void computeSparseAWpB(
  double *A,
  NumericsMatrix *W,
  double *B,
  NumericsMatrix *AWpB)
{

  /* unsigned int problemSize = W->size0; */

  SparseBlockStructuredMatrix* Wb = W->matrix1;

  assert((unsigned)W->size0 >= 3);
  assert((unsigned)W->size0 / 3 >= Wb->filled1 - 1);

  double Ai[9], Bi[9], tmp[9];

  for (unsigned int row = 0, ip9 = 0, i0 = 0;
       row < Wb->filled1 - 1; ++row, ip9 += 9, i0 += 3)
  {
    assert(ip9 < 3 *  (unsigned)W->size0 - 8);

    extract3x3(3, ip9, 0, A, Ai);
    extract3x3(3, ip9, 0, B, Bi);

    for (unsigned int blockn = (unsigned int) Wb->index1_data[row];
         blockn < Wb->index1_data[row + 1]; ++blockn)
    {

      unsigned int col = (unsigned int) Wb->index2_data[blockn];

      mm3x3(Ai, Wb->block[blockn], tmp);
      if (col == row) add3x3(Bi, tmp);

      cpy3x3(tmp, AWpB->matrix1->block[blockn]);
    }
  }
  /* Invalidation of sparse storage, if any. */
  if (AWpB->matrix2)
  {
    if (AWpB->matrix2->triplet)
    {
      cs_spfree(AWpB->matrix2->triplet);
      AWpB->matrix2->triplet = NULL;
    }
    if (AWpB->matrix2->csc)
    {
      cs_spfree(AWpB->matrix2->csc);
      AWpB->matrix2->csc = NULL;
    }
    if (AWpB->matrix2->trans_csc)
    {
      cs_spfree(AWpB->matrix2->trans_csc);
      AWpB->matrix2->trans_csc = NULL;
    }
  }
}
void computeDenseAWpB(
  double *A,
  NumericsMatrix *W,
  double *B,
  NumericsMatrix *AWpB)
{
  unsigned problemSize = W->size0;

  double* result = AWpB->matrix0;

  double* Wx = W->matrix0;

  assert(problemSize >= 3);

  double Wij[9], Ai[9], Bi[9], tmp[9];

  for (unsigned int ip3 = 0, ip9 = 0; ip3 < problemSize; ip3 += 3, ip9 += 9)
  {
    assert(ip9 < 3 * problemSize - 8);

    extract3x3(3, ip9, 0, A, Ai);
    extract3x3(3, ip9, 0, B, Bi);

    for (unsigned int jp3 = 0; jp3 < problemSize; jp3 += 3)
    {
      assert(jp3 < problemSize - 2);
      assert(ip3 < problemSize - 2);

      extract3x3(problemSize, ip3, jp3, Wx, Wij);
      mm3x3(Ai, Wij, tmp);
      if (jp3 == ip3) add3x3(Bi, tmp);
      insert3x3(problemSize, ip3, jp3, result, tmp);

    }
  }
}
int LocalNonsmoothNewtonSolver(FrictionContactProblem* localproblem, double * R, int *iparam, double *dparam)
{
  double mu = localproblem->mu[0];
  double * qLocal = localproblem->q;

  double * MLocal = localproblem->M->matrix0;





  double Tol = dparam[0];
  double itermax = iparam[0];


  int i, j, k, inew;

  // store the increment
  double dR[3] = {0., 0., 0.};

  // store the value fo the function
  double F[3] = {0., 0., 0.};

  // Store the (sub)-gradient of the function
  double A[9] = {0., 0., 0., 0., 0., 0., 0., 0., 0.};
  double B[9] = {0., 0., 0., 0., 0., 0., 0., 0., 0.};

  // Value of AW+B
  double AWplusB[9] = {0., 0., 0., 0., 0., 0., 0., 0., 0.};

  // Compute values of Rho (should be here ?)
  double rho[3] = {1., 1., 1.};
#ifdef OPTI_RHO
  computerho(localproblem, rho);
#endif

  // compute the velocity
  double velocity[3] = {0., 0., 0.};

  for (i = 0; i < 3; i++) velocity[i] = MLocal[i + 0 * 3] * R[0] + qLocal[i]
                                          + MLocal[i + 1 * 3] * R[1] +
                                          + MLocal[i + 2 * 3] * R[2] ;

  for (inew = 0 ; inew < itermax ; ++inew) // Newton iteration
  {
    //Update function and gradient

