Xmax[:,0] = x0
self._optimization_variables_upper_bounds = ci.veccat([ \
Umax,
Xmax,
])
def _set_measurement_data(self):
# The DOE problem does not depend on actual measurement values,
# the measurement deviations are only needed to set up the objective;
# therefore, dummy-values for the measurements can be used
# (see issue #7 for further information)
measurement_data = np.zeros((self._discretization.system.nphi, \
self._discretization.number_of_intervals + 1))
self._measurement_data_vectorized = ci.vec(measurement_data)
def _set_weightings(self, wv, weps_u):
input_error_weightings = \
inputchecks.check_input_error_weightings(weps_u, \
self._discretization.system.neps_u,
self._discretization.number_of_intervals)
measurement_weightings = \
inputchecks.check_measurement_weightings(wv, \
self._discretization.system.nphi, \
self._discretization.number_of_intervals + 1)
self._weightings_vectorized = ci.veccat([ \
input_error_weightings,
measurement_weightings,
])
def _set_measurement_deviations(self):
self._measurement_deviations = ci.vertcat([ \
ci.vec(self._discretization.measurements) - \
self._measurement_data_vectorized + \
ci.vec(self._discretization.optimization_variables["V"])
])
def _setup_constraints(self):
self._constraints = ci.vertcat([ \
self._measurement_deviations,
self._discretization.equality_constraints,
])
def _set_cov_matrix_derivative_directions(self):
# These correspond to the optimization variables of the parameter
# estimation problem; the evaluation of the covariance matrix, though,
# does not depend on the actual values of V, EPS_E and EPS_U, and with
# this, the DoE problem does not
self._cov_matrix_derivative_directions = ci.veccat([ \
self._discretization.optimization_variables["P"],
self._discretization.optimization_variables["X"],
self._discretization.optimization_variables["V"],
self._discretization.optimization_variables["EPS_U"],
])
def _setup_gauss_newton_lagrangian_hessian(self):
gauss_newton_lagrangian_hessian_diag = ci.vertcat([ \
ci.mx(self._cov_matrix_derivative_directions.shape[0] - \
self._weightings_vectorized.shape[0], 1), \
self._weightings_vectorized])
self._gauss_newton_lagrangian_hessian = ci.diag( \
gauss_newton_lagrangian_hessian_diag)
def _setup_covariance_matrix_for_evaluation(self):
covariance_matrix_free_variables = ci.veccat([ \
self._discretization.optimization_variables["P"],
self._discretization.optimization_variables["U"],
self._discretization.optimization_variables["Q"],
self._discretization.optimization_variables["X"],
self._discretization.optimization_variables["EPS_U"],
])
self._covariance_matrix_fcn = ci.mx_function("covariance_matrix_fcn", \
[covariance_matrix_free_variables], \
[self._covariance_matrix.covariance_matrix])
def _apply_parameters_to_objective(self):
# As mentioned above, the objective does not depend on the actual
# values of V, but on the values of P and EPS_U, while
# P is fed from pdata, and EPS_U is supposed to be 0
objective_free_variables = ci.veccat([ \
self._discretization.optimization_variables["P"],
self._discretization.optimization_variables["U"],
self._discretization.optimization_variables["Q"],
self._discretization.optimization_variables["X"],
self._discretization.optimization_variables["EPS_U"],
])
objective_free_variables_parameters_applied = ci.veccat([ \
self._pdata,
self._discretization.optimization_variables["U"],
self._discretization.optimization_variables["Q"],
self._discretization.optimization_variables["X"],
ci.mx(*self._discretization.optimization_variables["EPS_U"].shape),
])
objective_fcn = ci.mx_function("objective_fcn", \
[objective_free_variables], [self._objective_parameters_free])
self._objective = objective_fcn( \
objective_free_variables_parameters_applied)
def __init__(self, system, time_points, \
uinit = None, umin = None, umax = None, \
qinit = None, qmin = None, qmax = None, \
pdata = None, x0 = None, \
xmin = None, xmax = None, \
wv = None, weps_u = None, \
discretization_method = "collocation", \
optimality_criterion = "A", **kwargs):
r'''
:raises: AttributeError, NotImplementedError
:param system: system considered for parameter estimation, specified
using the :class:`casiopeia.system.System` class
:type system: casiopeia.system.System
:param time_points: time points :math:`t_\text{N} \in \mathbb{R}^\text{N}`
used to discretize the continuous time problem. Controls
will be applied at the first :math:`N-1` time points,
while measurements take place at all :math:`N` time points.
