278 lines
5.3 KiB
Text
278 lines
5.3 KiB
Text
/*
|
|
* kalmanfilter implementation
|
|
*
|
|
* Copyright (C) YuanJun, Li Wei, Li Ming, Andreas Nuechter,
|
|
*
|
|
* Released under the GPL version 3.
|
|
*
|
|
*/
|
|
|
|
/**
|
|
* @file
|
|
* @brief
|
|
*
|
|
* @author Andreas Nuechter. Jacobs University Bremen, Germany
|
|
* @author YuanJun, Wuhan University, China
|
|
* @author Li Wei, Wuhan University, China
|
|
* @author Li Ming, Wuhan University, China
|
|
*/
|
|
/***************By yuanjun**************************/
|
|
#include "veloslam/kalmanfilter.h"
|
|
#include "math.h"
|
|
#define delta_t 0.1
|
|
|
|
|
|
KalmanFilter::KalmanFilter()
|
|
{
|
|
|
|
}
|
|
|
|
KalmanFilter::KalmanFilter(clusterFeature &glu,double rollAngle)
|
|
{
|
|
X1.SetMatrixRowAndCol(4,1);
|
|
X2.SetMatrixRowAndCol(4,1);
|
|
A.SetMatrixRowAndCol(4,4);
|
|
H.SetMatrixRowAndCol(2,4);
|
|
Z.SetMatrixRowAndCol(2,1);
|
|
K.SetMatrixRowAndCol(4,2);
|
|
Q.SetMatrixRowAndCol(4,4);
|
|
R.SetMatrixRowAndCol(2,2);
|
|
P1.SetMatrixRowAndCol(4,4);
|
|
P2.SetMatrixRowAndCol(4,4);
|
|
CoordinateRoll.SetMatrixRowAndCol(4,4);
|
|
|
|
X1.m_pTMatrix[0][0]=glu.avg_x;
|
|
X1.m_pTMatrix[1][0]=glu.avg_z;
|
|
X1.m_pTMatrix[2][0]=X1.m_pTMatrix[3][0]=25;
|
|
|
|
|
|
A.Eye();
|
|
A.m_pTMatrix[0][2]=delta_t;
|
|
A.m_pTMatrix[1][3]=delta_t;
|
|
|
|
H.m_pTMatrix[0][0]=1;
|
|
H.m_pTMatrix[1][1]=1;
|
|
|
|
Q.Eye();
|
|
|
|
R.Eye();
|
|
|
|
P1.Eye();
|
|
|
|
angle=rollAngle;
|
|
|
|
CalCoorRoll(angle);
|
|
}
|
|
|
|
KalmanFilter::~KalmanFilter(void)
|
|
{
|
|
|
|
}
|
|
|
|
void KalmanFilter::InitialKalmanFilter(clusterFeature &glu)
|
|
{
|
|
X1.SetMatrixRowAndCol(4,1);
|
|
X2.SetMatrixRowAndCol(4,1);
|
|
A.SetMatrixRowAndCol(4,4);
|
|
H.SetMatrixRowAndCol(2,4);
|
|
Z.SetMatrixRowAndCol(2,1);
|
|
K.SetMatrixRowAndCol(4,2);
|
|
Q.SetMatrixRowAndCol(4,4);
|
|
R.SetMatrixRowAndCol(2,2);
|
|
P1.SetMatrixRowAndCol(4,4);
|
|
P2.SetMatrixRowAndCol(4,4);
|
|
|
|
X1.m_pTMatrix[0][0]=glu.avg_x;
|
|
X1.m_pTMatrix[1][0]=glu.avg_z;
|
|
X1.m_pTMatrix[2][0]=X1.m_pTMatrix[3][0]=1;
|
|
|
|
A.Eye();
|
|
A.m_pTMatrix[0][2]=delta_t;
|
|
A.m_pTMatrix[1][3]=delta_t;
|
|
|
|
H.m_pTMatrix[0][0]=1;
|
|
H.m_pTMatrix[1][1]=1;
|
|
|
|
Q.Eye();
|
|
|
|
R.Eye();
|
|
|
|
P1.Eye();
|
|
|
|
}
|
|
|
|
void KalmanFilter::timeUpdate()
|
|
{
|
|
CMatrix T1,T2;
|
|
X2=A*X1;
|
|
T1= A.Transpose();
|
|
P2= A * P1 * T1 + Q;
|
|
// X2=A*X1;
|
|
// P2=A*P1*A.Transpose()+Q;
|
|
}
|
|
|
|
void KalmanFilter::stateUpdate(clusterFeature &glu,double rollangle,double *pos)
|
|
{
|
|
CMatrix T1,T2;
|
|
//Z.m_pTMatrix[0][0]=glu.avg_x;
|
|
//Z.m_pTMatrix[1][0]=glu.avg_z;
|
|
|
|
//CMatrix temp(2,2);
|
|
// T1= H.Transpose();
|
|
//temp=H*P2*T1+R;
|
|
// T2= temp.Inverse();
|
|
//K=P2*T1*T2;
|
|
|
|
//CalCoorRoll(rollangle);
|
|
|
|
//X2=CoordinateRoll*X2;
|
|
|
|
// CMatrix T3,T4,T5;
|
|
|
|
// T4=H*X2;
|
|
// T3=Z-T4;
|
|
// T5=K*T3;
|
|
//X1=X2+T5;
|
|
|
|
//CMatrix I(4,4);
|
|
//I.Eye();
|
|
|
|
// CMatrix T6,T7,T8;
|
|
// T6=K*H;
|
|
// T7=I-T6;
|
|
//P1=T7*P2;
