fly316/PX4-Autopilot
7
0
Fork 0
mirror of https://gitee.com/mirrors_PX4/PX4-Autopilot.git synced 2026-10-03 07:28:54 +08:00
Code Issues Packages Projects Releases Wiki Activity

ekf2: do not auto-generate sideslip measurement jacobian

Browse Source 
This is to trade a bit of CPU load for more flash space.
This commit is contained in:
bresch
2025-01-15 11:05:50 -05:00
committed by Daniel Agar
parent e01fef755a
commit c99cb6e94b
4 changed files with 102 additions and 201 deletions
Download Patch File Download Diff File Expand all files Collapse all files
src/modules/ekf2/EKF/aid_sources/sideslip/sideslip_fusion.cpp
+3 -6
View File
@@ -43,7 +43,7 @@
#include "ekf.h"
#include <ekf_derivation/generated/compute_sideslip_innov_and_innov_var.h>
#include <ekf_derivation/generated/compute_sideslip_h_and_k.h>
#include <ekf_derivation/generated/compute_sideslip_h.h>
#include <mathlib/mathlib.h>
@@ -127,10 +127,9 @@ bool Ekf::fuseSideslip(estimator_aid_source1d_s &sideslip)
_fault_status.flags.bad_sideslip = false;
const float epsilon = 1e-3f;
VectorState H; // Observation jacobian
VectorState K; // Kalman gain vector
sym::ComputeSideslipHAndK(_state.vector(), P, sideslip.innovation_variance, epsilon, &H, &K);
const VectorState H = sym::ComputeSideslipH(_state.vector(), epsilon);
VectorState K = P * H / sideslip.innovation_variance;
if (update_wind_only) {
const Vector2f K_wind = K.slice<State::wind_vel.dof, 1>(State::wind_vel.idx, 0);
@@ -143,7 +142,5 @@ bool Ekf::fuseSideslip(estimator_aid_source1d_s &sideslip)
sideslip.fused = true;
sideslip.time_last_fuse = _time_delayed_us;
_fault_status.flags.bad_sideslip = false;
return true;
}
src/modules/ekf2/EKF/python/ekf_derivation/derivation.py
+3 -7
View File
@@ -354,10 +354,8 @@ def compute_sideslip_innov_and_innov_var(
return (innov, innov_var)
def compute_sideslip_h_and_k(
def compute_sideslip_h(
state: VState,
P: MTangent,
innov_var: sf.Scalar,
epsilon: sf.Scalar
) -> (VTangent, VTangent):
@@ -366,9 +364,7 @@ def compute_sideslip_h_and_k(
H = jacobian_chain_rule(sideslip_pred, state)
K = P * H.T / sf.Max(innov_var, epsilon)
return (H.T, K)
return (H.T)
def predict_vel_body(
state: VState
@@ -739,7 +735,7 @@ if not args.disable_wind:
generate_px4_function(compute_airspeed_innov_and_innov_var, output_names=["innov", "innov_var"])
generate_px4_function(compute_drag_x_innov_var_and_h, output_names=["innov_var", "Hx"])
generate_px4_function(compute_drag_y_innov_var_and_h, output_names=["innov_var", "Hy"])
generate_px4_function(compute_sideslip_h_and_k, output_names=["H", "K"])
generate_px4_function(compute_sideslip_h, output_names=None)
generate_px4_function(compute_sideslip_innov_and_innov_var, output_names=["innov", "innov_var"])
generate_px4_function(compute_wind_init_and_cov_from_airspeed, output_names=["wind", "P_wind"])
generate_px4_function(compute_wind_init_and_cov_from_wind_speed_and_direction, output_names=["wind", "P_wind"])
src/modules/ekf2/EKF/python/ekf_derivation/generated/compute_sideslip_h.h
+96
View File
@@ -0,0 +1,96 @@
// -----------------------------------------------------------------------------
// This file was autogenerated by symforce from template:
// function/FUNCTION.h.jinja
// Do NOT modify by hand.
// -----------------------------------------------------------------------------
#pragma once
#include <matrix/math.hpp>
namespace sym {
/**
* This function was autogenerated from a symbolic function. Do not modify by hand.
