Skip to content

Compute Kálmán filter chi2 from predicted state #1185

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Draft
stephenswat wants to merge 1 commit into
base: main
Choose a base branch
from
Draft

Compute Kálmán filter chi2 from predicted state #1185

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
49 changes: 25 additions & 24 deletions core/include/traccc/fitting/kalman_filter/gain_matrix_updater.hpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -74,7 +74,6 @@ struct gain_matrix_updater {
// Some identity matrices
// @TODO: Make constexpr work
const auto I66 = matrix::identity<bound_matrix_type>();
const auto I_m = matrix::identity<matrix_type<D, D>>();

// Measurement data on surface
matrix_type<D, 1> meas_local;
Expand Down Expand Up @@ -123,6 +122,31 @@ struct gain_matrix_updater {
const matrix_type<e_bound_size, D> projected_cov =
algebra::matrix::transposed_product<false, true>(predicted_cov, H);

{
// Calculate the chi square
const matrix_type<D, D> R =
V + (H * predicted_cov * algebra::matrix::transpose(H));
// Residual between measurement and predicted vector
const matrix_type<D, 1> residual = meas_local - H * predicted_vec;
const matrix_type<1, 1> chi2_mat =
algebra::matrix::transposed_product<true, false>(
residual, matrix::inverse(R)) *
residual;
const scalar chi2_val = getter::element(chi2_mat, 0, 0);

if (chi2_val < 0.f) {
TRACCC_ERROR_HOST_DEVICE("Chi2 negative");
return kalman_fitter_status::ERROR_UPDATER_CHI2_NEGATIVE;
}

if (!std::isfinite(chi2_val)) {
TRACCC_ERROR_HOST_DEVICE("Chi2 infinite");
return kalman_fitter_status::ERROR_UPDATER_CHI2_NOT_FINITE;
}

trk_state.filtered_chi2() = chi2_val;
}

const matrix_type<D, D> M = H * projected_cov + V;

// Kalman gain matrix
Expand Down Expand Up @@ -162,38 +186,15 @@ struct gain_matrix_updater {
return kalman_fitter_status::ERROR_QOP_ZERO;
}

// Residual between measurement and (projected) filtered vector
const matrix_type<D, 1> residual = meas_local - H * filtered_vec;

// Calculate the chi square
const matrix_type<D, D> R = (I_m - H * K) * V;
const matrix_type<1, 1> chi2 =
algebra::matrix::transposed_product<true, false>(
residual, matrix::inverse(R)) *
residual;

const scalar chi2_val{getter::element(chi2, 0, 0)};

TRACCC_VERBOSE_HOST("Filtered residual: " << residual);
TRACCC_DEBUG_HOST("R:\n" << R);
TRACCC_DEBUG_HOST_DEVICE("det(R): %f", matrix::determinant(R));
TRACCC_DEBUG_HOST("R_inv:\n" << matrix::inverse(R));
TRACCC_VERBOSE_HOST_DEVICE("Chi2: %f", chi2_val);

if (chi2_val < 0.f) {
TRACCC_ERROR_HOST_DEVICE("Chi2 negative");
return kalman_fitter_status::ERROR_UPDATER_CHI2_NEGATIVE;
}

if (!std::isfinite(chi2_val)) {
TRACCC_ERROR_HOST_DEVICE("Chi2 infinite");
return kalman_fitter_status::ERROR_UPDATER_CHI2_NOT_FINITE;
}

// Set the chi2 for this track and measurement
trk_state.filtered_params().set_vector(filtered_vec);
trk_state.filtered_params().set_covariance(filtered_cov);
trk_state.filtered_chi2() = chi2_val;

if (math::fmod(trk_state.filtered_params().theta(),
2.f * constant<traccc::scalar>::pi) == 0.f) {
Expand Down
104 changes: 51 additions & 53 deletions core/include/traccc/fitting/kalman_filter/two_filters_smoother.hpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -84,6 +84,27 @@ struct two_filters_smoother {
const matrix_type<e_bound_size, 1>& predicted_vec =
bound_params.vector();

matrix_type<D, e_bound_size> H =
measurements.at(trk_state.measurement_index())
.subs.template projector<D>();
if (dim == 1) {
getter::element(H, 1u, 0u) = 0.f;
getter::element(H, 1u, 1u) = 0.f;
}

// Spatial resolution (Measurement covariance)
matrix_type<D, D> V;
edm::get_measurement_covariance<algebra_t>(
measurements.at(trk_state.measurement_index()), V);
if (dim == 1) {
getter::element(V, 1u, 1u) = 1.f;
}

TRACCC_DEBUG_HOST("Measurement position: " << meas_local);
TRACCC_DEBUG_HOST("Measurement variance:\n" << V);
TRACCC_DEBUG_HOST("Predicted residual: " << meas_local -
H * predicted_vec);