:type time_points: numpy.ndarray, casadi.DMatrix, list
:param umin: optional, lower bounds of the time-varying controls
:math:`u_\text{min} \in \mathbb{R}^{\text{n}_\text{u}}`;
if not values are given, :math:`-\infty` will be used
:type umin: numpy.ndarray, casadi.DMatrix
:param umax: optional, upper bounds of the time-vaying controls
:math:`u_\text{max} \in \mathbb{R}^{\text{n}_\text{u}}`;
if not values are given, :math:`\infty` will be used
:type umax: numpy.ndarray, casadi.DMatrix
:param uinit: optional, initial guess for the values of the time-varying controls
:math:`u_\text{N} \in \mathbb{R}^{\text{n}_\text{u} \times \text{N}-1}`
that (might) change at the switching time points;
if no values are given, 0 will be used; note that a poorly
or wrongly chosen initial guess can cause the optimization
to fail, and note that the
the second dimension of :math:`u_N` is :math:`N-1` and not
:math:`N`, since there is no control value applied at the
last time point
:type uinit: numpy.ndarray, casadi.DMatrix
:param qmin: optional, lower bounds of the time-constant controls
:math:`q_\text{min} \in \mathbb{R}^{\text{n}_\text{q}}`;
if not values are given, :math:`-\infty` will be used
:type qmin: numpy.ndarray, casadi.DMatrix
:param qmax: optional, upper bounds of the time-constant controls
:math:`q_\text{max} \in \mathbb{R}^{\text{n}_\text{q}}`;
if not values are given, :math:`\infty` will be used
:type qmax: numpy.ndarray, casadi.DMatrix
:param qinit: optional, initial guess for the optimal values of the
time-constant controls
:math:`q_\text{init} \in \mathbb{R}^{\text{n}_\text{q}}`;
if not values are given, 0 will be used; note that a poorly
or wrongly chosen initial guess can cause the optimization
to fail
:type qinit: numpy.ndarray, casadi.DMatrix
:param pdata: values of the time-constant parameters
:math:`p \in \mathbb{R}^{\text{n}_\text{p}}`
:type pdata: numpy.ndarray, casadi.DMatrix
:param x0: state values :math:`x_0 \in \mathbb{R}^{\text{n}_\text{x}}`
at the first time point :math:`t_0`
:type x0: numpy.ndarray, casadi.DMatrix, list
:param xmin: optional, lower bounds of the states
:math:`x_\text{min} \in \mathbb{R}^{\text{n}_\text{x}}`;
if no value is given, :math:`-\infty` will be used
:type xmin: numpy.ndarray, casadi.DMatrix
:param xmax: optional, lower bounds of the states
:math:`x_\text{max} \in \mathbb{R}^{\text{n}_\text{x}}`;
if no value is given, :math:`\infty` will be used
:type xmax: numpy.ndarray, casadi.DMatrix
:param wv: weightings for the measurements
:math:`w_\text{v} \in \mathbb{R}^{\text{n}_\text{y} \times \text{N}}`
:type wv: numpy.ndarray, casadi.DMatrix
:param weps_u: weightings for the input errors
:math:`w_{\epsilon_\text{u}} \in \mathbb{R}^{\text{n}_{\epsilon_\text{u}}}`
(only necessary
if input errors are used within ``system``)
:type weps_u: numpy.ndarray, casadi.DMatrix
:param discretization_method: optional, the method that shall be used for
discretization of the continuous time
problem w. r. t. the time points given
in :math:`t_\text{N}`; possible values are
"collocation" (default) and
"multiple_shooting"
:type discretization_method: str
:param optimality_criterion: optional, the information function
:math:`I_\text{X}(\cdot)` to be used on the
covariance matrix, possible values are
`A` (default) and `D`, while
.. math ::
\begin{aligned}
I_\text{A}(\Sigma_\text{p}) & = \frac{1}{n_\text{p}} \text{Tr}(\Sigma_\text{p}),\\
I_\text{D}(\Sigma_\text{p}) & = \begin{vmatrix} \Sigma_\text{p} \end{vmatrix} ^{\frac{1}{n_\text{p}}},
\end{aligned}
for further information see e. g. [#f1]_
:type optimality_criterion: str
Depending on the discretization method specified in
`discretization_method`, the following parameters can be used
for further specification:
:param collocation_scheme: optional, scheme used for setting up the
collocation polynomials,
possible values are `radau` (default)
and `legendre`
:type collocation_scheme: str
:param number_of_collocation_points: optional, order of collocation
polynomials
:math:`d \in \mathbb{Z}` (default
values is 3)
:type number_of_collocation_points: int
:param integrator: optional, integrator to be used with multiple shooting.