*
* Symbolic function: compute_sideslip_h
*
* Args:
* state: Matrix25_1
* epsilon: Scalar
*
* Outputs:
* res: Matrix24_1
*/
template <typename Scalar>
matrix::Matrix<Scalar, 24, 1> ComputeSideslipH(const matrix::Matrix<Scalar, 25, 1>& state,
const Scalar epsilon) {
// Total ops: 131
// Input arrays
// Intermediate terms (37)
const Scalar _tmp0 = -state(22, 0) + state(4, 0);
const Scalar _tmp1 = 2 * state(1, 0);
const Scalar _tmp2 = 2 * state(6, 0);
const Scalar _tmp3 = _tmp2 * state(3, 0);
const Scalar _tmp4 = 1 - 2 * std::pow(state(3, 0), Scalar(2));
const Scalar _tmp5 = _tmp4 - 2 * std::pow(state(2, 0), Scalar(2));
const Scalar _tmp6 = 2 * state(0, 0);
const Scalar _tmp7 = _tmp6 * state(3, 0);
const Scalar _tmp8 = 2 * state(2, 0);
const Scalar _tmp9 = _tmp8 * state(1, 0);
const Scalar _tmp10 = _tmp7 + _tmp9;
const Scalar _tmp11 = -state(23, 0) + state(5, 0);
const Scalar _tmp12 = _tmp1 * state(3, 0) - _tmp8 * state(0, 0);
const Scalar _tmp13 = _tmp0 * _tmp5 + _tmp10 * _tmp11 + _tmp12 * state(6, 0);
const Scalar _tmp14 = _tmp13 + epsilon * ((((_tmp13) > 0) - ((_tmp13) < 0)) + Scalar(0.5));
const Scalar _tmp15 = Scalar(1.0) / (_tmp14);
const Scalar _tmp16 = _tmp2 * state(0, 0);
const Scalar _tmp17 = std::pow(_tmp14, Scalar(2));
const Scalar _tmp18 = _tmp4 - 2 * std::pow(state(1, 0), Scalar(2));
const Scalar _tmp19 = -_tmp7 + _tmp9;
const Scalar _tmp20 = _tmp6 * state(1, 0) + _tmp8 * state(3, 0);
const Scalar _tmp21 = _tmp0 * _tmp19 + _tmp11 * _tmp18 + _tmp20 * state(6, 0);
const Scalar _tmp22 = _tmp21 / _tmp17;
const Scalar _tmp23 = _tmp17 / (_tmp17 + std::pow(_tmp21, Scalar(2)));
const Scalar _tmp24 = (Scalar(1) / Scalar(2)) * _tmp23;
const Scalar _tmp25 = _tmp24 * (_tmp15 * (_tmp0 * _tmp1 + _tmp3) -
_tmp22 * (-4 * _tmp0 * state(2, 0) + _tmp1 * _tmp11 - _tmp16));
const Scalar _tmp26 = 2 * state(3, 0);
const Scalar _tmp27 = _tmp2 * state(1, 0);
const Scalar _tmp28 = _tmp2 * state(2, 0);
const Scalar _tmp29 =
_tmp24 * (_tmp15 * (-_tmp0 * _tmp26 + _tmp27) - _tmp22 * (_tmp11 * _tmp26 - _tmp28));
const Scalar _tmp30 = _tmp24 * (_tmp15 * (_tmp0 * _tmp8 - 4 * _tmp11 * state(1, 0) + _tmp16) -
_tmp22 * (_tmp11 * _tmp8 + _tmp3));
const Scalar _tmp31 = 4 * state(3, 0);
const Scalar _tmp32 = _tmp24 * (_tmp15 * (-_tmp0 * _tmp6 - _tmp11 * _tmp31 + _tmp28) -
_tmp22 * (-_tmp0 * _tmp31 + _tmp11 * _tmp6 + _tmp27));
const Scalar _tmp33 = _tmp22 * _tmp5;
