def resample(self):
if self.parameters.state_parameters["resample_threshold"] < 0:
return
# Effective number of particles
eff_nparticles = 1/np.power(self.weights, 2).sum()
resample_threshold = (
eff_nparticles/self.parameters.state_parameters["nparticles"])
# Check if we have particle depletion
if (resample_threshold >
self.parameters.state_parameters["resample_threshold"]):
return
# Otherwise we need to resample
max_wt_index = self.weights.argmax()
max_wt_state = self.states[max_wt_index].copy()
max_wt_state[4:] = 0
max_wt_map = self.maps[max_wt_index].copy()
resample_index = misctools.get_resample_index(self.weights,
self.parameters.state_parameters["nparticles"]-1)
# self.states is a numpy array so the indexing operation forces a copy
resampled_states = self.states[resample_index]
resampled_states.resize((resampled_states.shape[0]+1, resampled_states.shape[1]))
resampled_states[-1] = max_wt_state
resampled_maps = [self.maps[i].copy() for i in resample_index]
resampled_maps.append(max_wt_map)
resampled_weights = (
np.ones(self.parameters.state_parameters["nparticles"], dtype=float)*
1/float(self.parameters.state_parameters["nparticles"]))
self.weights = resampled_weights
self.states = resampled_states
self.maps = resampled_maps
def phdResample(self, FORCE_COPY=True):
sum_weights = [_weights_.sum() for _weights_ in self._weights_]
nparticles = np.array(sum_weights)*self.parameters.phd_parameters["nparticles"]
for count in range(len(self._weights_)):
resample_indices = get_resample_index(self._weights_[count], nparticles[count])
self._weights_[count] = (1.0/nparticles[count])*np.ones(nparticles[count])
if FORCE_COPY:
self._states_[count] = copy.deepcopy([self._states_[count][ridx] for ridx in resample_indices])
else:
self._states_[count] = [self._states_[count][ridx] for ridx in resample_indices]
def resample(self):
# Effective number of particles
nparticles = self.vars.nparticles
eff_nparticles = 1/np.power(self.vehicle.weights, 2).sum()
resample_threshold = eff_nparticles/nparticles
# Check if we have particle depletion
if (resample_threshold > self.vars.resample_threshold):
return
# Otherwise we need to resample
resample_index = misctools.get_resample_index(self.vehicle.weights)
# self.states is a numpy array so the indexing operation forces a copy
self.vehicle.weights = np.ones(nparticles, dtype=float)/nparticles
self.vehicle.states = self.vehicle.states[resample_index]
self.vehicle.covs = self.vehicle.covs[resample_index]
self.vehicle.maps = [self.vehicle.maps[i].copy()
for i in resample_index]
