src/daComposant/daAlgorithms/ExtendedKalmanFilter.py

   1 # -*- coding: utf-8 -*-
   2 #
   3 # Copyright (C) 2008-2020 EDF R&D
   4 #
   5 # This library is free software; you can redistribute it and/or
   6 # modify it under the terms of the GNU Lesser General Public
   7 # License as published by the Free Software Foundation; either
   8 # version 2.1 of the License.
   9 #
  10 # This library is distributed in the hope that it will be useful,
  11 # but WITHOUT ANY WARRANTY; without even the implied warranty of
  12 # MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
  13 # Lesser General Public License for more details.
  14 #
  15 # You should have received a copy of the GNU Lesser General Public
  16 # License along with this library; if not, write to the Free Software
  17 # Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307 USA
  18 #
  19 # See http://www.salome-platform.org/ or email : webmaster.salome@opencascade.com
  20 #
  21 # Author: Jean-Philippe Argaud, jean-philippe.argaud@edf.fr, EDF R&D
  22
  23 import logging
  24 from daCore import BasicObjects
  25 import numpy
  26
  27 # ==============================================================================
  28 class ElementaryAlgorithm(BasicObjects.Algorithm):
  29     def __init__(self):
  30         BasicObjects.Algorithm.__init__(self, "EXTENDEDKALMANFILTER")
  31         self.defineRequiredParameter(
  32             name     = "ConstrainedBy",
  33             default  = "EstimateProjection",
  34             typecast = str,
  35             message  = "Prise en compte des contraintes",
  36             listval  = ["EstimateProjection"],
  37             )
  38         self.defineRequiredParameter(
  39             name     = "EstimationOf",
  40             default  = "State",
  41             typecast = str,
  42             message  = "Estimation d'etat ou de parametres",
  43             listval  = ["State", "Parameters"],
  44             )
  45         self.defineRequiredParameter(
  46             name     = "StoreInternalVariables",
  47             default  = False,
  48             typecast = bool,
  49             message  = "Stockage des variables internes ou intermédiaires du calcul",
  50             )
  51         self.defineRequiredParameter(
  52             name     = "StoreSupplementaryCalculations",
  53             default  = [],
  54             typecast = tuple,
  55             message  = "Liste de calculs supplémentaires à stocker et/ou effectuer",
  56             listval  = [
  57                 "Analysis",
  58                 "APosterioriCorrelations",
  59                 "APosterioriCovariance",
  60                 "APosterioriStandardDeviations",
  61                 "APosterioriVariances",
  62                 "BMA",
  63                 "CostFunctionJ",
  64                 "CostFunctionJAtCurrentOptimum",
  65                 "CostFunctionJb",
  66                 "CostFunctionJbAtCurrentOptimum",
  67                 "CostFunctionJo",
  68                 "CostFunctionJoAtCurrentOptimum",
  69                 "CurrentIterationNumber",
  70                 "CurrentOptimum",
  71                 "CurrentState",
  72                 "ForecastState",
  73                 "IndexOfOptimum",
  74                 "InnovationAtCurrentAnalysis",
  75                 "InnovationAtCurrentState",
  76                 "SimulatedObservationAtCurrentAnalysis",
  77                 "SimulatedObservationAtCurrentOptimum",
  78                 "SimulatedObservationAtCurrentState",
  79                 ]
  80             )
  81         self.defineRequiredParameter( # Pas de type
  82             name     = "Bounds",
  83             message  = "Liste des valeurs de bornes",
  84             )
  85         self.requireInputArguments(
  86             mandatory= ("Xb", "Y", "HO", "R", "B"),
  87             optional = ("U", "EM", "CM", "Q"),
  88             )
  89         self.setAttributes(tags=(
  90             "DataAssimilation",
  91             "NonLinear",
  92             "Filter",
