wh200720041 / floam

Line 83 in 4ba1492

    
           bool PoseSE3Parameterization::Plus(const double *x, const double *delta, double *x_plus_delta) const

Thanks for your great work!
But I have a question why your plus implement is different from the plus implement of the poses in VINS.
As far as I know, your pose is 7 dim global pose(rotation(x,y,z,w),trans(x,y,z)) , and VINS's pose is the same(trans(x,y,z),rotation(x,y,z,w)), and as a consequence, they should be equal to each other, but in fact, they are not.
I don't know the reason for this, can you tell me the reason for this?
Sincerely!
Thank you!

Hi xuwu,

The update is performed on se(3) instead of SE(3), they are different

T_op = exp(delta) * T

If I am not mistaken, in vins it is
T = delta T * T

Thanks

Thank you!

Hi wh,
I'm very sorry to reopen this issue.
But after a lot of attempts, I still failed. The problem is still the implement plus function of localparameterization.
First of all, the local parameter is se(3) , which represents the poses.
In your implement:
quater_plus = delta_q * quater;
trans_plus = delta_q * trans + delta_t;
I know this is right, because it means that T_op = exp(delta)*T;

But the implement of VINS is:
p = _p + dp;
q = (_q * dq).normalized();

The differences between VINSs' and your implement mainly remain on 2 two point:
1.The p_plus of yours is q_deltap+p_delta(I know that it is derived from TT_delta), but the p_plus of VINS is just p+delta_p(I don't know why, at first, I think this is an approximation. But after some experiments, it shows that I was wrong)
2.The q_plus of yours is qq_delta, but in VINS, q_plus is q_deltaq.The order is completely reversed.

I really don't know the reason for this.
Would you solve my problem?
Thank you!
Best wishes.

I know that, the difference between them is the left perturbation model or the right perturbation model.
Thank you.

problem about the plus of localparameterization