Jacobian question again

[vonzy]

Hello, I have read paper 'Tightly-Coupled Monocular Visual-Inertial Fusion for Autonomous Flight of Rotorcraft MAVs' ,'Technical Report VINS-Mono: A Robust and Versatile Monocular Visual-Inertial State Estimator' and 'IMU Preintegration on Manifold for Efficient Visual-Inertial Maximum-a-Posteriori Estimation' carefully, and found there are some differences on jacobian calculation between the code and the papers. In paper 'Tightly-Coupled Monocular Visual-Inertial Fusion for Autonomous Flight of Rotorcraft MAVs' ,the jacobian for p,v,q part is but in the code, it's Also, in the paper 'IMU Preintegration on Manifold for Efficient Visual-Inertial Maximum-a-Posteriori Estimation', the jacobians are There still some differences for bias part. I wonder what cause the difference. And I also noticed that in the code the jacobian for attitude is defined for two situations: I don't know what this is for, could you please explain it?

[vonzy]

@shaojie @qintony @PeiliangLi

[qintonguav]

The difference is caused by the different definition of variables and residual function. I suggest you derivate it by yourself then you can know the difference behind them.

[vonzy]

So why the jacobian of attitude is defined for two situation?

[qintonguav]

The comment version is used when bias estimation is off. We keep it in debug purpose, to valid the IMU correction is right.