|
|
|
|
Z.m_pTMatrix[0][0]=glu.avg_x;
|
|
Z.m_pTMatrix[1][0]=glu.avg_z;
|
|
|
|
// CMatrix temp(2,2);
|
|
// temp=H*P2*H.Transpose()+R;
|
|
|
|
CMatrix temp(2,2);
|
|
T1= H.Transpose();
|
|
temp=H*P2*T1+R;
|
|
|
|
T2= temp.Inverse();
|
|
K=P2*T1*T2;
|
|
|
|
// K=P2*H.Transpose()*temp.Inverse();
|
|
|
|
CalCoorRoll(rollangle);
|
|
|
|
X2=CoordinateRoll*X2;
|
|
|
|
X2.m_pTMatrix[0][0]+=pos[0];
|
|
X2.m_pTMatrix[1][0]+=pos[2];
|
|
|
|
|
|
// X1=X2+K*(Z-H*X2);
|
|
|
|
CMatrix T3,T4,T5;
|
|
|
|
T4=H*X2;
|
|
T3=Z-T4;
|
|
T5=K*T3;
|
|
X1=X2+T5;
|
|
|
|
CMatrix I(4,4);
|
|
I.Eye();
|
|
|
|
// P1=(I-K*H)*P2;
|
|
|
|
CMatrix T6,T7,T8;
|
|
T6=K*H;
|
|
T7=I-T6;
|
|
P1=T7*P2;
|
|
|
|
}
|
|
|
|
ObjectState KalmanFilter::GetCurrentState()
|
|
{
|
|
ObjectState movestate;
|
|
movestate.x_position=X1.m_pTMatrix[0][0];
|
|
movestate.z_positon=X1.m_pTMatrix[1][0];
|
|
movestate.x_speed=X1.m_pTMatrix[2][0];
|
|
movestate.z_speed=X1.m_pTMatrix[3][0];
|
|
|
|
return movestate;
|
|
}
|
|
|
|
ObjectState KalmanFilter::GetPredictState()
|
|
{
|
|
ObjectState movestate;
|
|
|
|
movestate.x_position=X2.m_pTMatrix[0][0];
|
|
movestate.z_positon=X2.m_pTMatrix[1][0];
|
|
movestate.x_speed=X2.m_pTMatrix[2][0];
|
|
movestate.z_speed=X2.m_pTMatrix[3][0];
|
|
|
|
return movestate;
|
|
}
|
|
|
|
Measurement KalmanFilter::GetPredictMeasurement(double rollAngle,double *pos)
|
|
{
|
|
Measurement predictMeasurement;
|
|
CMatrix temp(2,1);
|
|
CalCoorRoll(rollAngle);
|
|
X2=CoordinateRoll*X2;
|
|
X2.m_pTMatrix[0][0]+=pos[0];
|
|
X2.m_pTMatrix[1][0]+=pos[2];
|
|
temp=H*X2;
|
|
predictMeasurement.x_measurement=temp.m_pTMatrix[0][0];
|
|
predictMeasurement.z_measurement=temp.m_pTMatrix[1][0];
|
|
|
|
return predictMeasurement;
|
|
}
|
|
|
|
CMatrix KalmanFilter::CalMeasureDeviation()
|
|
{
|
|
CMatrix Deviation(2,2),standardDeviation(2,2);
|
|
|
|
CMatrix T6,T7,T8;
|
|
T6=H.Transpose();
|
|
T7=P2*T6;
|
|
Deviation=H*T7+R;
|
|
|
|
// Deviation=H*P2*H.Transpose()+R;
|
|
|
|
for (int i=0;i<2;i++)
|
|
{
|
|
for(int j=0;j<2;j++)
|
|
{
|
|
standardDeviation.m_pTMatrix[i][j]=sqrt(Deviation.m_pTMatrix[i][j]);
|
|
}
|
|
}
|
|
|
|
return standardDeviation;
|
|
}
|
|
|
|
KalmanFilter& KalmanFilter::operator = (const KalmanFilter& anotherKF)
|
|
{
|
|
A=anotherKF.A;
|
|
H=anotherKF.H;
|
|
X1=anotherKF.X1;
|
|
X2=anotherKF.X2;
|
|
K=anotherKF.K;
|
|
Z=anotherKF.Z;
|
|
P1=anotherKF.P1;
|
|
P2=anotherKF.P2;
|
|
Q=anotherKF.Q;
|
|
R=anotherKF.R;
|
|
CoordinateRoll=anotherKF.CoordinateRoll;
|
|
angle=anotherKF.angle;
|
|
return *this;
|
|
}
|
|
|
|
void KalmanFilter::CalCoorRoll(double angle)
|
|
{
|
|
CoordinateRoll.m_pTMatrix[0][0]=cos(angle);//angle is radian??
|
|
CoordinateRoll.m_pTMatrix[0][1]=sin(angle);
|
|
CoordinateRoll.m_pTMatrix[1][0]=-sin(angle);
|
|
CoordinateRoll.m_pTMatrix[1][1]=cos(angle);
|
|
CoordinateRoll.m_pTMatrix[2][2]=cos(angle);
|
|
CoordinateRoll.m_pTMatrix[2][3]=sin(angle);
|
|
CoordinateRoll.m_pTMatrix[3][2]=-sin(angle);
|
|
CoordinateRoll.m_pTMatrix[3][3]=cos(angle);
|
|
}
|
|
|
|
|