See the CasADi documentation for a list of
all available integrators. As a default, `cvodes`
is used.
:type integrator: str
:param integrator_options: optional, options to be passed to the CasADi
integrator used with multiple shooting
(see the CasADi documentation for a list of
all possible options)
:type integrator_options: dict
You do not need to specify initial guesses for the estimated states,
since these are obtained with a system simulation using the initial
states and the provided initial guesses for the controls.
The resulting optimization problem has the following form:
.. math::
\begin{aligned}
\text{arg}\,\underset{u, q, x}{\text{min}} & & I(\Sigma_{\text{p}}(x, u, q; p)) &\\
\text{subject to:} & & g(x, u, q; p) & = 0\\
& & u_\text{min} \leq u_\text{k} & \leq u_\text{max} \hspace{1cm} k = 1, \dots, N-1\\
& & x_\text{min} \leq x_\text{k} & \leq x_\text{max} \hspace{1cm} k = 1, \dots, N\\
& & x_1 \leq x(t_1) & \leq x_1
\end{aligned}
where :math:`\Sigma_p = \text{Cov}(p)` and :math:`g(\cdot)` contains the
discretized system dynamics
according to the specified discretization method. If the system is
non-dynamic, it only contains the user-provided equality constraints.
.. rubric:: References
.. [#f1] |linkf1|_
.. _linkf1: http://ginger.iwr.uni-heidelberg.de/vplan/images/5/54/Koerkel2002.pdf
.. |linkf1| replace:: *Körkel, Stefan: Numerische Methoden für Optimale Versuchsplanungsprobleme bei nichtlinearen DAE-Modellen, PhD Thesis, Heidelberg university, 2002, pages 74/75.*
'''
intro()
self._discretize_system( \
system, time_points, discretization_method, **kwargs)
self._set_parameter_guess(pdata)
self._apply_parameters_to_discretization()
self._set_optimization_variables()
self._set_optimization_variables_initials(qinit, x0, uinit)
self._set_optimization_variables_lower_bounds(umin, qmin, xmin, x0)
self._set_optimization_variables_upper_bounds(umax, qmax, xmax, x0)
self._set_measurement_data()
self._set_weightings(wv, weps_u)
self._set_measurement_deviations()
self._set_cov_matrix_derivative_directions()
self._setup_constraints()
self._setup_gauss_newton_lagrangian_hessian()
self._setup_covariance_matrix()
self._setup_covariance_matrix_for_evaluation()
self._set_optimiality_criterion(optimality_criterion)
self._setup_objective()
self._apply_parameters_to_objective()
self._setup_nlp()
def _print_experimental_properties(self, covariance_matrix):
np.set_printoptions(linewidth = 200, \
formatter={'float': lambda x: format(x, ' 10.8e')})
print("\nParameters p_i:")
for k, pk in enumerate(self._pdata):
print(" p_{0:<3} = {1} +/- {2}".format( \
k, pk, ci.sqrt(abs(ci.diag(covariance_matrix)[k]))))
print("\nCovariance matrix for this setup:")
print(np.atleast_2d(covariance_matrix))
def _compute_initial_covariance_matrix(self):
covariance_matrix_initial_input = ci.veccat([ \
self._pdata,
self._optimization_variables_initials,
np.zeros(self._discretization.optimization_variables["EPS_U"].shape)
])
self._covariance_matrix_initial = self._covariance_matrix_fcn( \
covariance_matrix_initial_input)
def _compute_optimized_covariance_matrix(self):
covariance_matrix_optimized_input = ci.veccat([ \
self._pdata,
self.design_results["x"],
np.zeros(self._discretization.optimization_variables["EPS_U"].shape)
])
self._covariance_matrix_optimized = self._covariance_matrix_fcn( \
covariance_matrix_optimized_input)
def plot_confidence_ellipsoids(self, properties = "initial"):
r'''
:param properties: Set whether the experimental properties for the
initial setup ("initial", default), the optimized setup
("optimized") or for both setups ("all") shall be
plotted. In the later case, both ellipsoids for one
pair of parameters will be displayed within one plot.
:type properties: str
Plot confidence ellipsoids for all parameter pairs.