    Function(R, velocity, mu, rho, F, A, B);

#ifndef AC_Generated
#ifndef NDEBUG
    double Fg[3] = {0., 0., 0.};
    double Ag[9] = {0., 0., 0., 0., 0., 0., 0., 0., 0.};
    double Bg[9] = {0., 0., 0., 0., 0., 0., 0., 0., 0.};
    double AWpB[9];


    assert(*rho > 0. && *(rho + 1) > 0. && *(rho + 2) > 0.);

#ifdef AC_STD
    frictionContact3D_AlartCurnierFunctionGenerated(R, velocity, mu, rho, Fg, Ag, Bg);
#endif

#ifdef AC_JeanMoreau
    frictionContact3D_AlartCurnierJeanMoreauFunctionGenerated(R, velocity, mu, rho, Fg, Ag, Bg);
#endif

    sub3(F, Fg);
    sub3x3(A, Ag);
    sub3x3(B, Bg);

    assert(hypot3(Fg) <= 1e-7);
    assert(hypot9(Ag) <= 1e-7);
    assert(hypot9(Bg) <= 1e-7);
    cpy3x3(A, Ag);
    cpy3x3(B, Bg);
    mm3x3(A, MLocal, AWpB);
    add3x3(B, AWpB);

#endif
#endif


    // compute -(A MLocal +B)
    for (i = 0; i < 3; i++)
    {
      for (j = 0; j < 3; j++)
      {
        AWplusB[i + 3 * j] = 0.0;
        for (k = 0; k < 3; k++)
        {
          AWplusB[i + 3 * j] -= A[i + 3 * k] * MLocal[k + j * 3];
        }
        AWplusB[i + 3 * j] -= B[i + 3 * j];
      }
    }

#ifdef AC_STD
#ifndef NDEBUG
    scal3x3(-1., AWpB);
    sub3x3(AWplusB, AWpB);
    assert(hypot9(AWpB) <= 1e-7);
#endif
#endif

    // Solve the linear system
    solv3x3(AWplusB, dR, F);
    // upate iterates
    R[0] += dR[0];
    R[1] += dR[1];
    R[2] += dR[2];
    // compute new residue
    for (i = 0; i < 3; i++) velocity[i] = MLocal[i + 0 * 3] * R[0] + qLocal[i]
                                            + MLocal[i + 1 * 3] * R[1] +
                                            + MLocal[i + 2 * 3] * R[2] ;
    Function(R, velocity, mu, rho, F, NULL, NULL);
    dparam[1] = 0.5 * (F[0] * F[0] + F[1] * F[1] + F[2] * F[2]) / (1.0 + sqrt(R[0] * R[0] + R[1] * R[1] + R[2] * R[2])) ; // improve with relative tolerance

    /*      dparam[2] =0.0;
            FrictionContact3D_unitary_compute_and_add_error( R , velocity,mu, &(dparam[2]));*/




    if (verbose > 1) printf("-----------------------------------    LocalNewtonSolver number of iteration = %i  error = %.10e \n", inew, dparam[1]);

    if (dparam[1] < Tol)
    {
      /*    printf("-----------------------------------    LocalNewtonSolver number of iteration = %i  error = %.10e \t error2 = %.10e \n",inew,dparam[1], dparam[2]); */

      return 0;
    }

  }// End of the Newton iteration

  /*  printf("-----------------------------------    LocalNewtonSolver number of iteration = %i  error = %.10e \t error2 = %.10e \n",inew,dparam[1], dparam[2]); */
  return 1;

}
Exemple #4
0
void kf(float prevCov[3][3], algp1_msgs::Pose2DWithCovariance action, algp1_msgs::Pose2DWithCovariance measure){
	bool kill = false;
	if(kalmanIteration > 30){ //Base case death
		kill = true;
	}