  93             "Dynamic",
  94             ))
  95
  96     def run(self, Xb=None, Y=None, U=None, HO=None, EM=None, CM=None, R=None, B=None, Q=None, Parameters=None):
  97         self._pre_run(Parameters, Xb, Y, U, HO, EM, CM, R, B, Q)
  98         #
  99         if self._parameters["EstimationOf"] == "Parameters":
 100             self._parameters["StoreInternalVariables"] = True
 101         #
 102         # Opérateurs
 103         # ----------
 104         Hm = HO["Direct"].appliedControledFormTo
 105         #
 106         if self._parameters["EstimationOf"] == "State":
 107             Mm = EM["Direct"].appliedControledFormTo
 108         #
 109         if CM is not None and "Tangent" in CM and U is not None:
 110             Cm = CM["Tangent"].asMatrix(Xb)
 111         else:
 112             Cm = None
 113         #
 114         # Nombre de pas identique au nombre de pas d'observations
 115         # -------------------------------------------------------
 116         if hasattr(Y,"stepnumber"):
 117             duration = Y.stepnumber()
 118         else:
 119             duration = 2
 120         #
 121         # Précalcul des inversions de B et R
 122         # ----------------------------------
 123         if self._parameters["StoreInternalVariables"] \
 124             or self._toStore("CostFunctionJ") \
 125             or self._toStore("CostFunctionJb") \
 126             or self._toStore("CostFunctionJo") \
 127             or self._toStore("CurrentOptimum") \
 128             or self._toStore("APosterioriCovariance"):
 129             BI = B.getI()
 130             RI = R.getI()
 131         #
 132         # Initialisation
 133         # --------------
 134         Xn = Xb
 135         Pn = B
 136         #
 137         if len(self.StoredVariables["Analysis"])==0 or not self._parameters["nextStep"]:
 138             self.StoredVariables["CurrentIterationNumber"].store( len(self.StoredVariables["Analysis"]) )
 139             self.StoredVariables["Analysis"].store( numpy.ravel(Xn) )
 140             if self._toStore("APosterioriCovariance"):
 141                 self.StoredVariables["APosterioriCovariance"].store( Pn.asfullmatrix(Xn.size) )
 142                 covarianceXa = Pn
 143         #
 144         Xa               = Xn
 145         XaMin            = Xn
 146         previousJMinimum = numpy.finfo(float).max
 147         #
 148         for step in range(duration-1):
 149             if hasattr(Y,"store"):
 150                 Ynpu = numpy.asmatrix(numpy.ravel( Y[step+1] )).T
 151             else:
 152                 Ynpu = numpy.asmatrix(numpy.ravel( Y )).T
 153             #
 154             Ht = HO["Tangent"].asMatrix(ValueForMethodForm = Xn)
 155             Ht = Ht.reshape(Ynpu.size,Xn.size) # ADAO & check shape
 156             Ha = HO["Adjoint"].asMatrix(ValueForMethodForm = Xn)
 157             Ha = Ha.reshape(Xn.size,Ynpu.size) # ADAO & check shape
 158             #
 159             if self._parameters["EstimationOf"] == "State":
 160                 Mt = EM["Tangent"].asMatrix(ValueForMethodForm = Xn)
 161                 Mt = Mt.reshape(Xn.size,Xn.size) # ADAO & check shape
 162                 Ma = EM["Adjoint"].asMatrix(ValueForMethodForm = Xn)
 163                 Ma = Ma.reshape(Xn.size,Xn.size) # ADAO & check shape
 164             #
 165             if U is not None:
 166                 if hasattr(U,"store") and len(U)>1:
 167                     Un = numpy.asmatrix(numpy.ravel( U[step] )).T
 168                 elif hasattr(U,"store") and len(U)==1:
 169                     Un = numpy.asmatrix(numpy.ravel( U[0] )).T
 170                 else:
 171                     Un = numpy.asmatrix(numpy.ravel( U )).T
 172             else:
 173                 Un = None
 174             #
 175             if self._parameters["EstimationOf"] == "State":
 176                 Xn_predicted = numpy.asmatrix(numpy.ravel( Mm( (Xn, Un) ) )).T
 177                 if Cm is not None and Un is not None: # Attention : si Cm est aussi dans M, doublon !