Since the number of plots is possibly big, all plots will be saved
within a folder *confidence_ellipsoids_scriptname* in you current
working directory rather than being displayed directly.
'''
self._plot_confidence_ellipsoids(pdata = self._pdata, \
properties = properties)
def my_calculation(arguments):
"""
This function returns the assets after death for the given arguments.
"""
args, rate_of_return, years_to_wait = arguments
#
# Calculate the most efficient Roth conversion amount.
#
most_assets = 0
roth_conversion_amount = 0
best_roth_conversion_amount = 0
while True:
simulation = Simulation(
args.starting_balance_hsa,
args.starting_balance_taxable,
args.starting_balance_trad_401k,
args.starting_balance_trad_ira,
args.starting_balance_roth_401k,
args.starting_balance_roth_ira,
rate_of_return,
years_to_wait,
args.current_age,
args.age_of_retirement,
args.age_to_start_rmds,
args.age_of_death,
roth_conversion_amount, # this is our variable
args.income,
args.yearly_income_raise,
args.max_income,
args.age_of_marriage,
args.spending,
args.contribution_limit_hsa,
args.contribution_catch_up_amount_hsa,
args.contribution_catch_up_age_hsa,
args.contribution_limit_401k,
args.contribution_limit_401k_total,
args.contribution_catch_up_amount_401k,
args.contribution_catch_up_age_401k,
args.contribution_limit_ira,
args.contribution_catch_up_amount_ira,
args.contribution_catch_up_age_ira,
args.do_mega_backdoor_roth,
args.work_state,
args.retirement_state,
args.add_dependent,
args.public_safety_employee,
args.employer_match_401k,
args.max_contribution_percentage_401k,
args.employer_contribution_hsa)
simulation.simulate()
if round(simulation.get_total_assets_after_death(), 2) >= round(
most_assets, 2):
best_roth_conversion_amount = roth_conversion_amount
most_assets = simulation.get_total_assets_after_death()
traditional_money = (simulation.accounts.trad_401k.get_value() +
simulation.accounts.trad_ira.get_value())
if round(traditional_money, 2) == 0:
break
roth_conversion_amount += args.roth_conversion_unit
simulation = Simulation(
args.starting_balance_hsa, args.starting_balance_taxable,
args.starting_balance_trad_401k, args.starting_balance_trad_ira,
args.starting_balance_roth_401k, args.starting_balance_roth_ira,
rate_of_return, years_to_wait, args.current_age,
args.age_of_retirement, args.age_to_start_rmds, args.age_of_death,
best_roth_conversion_amount, args.income, args.yearly_income_raise,
args.max_income, args.age_of_marriage, args.spending,
args.contribution_limit_hsa, args.contribution_catch_up_amount_hsa,
args.contribution_catch_up_age_hsa, args.contribution_limit_401k,
args.contribution_limit_401k_total,
args.contribution_catch_up_amount_401k,
args.contribution_catch_up_age_401k, args.contribution_limit_ira,
args.contribution_catch_up_amount_ira,
args.contribution_catch_up_age_ira, args.do_mega_backdoor_roth,
args.work_state, args.retirement_state, args.add_dependent,
args.public_safety_employee, args.employer_match_401k,
args.max_contribution_percentage_401k, args.employer_contribution_hsa)
simulation.simulate()
return simulation.get_total_assets_after_death()
def startSim(self):
self.sim[0] = Simulation(self.population, self.screen, self.clock)
#!/usr/bin/env python
from parse import Parser
from sim import Simulation
#----- Main
parser = Parser("sampledata.csv")
simulation = Simulation()
simulation.setUp(parser.getData())
print("Running ...")
simulation.runFullSim()
print(simulation.getAllPaths())
print(simulation.getAllHeatData())
'''
print('This is a test!')
parser = Parser("sampledata.csv")
simulation = Simulation()
simulation.setUp(parser.getData())
done = False
while not done:
print("(25) to see all paths thus far. (35) to see all heat data. (99) to quit.")
print("-xx to completely run sim, in increments of xx.")
print("Total time so far: " + str(simulation.getTotalTime()))
inputStr = str(input("How long to move? "))
if(inputStr[0] == '-'):
simulation.oneStep(int(inputStr[1:]))
if (int(inputStr) == 99):
done = True
elif (int(inputStr) == 25):
def run_sim(model):
odefunction = lambda t, x: model.ode(t, x)
sim_params = get_sim_parameters()
sim = Simulation(sim_params)
return sim.simulate(odefunction)