Confused about code for initial alignment

[vonzy]

Confused about code for initial alignment

[qintonguav]

That is the Jacobian Matrix propagated through each integration step.

see the integration_base_h.

[vonzy]

@qintony I mean I don't know how to construct the Matrix A and b for least square estimation ,for example why the code add tmp_A.transpose() * tmp_A to A in the loop.

[qintonguav]

To solve Ax = b. If A is a rectangle, you can not do the inverse.
A'Ax = A'b If A'A is full rank, you can do the inverse operation and solve it.
sum (tmp_A' * tmp_A) = A'A. prove it by yourself.

[vonzy]

yes! that's right! thank you very much!!!

[vonzy]

@qintony but I am still not aware of why must sum tmp_A to get A, I mean it seems that tmp_Ax=tmp_b,is it right to construct matrix equation like
image

[qintonguav]

How to solve tmp_Ax = tmp_b when tmp_A is the rectangle?
tmp_A' * tmpA... is equal to sum every one.