const Scalar _tmp34 = _tmp15 * _tmp19;
const Scalar _tmp35 = _tmp15 * _tmp18;
const Scalar _tmp36 = _tmp10 * _tmp22;
// Output terms (1)
matrix::Matrix<Scalar, 24, 1> _res;
_res.setZero();
_res(0, 0) =
-_tmp25 * state(3, 0) - _tmp29 * state(1, 0) + _tmp30 * state(0, 0) + _tmp32 * state(2, 0);
_res(1, 0) =
_tmp25 * state(0, 0) - _tmp29 * state(2, 0) + _tmp30 * state(3, 0) - _tmp32 * state(1, 0);
_res(2, 0) =
_tmp25 * state(1, 0) - _tmp29 * state(3, 0) - _tmp30 * state(2, 0) + _tmp32 * state(0, 0);
_res(3, 0) = _tmp23 * (-_tmp33 + _tmp34);
_res(4, 0) = _tmp23 * (_tmp35 - _tmp36);
_res(5, 0) = _tmp23 * (-_tmp12 * _tmp22 + _tmp15 * _tmp20);
_res(21, 0) = _tmp23 * (_tmp33 - _tmp34);
_res(22, 0) = _tmp23 * (-_tmp35 + _tmp36);
return _res;
} // NOLINT(readability/fn_size)
// NOLINTNEXTLINE(readability/fn_size)
} // namespace sym
src/modules/ekf2/EKF/python/ekf_derivation/generated/compute_sideslip_h_and_k.h
-188
View File
@@ -1,188 +0,0 @@
// -----------------------------------------------------------------------------
// This file was autogenerated by symforce from template:
// function/FUNCTION.h.jinja
// Do NOT modify by hand.
// -----------------------------------------------------------------------------
#pragma once
#include <matrix/math.hpp>
namespace sym {
/**
* This function was autogenerated from a symbolic function. Do not modify by hand.
*
* Symbolic function: compute_sideslip_h_and_k
*
* Args:
* state: Matrix25_1
* P: Matrix24_24
* innov_var: Scalar
* epsilon: Scalar
*
* Outputs:
* H: Matrix24_1
* K: Matrix24_1
*/
template <typename Scalar>
void ComputeSideslipHAndK(const matrix::Matrix<Scalar, 25, 1>& state,
const matrix::Matrix<Scalar, 24, 24>& P, const Scalar innov_var,
const Scalar epsilon, matrix::Matrix<Scalar, 24, 1>* const H = nullptr,
matrix::Matrix<Scalar, 24, 1>* const K = nullptr) {
// Total ops: 518
// Input arrays
// Intermediate terms (50)
const Scalar _tmp0 = -state(22, 0) + state(4, 0);
const Scalar _tmp1 = 2 * _tmp0;
const Scalar _tmp2 = 2 * state(3, 0);
const Scalar _tmp3 = _tmp2 * state(6, 0);
const Scalar _tmp4 = 1 - 2 * std::pow(state(3, 0), Scalar(2));
const Scalar _tmp5 = _tmp4 - 2 * std::pow(state(2, 0), Scalar(2));
const Scalar _tmp6 = _tmp2 * state(0, 0);
const Scalar _tmp7 = 2 * state(1, 0);
const Scalar _tmp8 = _tmp7 * state(2, 0);
const Scalar _tmp9 = _tmp6 + _tmp8;
const Scalar _tmp10 = -state(23, 0) + state(5, 0);