	//The initial state prediction goes here
	float X_o [3]={		1,
					 	1,
					 	0
					};

	float Z_n [3];
	Z_n[0] = sense_model_.pose2d.x;
	Z_n[1] = sense_model_.pose2d.y;
	Z_n[2] = sense_model_.pose2d.theta;


	float A [3][3] = {{	1,		0,		0},
					 {	0,		1,		0},
					 {	0,		0,		1}};
	float A_t [3][3] =	{{	1,		0,		0},
					 	{	0,		1,		0},
					 	{	0,		0,		1}};
					 				
	float H [3][3] = {{	1,		0,		0},	//plz make this stay I so everything below works
					 {	0,		1,		0},
					 {	0,		0,		1}};

	float P [3][3];
	for(int i = 0; i < 3; i++){
		for(int j = 0; j < 3; j++){
			P[i][j] = prevCov[i][j];
		}
	}

	float Q [3][3] = {{	0.01,	0,		0},
					 {	0,		0.01,	0},
					 {	0,		0,		0.01}};

	float R [3][3] = {{	1,		0,		0},
					 {	0,		1,		0},
					 {	0,		0,		1}};

	float I [3][3] = {{	1,		0,		0},
					 {	0,		1,		0},
					 {	0,		0,		1}};

	//------ State prediction (from velocity model)
	float X_p[2];	
	X_p[0] = action.pose2d.x;
	X_p[1] = action.pose2d.y;
	X_p[2] = action.pose2d.theta;

	
	//------ Covariance Prediction
	float APnm1[3][3];
	multiply3x3(APnm1, A, P);
	
	float APnm1At[3][3];
	multiply3x3(APnm1At, APnm1, A_t);
	
	float P_p[3][3];
	add3x3(P_p, APnm1At, Q);


	//------Innovation
	float Y_tilde[3];

	sub3x1(Y_tilde, Z_n, X_p); //NOTE: In a fully implemented kalman filter where H is not [I], this step is incorrect.

	//------Innovation Covariance
	float HP_p[3][3];
	multiply3x3(HP_p, H, P_p);

	float HP_pHt[3][3];
	multiply3x3(HP_pHt, HP_p, H); //NOTE: In this case I assume H = H^T

	float S[3][3];
	add3x3(S, HP_pHt, R);

	//------Kalman Gain
	float S_inv[3][3];

	try{
		InvMatrix(S, S_inv);
	} catch(const char* errorMessage){
		std::cout << errorMessage << "\n";
	}

	float P_pHt [3][3];
	multiply3x3(P_pHt, P_p, H);//NOTE: In this case I assume H = H^T

	float K[3][3];
	multiply3x3(K, P_pHt, S_inv);

	//------State update (yay!)
	float Ky[3];
	multiply3x1and3x3(Ky, K, Y_tilde);

	float X_n[3];				//New state!
	add3x1(X_n, X_p, Ky);

	//------Covariance update (yay yay!)
	float KH[3][3];
	multiply3x3(KH, K, H);

	float IminusKH[3][3];
	sub3x3(IminusKH, I, KH);

	float P_n[3][3];
	multiply3x3(P_n, IminusKH, P_p);

	//Publish state update
	algp1_msgs::Pose2DWithCovariance state;
	geometry_msgs::PoseStamped stamped;

	state.pose2d.x = X_n[0];
	state.pose2d.y = X_n[1];
	state.pose2d.theta = X_n[2];

	stamped.pose.position.x = X_n[0];
	stamped.pose.position.y = X_n[1];
	stamped.header.seq = 0;
	stamped.header.stamp = ros::Time::now();
	stamped.header.frame_id = "map_static";



	state.covariance[0] = P_n[0][0];
	state.covariance[1] = P_n[0][1];
	state.covariance[2] = P_n[0][2];
	state.covariance[3] = P_n[1][0];
	state.covariance[4] = P_n[1][1];
	state.covariance[5] = P_n[1][2];
	state.covariance[6] = P_n[2][0];
	state.covariance[7] = P_n[2][1];
	state.covariance[8] = P_n[2][2];

	std::cout << "xn"<< std::endl;

	for (int foo = 0; foo < 3; foo++) {
        std::cout << X_n[foo] << " ";
        std::cout << "\n";
    }
    
    std::cout << std::endl;
    std::cout << "pn"<< std::endl;

	print3x3(P_n);


	pub_state_.publish(state);
	pub_stamp_.publish(stamped);
	ROS_INFO("pub'd");

	kalmanIteration++;
	ros::spinOnce();
	ros::Duration(.7).sleep();
	
	if(!kill)
		kf(P_n, vel_model_, sense_model_);
	//keep repeating until base case has been met
}