 178                     Cm = Cm.reshape(Xn.size,Un.size) # ADAO & check shape
 179                     Xn_predicted = Xn_predicted + Cm * Un
 180                 Pn_predicted = Q + Mt * Pn * Ma
 181             elif self._parameters["EstimationOf"] == "Parameters":
 182                 # --- > Par principe, M = Id, Q = 0
 183                 Xn_predicted = Xn
 184                 Pn_predicted = Pn
 185             #
 186             if self._parameters["Bounds"] is not None and self._parameters["ConstrainedBy"] == "EstimateProjection":
 187                 Xn_predicted = numpy.max(numpy.hstack((Xn_predicted,numpy.asmatrix(self._parameters["Bounds"])[:,0])),axis=1)
 188                 Xn_predicted = numpy.min(numpy.hstack((Xn_predicted,numpy.asmatrix(self._parameters["Bounds"])[:,1])),axis=1)
 189             #
 190             if self._parameters["EstimationOf"] == "State":
 191                 _HX          = numpy.asmatrix(numpy.ravel( Hm( (Xn_predicted, None) ) )).T
 192                 _Innovation  = Ynpu - _HX
 193             elif self._parameters["EstimationOf"] == "Parameters":
 194                 _HX          = numpy.asmatrix(numpy.ravel( Hm( (Xn_predicted, Un) ) )).T
 195                 _Innovation  = Ynpu - _HX
 196                 if Cm is not None and Un is not None: # Attention : si Cm est aussi dans H, doublon !
 197                     _Innovation = _Innovation - Cm * Un
 198             #
 199             _A = R + numpy.dot(Ht, Pn_predicted * Ha)
 200             _u = numpy.linalg.solve( _A , _Innovation )
 201             Xn = Xn_predicted + Pn_predicted * Ha * _u
 202             Kn = Pn_predicted * Ha * (R + numpy.dot(Ht, Pn_predicted * Ha)).I
 203             Pn = Pn_predicted - Kn * Ht * Pn_predicted
 204             Xa, _HXa = Xn, _HX # Pointeurs
 205             #
 206             self.StoredVariables["CurrentIterationNumber"].store( len(self.StoredVariables["Analysis"]) )
 207             # ---> avec analysis
 208             self.StoredVariables["Analysis"].store( Xa )
 209             if self._toStore("SimulatedObservationAtCurrentAnalysis"):
 210                 self.StoredVariables["SimulatedObservationAtCurrentAnalysis"].store( _HXa )
 211             if self._toStore("InnovationAtCurrentAnalysis"):
 212                 self.StoredVariables["InnovationAtCurrentAnalysis"].store( _Innovation )
 213             # ---> avec current state
 214             if self._parameters["StoreInternalVariables"] \
 215                 or self._toStore("CurrentState"):
 216                 self.StoredVariables["CurrentState"].store( Xn )
 217             if self._toStore("ForecastState"):
 218                 self.StoredVariables["ForecastState"].store( Xn_predicted )
 219             if self._toStore("BMA"):
 220                 self.StoredVariables["BMA"].store( Xn_predicted - Xa )
 221             if self._toStore("InnovationAtCurrentState"):
 222                 self.StoredVariables["InnovationAtCurrentState"].store( _Innovation )
 223             if self._toStore("SimulatedObservationAtCurrentState") \
 224                 or self._toStore("SimulatedObservationAtCurrentOptimum"):
 225                 self.StoredVariables["SimulatedObservationAtCurrentState"].store( _HX )
 226             # ---> autres
 227             if self._parameters["StoreInternalVariables"] \
 228                 or self._toStore("CostFunctionJ") \
 229                 or self._toStore("CostFunctionJb") \
 230                 or self._toStore("CostFunctionJo") \
 231                 or self._toStore("CurrentOptimum") \