const Scalar _tmp11 = 2 * state(2, 0);
const Scalar _tmp12 = -_tmp11 * state(0, 0) + _tmp2 * state(1, 0);
const Scalar _tmp13 = _tmp0 * _tmp5 + _tmp10 * _tmp9 + _tmp12 * state(6, 0);
const Scalar _tmp14 = _tmp13 + epsilon * ((((_tmp13) > 0) - ((_tmp13) < 0)) + Scalar(0.5));
const Scalar _tmp15 = Scalar(1.0) / (_tmp14);
const Scalar _tmp16 = 2 * _tmp10;
const Scalar _tmp17 = 2 * state(0, 0) * state(6, 0);
const Scalar _tmp18 = std::pow(_tmp14, Scalar(2));
const Scalar _tmp19 = _tmp4 - 2 * std::pow(state(1, 0), Scalar(2));
const Scalar _tmp20 = -_tmp6 + _tmp8;
const Scalar _tmp21 = _tmp2 * state(2, 0) + _tmp7 * state(0, 0);
const Scalar _tmp22 = _tmp0 * _tmp20 + _tmp10 * _tmp19 + _tmp21 * state(6, 0);
const Scalar _tmp23 = _tmp22 / _tmp18;
const Scalar _tmp24 = _tmp15 * (_tmp1 * state(1, 0) + _tmp3) -
_tmp23 * (-4 * _tmp0 * state(2, 0) + _tmp16 * state(1, 0) - _tmp17);
const Scalar _tmp25 = _tmp18 / (_tmp18 + std::pow(_tmp22, Scalar(2)));
const Scalar _tmp26 = (Scalar(1) / Scalar(2)) * _tmp25;
const Scalar _tmp27 = _tmp26 * state(3, 0);
const Scalar _tmp28 = _tmp7 * state(6, 0);
const Scalar _tmp29 = _tmp11 * state(6, 0);
const Scalar _tmp30 =
_tmp15 * (-_tmp1 * state(3, 0) + _tmp28) - _tmp23 * (_tmp16 * state(3, 0) - _tmp29);
const Scalar _tmp31 = _tmp26 * state(1, 0);
const Scalar _tmp32 = _tmp15 * (_tmp1 * state(2, 0) - 4 * _tmp10 * state(1, 0) + _tmp17) -
_tmp23 * (_tmp16 * state(2, 0) + _tmp3);
const Scalar _tmp33 = _tmp26 * state(0, 0);
const Scalar _tmp34 = 4 * state(3, 0);
const Scalar _tmp35 = _tmp15 * (-_tmp1 * state(0, 0) - _tmp10 * _tmp34 + _tmp29) -
_tmp23 * (-_tmp0 * _tmp34 + _tmp16 * state(0, 0) + _tmp28);
const Scalar _tmp36 = _tmp26 * state(2, 0);
const Scalar _tmp37 = -_tmp24 * _tmp27 - _tmp30 * _tmp31 + _tmp32 * _tmp33 + _tmp35 * _tmp36;
const Scalar _tmp38 = _tmp24 * _tmp33 + _tmp27 * _tmp32 - _tmp30 * _tmp36 - _tmp31 * _tmp35;
const Scalar _tmp39 = _tmp24 * _tmp31 - _tmp27 * _tmp30 - _tmp32 * _tmp36 + _tmp33 * _tmp35;
const Scalar _tmp40 = _tmp23 * _tmp5;
const Scalar _tmp41 = _tmp15 * _tmp20;
const Scalar _tmp42 = _tmp25 * (-_tmp40 + _tmp41);
const Scalar _tmp43 = _tmp15 * _tmp19;
const Scalar _tmp44 = _tmp23 * _tmp9;
const Scalar _tmp45 = _tmp25 * (_tmp43 - _tmp44);
const Scalar _tmp46 = _tmp25 * (-_tmp12 * _tmp23 + _tmp15 * _tmp21);
const Scalar _tmp47 = _tmp25 * (_tmp40 - _tmp41);
const Scalar _tmp48 = _tmp25 * (-_tmp43 + _tmp44);
