hello，It's still the issue we discussed the day before yesterday about weight.

[alonly2024]

@eupedrosa hello，It's still the issue we discussed the day before yesterday.
I want to integrate the wheel speed odometer into solver preblem.
Therefore, I fabricated two points (0,0.25) and (0.25.0), which will participate in the solver as new target points calculated based on wheel speed mileage(odom).
The following figure.

but two opints weight is too small.
so i add Many identical points to increase weight.At present, this approach has a good effect.Only matching a small outline can achieve localization.
But I still think it's not good to do this. I want to increase the weight of these two points separately to 10 or even 100.
So,i change solver.cpp the weight.But none of them are correct.
So,Can you tell me how to add weights in solver.cpp，Realize the effect that one point is equivalent to multiple points.
or,do you have a better way to integrate wheel speed odometry, laser odometry, or other characterized points into this minimal quadratic solver.
Thank you very much.

[eupedrosa]

It may not be the best option to generate artificial points to integrate the wheel speed odometry. By adding multiple points you are adding bias to the scan matching. LaMa's solver.cpp is not the best solver for what you want.
Maybe you should use the wheel speed to estimate an odometry value and feed that to the algorithm.

you have a better way to integrate wheel speed odometry, laser odometry, or other characterized points into this minimal quadratic solver.

You have the classic Kalman filter to fuse your estimates.