 232                 or self._toStore("APosterioriCovariance"):
 233                 Jb  = float( 0.5 * (Xa - Xb).T * BI * (Xa - Xb) )
 234                 Jo  = float( 0.5 * _Innovation.T * RI * _Innovation )
 235                 J   = Jb + Jo
 236                 self.StoredVariables["CostFunctionJb"].store( Jb )
 237                 self.StoredVariables["CostFunctionJo"].store( Jo )
 238                 self.StoredVariables["CostFunctionJ" ].store( J )
 239                 #
 240                 if self._toStore("IndexOfOptimum") \
 241                     or self._toStore("CurrentOptimum") \
 242                     or self._toStore("CostFunctionJAtCurrentOptimum") \
 243                     or self._toStore("CostFunctionJbAtCurrentOptimum") \
 244                     or self._toStore("CostFunctionJoAtCurrentOptimum") \
 245                     or self._toStore("SimulatedObservationAtCurrentOptimum"):
 246                     IndexMin = numpy.argmin( self.StoredVariables["CostFunctionJ"][nbPreviousSteps:] ) + nbPreviousSteps
 247                 if self._toStore("IndexOfOptimum"):
 248                     self.StoredVariables["IndexOfOptimum"].store( IndexMin )
 249                 if self._toStore("CurrentOptimum"):
 250                     self.StoredVariables["CurrentOptimum"].store( self.StoredVariables["Analysis"][IndexMin] )
 251                 if self._toStore("SimulatedObservationAtCurrentOptimum"):
 252                     self.StoredVariables["SimulatedObservationAtCurrentOptimum"].store( self.StoredVariables["SimulatedObservationAtCurrentAnalysis"][IndexMin] )
 253                 if self._toStore("CostFunctionJbAtCurrentOptimum"):
 254                     self.StoredVariables["CostFunctionJbAtCurrentOptimum"].store( self.StoredVariables["CostFunctionJb"][IndexMin] )
 255                 if self._toStore("CostFunctionJoAtCurrentOptimum"):
 256                     self.StoredVariables["CostFunctionJoAtCurrentOptimum"].store( self.StoredVariables["CostFunctionJo"][IndexMin] )
 257                 if self._toStore("CostFunctionJAtCurrentOptimum"):
 258                     self.StoredVariables["CostFunctionJAtCurrentOptimum" ].store( self.StoredVariables["CostFunctionJ" ][IndexMin] )
 259             if self._toStore("APosterioriCovariance"):
 260                 self.StoredVariables["APosterioriCovariance"].store( Pn )
 261             if self._parameters["EstimationOf"] == "Parameters" \
 262                 and J < previousJMinimum:
 263                 previousJMinimum    = J
 264                 XaMin               = Xa
 265                 if self._toStore("APosterioriCovariance"):
 266                     covarianceXaMin = Pn
 267         #
 268         # Stockage final supplémentaire de l'optimum en estimation de paramètres
 269         # ----------------------------------------------------------------------
 270         if self._parameters["EstimationOf"] == "Parameters":
 271             self.StoredVariables["CurrentIterationNumber"].store( len(self.StoredVariables["Analysis"]) )
 272             self.StoredVariables["Analysis"].store( XaMin )
 273             if self._toStore("APosterioriCovariance"):
 274                 self.StoredVariables["APosterioriCovariance"].store( covarianceXaMin )
 275             if self._toStore("BMA"):
 276                 self.StoredVariables["BMA"].store( numpy.ravel(Xb) - numpy.ravel(XaMin) )
 277         #
 278         self._post_run(HO)
 279         return 0
 280
 281 # ==============================================================================
 282 if __name__ == "__main__":
 283     print('\n AUTODIAGNOSTIC\n')