const Scalar _tmp49 = Scalar(1.0) / (math::max<Scalar>(epsilon, innov_var));
// Output terms (2)
if (H != nullptr) {
matrix::Matrix<Scalar, 24, 1>& _h = (*H);
_h.setZero();
_h(0, 0) = _tmp37;
_h(1, 0) = _tmp38;
_h(2, 0) = _tmp39;
_h(3, 0) = _tmp42;
_h(4, 0) = _tmp45;
_h(5, 0) = _tmp46;
_h(21, 0) = _tmp47;
_h(22, 0) = _tmp48;
}
if (K != nullptr) {
matrix::Matrix<Scalar, 24, 1>& _k = (*K);
_k(0, 0) =
_tmp49 * (P(0, 0) * _tmp37 + P(0, 1) * _tmp38 + P(0, 2) * _tmp39 + P(0, 21) * _tmp47 +
P(0, 22) * _tmp48 + P(0, 3) * _tmp42 + P(0, 4) * _tmp45 + P(0, 5) * _tmp46);
_k(1, 0) =
_tmp49 * (P(1, 0) * _tmp37 + P(1, 1) * _tmp38 + P(1, 2) * _tmp39 + P(1, 21) * _tmp47 +
P(1, 22) * _tmp48 + P(1, 3) * _tmp42 + P(1, 4) * _tmp45 + P(1, 5) * _tmp46);
_k(2, 0) =
_tmp49 * (P(2, 0) * _tmp37 + P(2, 1) * _tmp38 + P(2, 2) * _tmp39 + P(2, 21) * _tmp47 +
P(2, 22) * _tmp48 + P(2, 3) * _tmp42 + P(2, 4) * _tmp45 + P(2, 5) * _tmp46);
_k(3, 0) =
_tmp49 * (P(3, 0) * _tmp37 + P(3, 1) * _tmp38 + P(3, 2) * _tmp39 + P(3, 21) * _tmp47 +
P(3, 22) * _tmp48 + P(3, 3) * _tmp42 + P(3, 4) * _tmp45 + P(3, 5) * _tmp46);
_k(4, 0) =
_tmp49 * (P(4, 0) * _tmp37 + P(4, 1) * _tmp38 + P(4, 2) * _tmp39 + P(4, 21) * _tmp47 +
P(4, 22) * _tmp48 + P(4, 3) * _tmp42 + P(4, 4) * _tmp45 + P(4, 5) * _tmp46);
_k(5, 0) =
_tmp49 * (P(5, 0) * _tmp37 + P(5, 1) * _tmp38 + P(5, 2) * _tmp39 + P(5, 21) * _tmp47 +
P(5, 22) * _tmp48 + P(5, 3) * _tmp42 + P(5, 4) * _tmp45 + P(5, 5) * _tmp46);
_k(6, 0) =
_tmp49 * (P(6, 0) * _tmp37 + P(6, 1) * _tmp38 + P(6, 2) * _tmp39 + P(6, 21) * _tmp47 +
P(6, 22) * _tmp48 + P(6, 3) * _tmp42 + P(6, 4) * _tmp45 + P(6, 5) * _tmp46);
_k(7, 0) =
_tmp49 * (P(7, 0) * _tmp37 + P(7, 1) * _tmp38 + P(7, 2) * _tmp39 + P(7, 21) * _tmp47 +
P(7, 22) * _tmp48 + P(7, 3) * _tmp42 + P(7, 4) * _tmp45 + P(7, 5) * _tmp46);
_k(8, 0) =
_tmp49 * (P(8, 0) * _tmp37 + P(8, 1) * _tmp38 + P(8, 2) * _tmp39 + P(8, 21) * _tmp47 +
P(8, 22) * _tmp48 + P(8, 3) * _tmp42 + P(8, 4) * _tmp45 + P(8, 5) * _tmp46);
_k(9, 0) =
_tmp49 * (P(9, 0) * _tmp37 + P(9, 1) * _tmp38 + P(9, 2) * _tmp39 + P(9, 21) * _tmp47 +
P(9, 22) * _tmp48 + P(9, 3) * _tmp42 + P(9, 4) * _tmp45 + P(9, 5) * _tmp46);
_k(10, 0) =
_tmp49 * (P(10, 0) * _tmp37 + P(10, 1) * _tmp38 + P(10, 2) * _tmp39 + P(10, 21) * _tmp47 +
P(10, 22) * _tmp48 + P(10, 3) * _tmp42 + P(10, 4) * _tmp45 + P(10, 5) * _tmp46);
_k(11, 0) =
_tmp49 * (P(11, 0) * _tmp37 + P(11, 1) * _tmp38 + P(11, 2) * _tmp39 + P(11, 21) * _tmp47 +
