mirror of
https://bitbucket.org/librepilot/librepilot.git
synced 2024-11-29 07:24:13 +01:00
900780e10c
the gains on acceleration and velocity feedback terms (they are a problem when biased).
70 lines
2.5 KiB
Matlab
70 lines
2.5 KiB
Matlab
% Generate the symbolic code for the kalman filter on altitude
|
|
|
|
dT = sym('dT','real');
|
|
A = [1 dT 0 0; 0 1 dT 0; 0 0 1 0; 0 0 0 1];
|
|
%Nu = diag([sym('V[1]') sym('V[2]') sym('V[3]') sym('V[4]')]);
|
|
%Nu = [sym('V[1][1]') 0 0 0; ...
|
|
% 0 sym('V[2][2]') 0 0; ...
|
|
% 0 0 sym('V[3][3]') sym('V[3][4]'); ...
|
|
% 0 0 sym('V[4][3]') sym('V[4][4]')];
|
|
Nu = [sym('V[1][1]') sym('V[1][2]') sym('V[1][3]') sym('V[1][4]'); ...
|
|
sym('V[2][1]') sym('V[2][2]') sym('V[2][3]') sym('V[2][4]'); ...
|
|
sym('V[3][1]') sym('V[3][2]') sym('V[3][3]') sym('V[3][4]'); ...
|
|
sym('V[4][1]') sym('V[4][2]') sym('V[4][3]') sym('V[4][4]');];
|
|
|
|
|
|
Gamma = diag([sym('G[1]') sym('G[2]') sym('G[3]') sym('G[4]')]);
|
|
Sigma = diag([sym('S[1]') sym('S[2]')]);
|
|
C = [1 0 0 0; 0 0 1 1];
|
|
state = [sym('z[1]'); sym('z[2]'); sym('z[3]'); sym('z[4]')];
|
|
measure = [sym('x[1]'); sym('x[2]')];
|
|
|
|
P = simplify(A * Nu * A' + Gamma);
|
|
K = simplify(P*C'*(C*P*C'+Sigma)^-1);
|
|
|
|
% fill in the zeros from above equations to make next calculation sparse
|
|
P_mat = [sym('P[1][1]') sym('P[1][2]') sym('P[1][3]') sym('P[1][4]'); ...
|
|
sym('P[2][1]') sym('P[2][2]') sym('P[2][3]') sym('P[2][4]'); ...
|
|
sym('P[3][1]') sym('P[3][2]') sym('P[3][3]') sym('P[3][4]'); ...
|
|
sym('P[4][1]') sym('P[4][3]') sym('P[4][3]') sym('P[4][4]')];
|
|
K_mat = [sym('K[1][1]') sym('K[1][2]'); ...
|
|
sym('K[2][1]') sym('K[2][2]'); ...
|
|
sym('K[3][1]') sym('K[3][2]'); ...
|
|
sym('K[4][1]') sym('K[4][2]')];
|
|
|
|
z_new = A * state + K_mat * (measure - C * A * state);
|
|
V = (eye(4) - K_mat * C) * P_mat;
|
|
|
|
ccode(P)
|
|
ccode(K)
|
|
ccode(z_new)
|
|
ccode(V)
|
|
|
|
|
|
%% For when there is no baro update
|
|
% Generate the symbolic code for the kalman filter on altitude
|
|
C = [0 0 1 1];
|
|
Sigma = sym('S[2]');
|
|
measure = [sym('x[2]')];
|
|
|
|
P = simplify(A * Nu * A' + Gamma);
|
|
K = simplify(P*C'*(C*P*C'+Sigma)^-1);
|
|
|
|
% fill in the zeros from above equations to make next calculation sparse
|
|
P_mat = [sym('P[1][1]') sym('P[1][2]') sym('P[1][3]') sym('P[1][4]'); ...
|
|
sym('P[2][1]') sym('P[2][2]') sym('P[2][3]') sym('P[2][4]'); ...
|
|
sym('P[3][1]') sym('P[3][2]') sym('P[3][3]') sym('P[3][4]'); ...
|
|
sym('P[4][1]') sym('P[4][3]') sym('P[4][3]') sym('P[4][4]')];
|
|
K_mat = [sym('K[1][1]'); ...
|
|
sym('K[2][1]'); ...
|
|
sym('K[3][1]'); ...
|
|
sym('K[4][1]')];
|
|
|
|
z_new = A * state + K_mat * (measure - C * A * state);
|
|
V = (eye(4) - K_mat * C) * P_mat;
|
|
|
|
ccode(P)
|
|
ccode(K)
|
|
ccode(z_new)
|
|
ccode(V)
|