// Predicted covaraince of bound track parameters
const matrix_type<e_bound_size, e_bound_size>& predicted_cov =
bound_params.covariance();
Expand Down Expand Up @@ -134,29 +155,8 @@ struct two_filters_smoother {
// Wrap the phi and theta angles in their valid ranges
normalize_angles(trk_state.smoothed_params());

matrix_type<D, e_bound_size> H =
measurements.at(trk_state.measurement_index())
.subs.template projector<D>();
if (dim == 1) {
getter::element(H, 1u, 0u) = 0.f;
getter::element(H, 1u, 1u) = 0.f;
}

const matrix_type<D, 1> residual_smt = meas_local - H * smoothed_vec;

// Spatial resolution (Measurement covariance)
matrix_type<D, D> V;
edm::get_measurement_covariance<algebra_t>(
measurements.at(trk_state.measurement_index()), V);
if (dim == 1) {
getter::element(V, 1u, 1u) = 1.f;
}

TRACCC_DEBUG_HOST("Measurement position: " << meas_local);
TRACCC_DEBUG_HOST("Measurement variance:\n" << V);
TRACCC_DEBUG_HOST("Predicted residual: " << meas_local -
H * predicted_vec);

// Eq (3.39) of "Pattern Recognition, Tracking and Vertex
// Reconstruction in Particle Detectors"
const matrix_type<D, D> R_smt =
Expand Down Expand Up @@ -204,7 +204,6 @@ struct two_filters_smoother {

const auto I66 =
matrix::identity<matrix_type<e_bound_size, e_bound_size>>();
const auto I_m = matrix::identity<matrix_type<D, D>>();

const matrix_type<e_bound_size, D> projected_cov =
algebra::matrix::transposed_product<false, true>(predicted_cov, H);
Expand Down Expand Up @@ -252,39 +251,38 @@ struct two_filters_smoother {

bound_params.set_covariance(filtered_cov);

// Residual between measurement and (projected) filtered vector
const matrix_type<D, 1> residual = meas_local - H * filtered_vec;

// Calculate backward chi2
const matrix_type<D, D> R = (I_m - H * K) * V;
// assert(matrix::determinant(R) != 0.f); // @TODO: This fails
assert(std::isfinite(matrix::determinant(R)));
const matrix_type<1, 1> chi2 =
algebra::matrix::transposed_product<true, false>(
residual, matrix::inverse(R)) *
residual;

const scalar chi2_val{getter::element(chi2, 0, 0)};

TRACCC_VERBOSE_HOST("Filtered residual: " << residual);
TRACCC_DEBUG_HOST("R:\n" << R);
TRACCC_DEBUG_HOST_DEVICE("det(R): %f", matrix::determinant(R));
TRACCC_DEBUG_HOST("R_inv:\n" << matrix::inverse(R));
TRACCC_VERBOSE_HOST_DEVICE("Chi2: %f", chi2_val);

if (chi2_val < 0.f) {
TRACCC_ERROR_HOST_DEVICE("Chi2 negative: %f", chi2_val);
return kalman_fitter_status::ERROR_SMOOTHER_CHI2_NEGATIVE;
}

if (!std::isfinite(chi2_val)) {
TRACCC_ERROR_HOST_DEVICE("Chi2 infinite");
return kalman_fitter_status::ERROR_SMOOTHER_CHI2_NOT_FINITE;
{
// Calculate the chi square
const matrix_type<D, D> R =
V - (H * predicted_cov * algebra::matrix::transpose(H));
// Residual between measurement and predicted vector
const matrix_type<D, 1> residual = meas_local - H * predicted_vec;
const matrix_type<1, 1> chi2_mat =
algebra::matrix::transposed_product<true, false>(
residual, matrix::inverse(R)) *
residual;
const scalar chi2_val = getter::element(chi2_mat, 0, 0);

TRACCC_VERBOSE_HOST("Predicted residual: " << residual);
TRACCC_DEBUG_HOST("R:\n" << R);
TRACCC_DEBUG_HOST_DEVICE("det(R): %f", matrix::determinant(R));
TRACCC_DEBUG_HOST("R_inv:\n" << matrix::inverse(R));
TRACCC_VERBOSE_HOST_DEVICE("Chi2: %f", chi2_val);

if (chi2_val < 0.f) {
TRACCC_ERROR_HOST_DEVICE("Chi2 negative: %f", chi2_val);
return kalman_fitter_status::ERROR_SMOOTHER_CHI2_NEGATIVE;
}

if (!std::isfinite(chi2_val)) {
TRACCC_ERROR_HOST_DEVICE("Chi2 infinite");
return kalman_fitter_status::ERROR_SMOOTHER_CHI2_NOT_FINITE;
}

// Set backward chi2
trk_state.backward_chi2() = chi2_val;
}

// Set backward chi2
trk_state.backward_chi2() = chi2_val;

// Wrap the phi and theta angles in their valid ranges
normalize_angles(bound_params);

Expand Down
Loading