P(11, 22) * _tmp48 + P(11, 3) * _tmp42 + P(11, 4) * _tmp45 + P(11, 5) * _tmp46);
_k(12, 0) =
_tmp49 * (P(12, 0) * _tmp37 + P(12, 1) * _tmp38 + P(12, 2) * _tmp39 + P(12, 21) * _tmp47 +
P(12, 22) * _tmp48 + P(12, 3) * _tmp42 + P(12, 4) * _tmp45 + P(12, 5) * _tmp46);
_k(13, 0) =
_tmp49 * (P(13, 0) * _tmp37 + P(13, 1) * _tmp38 + P(13, 2) * _tmp39 + P(13, 21) * _tmp47 +
P(13, 22) * _tmp48 + P(13, 3) * _tmp42 + P(13, 4) * _tmp45 + P(13, 5) * _tmp46);
_k(14, 0) =
_tmp49 * (P(14, 0) * _tmp37 + P(14, 1) * _tmp38 + P(14, 2) * _tmp39 + P(14, 21) * _tmp47 +
P(14, 22) * _tmp48 + P(14, 3) * _tmp42 + P(14, 4) * _tmp45 + P(14, 5) * _tmp46);
_k(15, 0) =
_tmp49 * (P(15, 0) * _tmp37 + P(15, 1) * _tmp38 + P(15, 2) * _tmp39 + P(15, 21) * _tmp47 +
P(15, 22) * _tmp48 + P(15, 3) * _tmp42 + P(15, 4) * _tmp45 + P(15, 5) * _tmp46);
_k(16, 0) =
_tmp49 * (P(16, 0) * _tmp37 + P(16, 1) * _tmp38 + P(16, 2) * _tmp39 + P(16, 21) * _tmp47 +
P(16, 22) * _tmp48 + P(16, 3) * _tmp42 + P(16, 4) * _tmp45 + P(16, 5) * _tmp46);
_k(17, 0) =
_tmp49 * (P(17, 0) * _tmp37 + P(17, 1) * _tmp38 + P(17, 2) * _tmp39 + P(17, 21) * _tmp47 +
P(17, 22) * _tmp48 + P(17, 3) * _tmp42 + P(17, 4) * _tmp45 + P(17, 5) * _tmp46);
_k(18, 0) =
_tmp49 * (P(18, 0) * _tmp37 + P(18, 1) * _tmp38 + P(18, 2) * _tmp39 + P(18, 21) * _tmp47 +
P(18, 22) * _tmp48 + P(18, 3) * _tmp42 + P(18, 4) * _tmp45 + P(18, 5) * _tmp46);
_k(19, 0) =
_tmp49 * (P(19, 0) * _tmp37 + P(19, 1) * _tmp38 + P(19, 2) * _tmp39 + P(19, 21) * _tmp47 +
P(19, 22) * _tmp48 + P(19, 3) * _tmp42 + P(19, 4) * _tmp45 + P(19, 5) * _tmp46);
_k(20, 0) =
_tmp49 * (P(20, 0) * _tmp37 + P(20, 1) * _tmp38 + P(20, 2) * _tmp39 + P(20, 21) * _tmp47 +
P(20, 22) * _tmp48 + P(20, 3) * _tmp42 + P(20, 4) * _tmp45 + P(20, 5) * _tmp46);
_k(21, 0) =
_tmp49 * (P(21, 0) * _tmp37 + P(21, 1) * _tmp38 + P(21, 2) * _tmp39 + P(21, 21) * _tmp47 +
P(21, 22) * _tmp48 + P(21, 3) * _tmp42 + P(21, 4) * _tmp45 + P(21, 5) * _tmp46);
_k(22, 0) =
_tmp49 * (P(22, 0) * _tmp37 + P(22, 1) * _tmp38 + P(22, 2) * _tmp39 + P(22, 21) * _tmp47 +
P(22, 22) * _tmp48 + P(22, 3) * _tmp42 + P(22, 4) * _tmp45 + P(22, 5) * _tmp46);
_k(23, 0) =
_tmp49 * (P(23, 0) * _tmp37 + P(23, 1) * _tmp38 + P(23, 2) * _tmp39 + P(23, 21) * _tmp47 +
P(23, 22) * _tmp48 + P(23, 3) * _tmp42 + P(23, 4) * _tmp45 + P(23, 5) * _tmp46);
}
} // NOLINT(readability/fn_size)
// NOLINTNEXTLINE(readability/fn_size)
} // namespace sym