koide3 / caratheodory2 Goto Github PK

@koide3 Hi author and everyone,
Currently, I am doing the experiment on KITTI dataset.
My approach is that I construct a pose graph with relative pose constraints between the consecutive frames. Then I use GTSAM to optimize the pose graph. Then, I again optimize the sensor poses using pose graph optimization and global registration error minimization (based on GICP scan matching for all frame pairs).
Is my approach right? If it's right, please tell me how to apply the error of GICP error to GTSAM. (Can anyone explain in detail?)
Thank you very much

Hi everyone,

I want to compute the Jacobian matrix from residual. I don't understand the derivatives of a residual ek.
Can anyone help me? Thanh you very much.
Duy Nhat

@koide3 Hi author and everyone,
Currently, I am using g2o to optimize graph. But I want to implement this approach (based on the GICP score) to improve the result. I have some questions:
1> When implementing the GICP factor class, what class can I derive from?
2> Does the error function of the GICP factor class return the score between two corresponding point clouds, representing their respective poses?
3> What are the meanings of parameters like H, b, c in the quadratic function within the GICP factor?
I hope the author and everyone can help me with the answers.
Sincerely.