Covariance matrix?

[facontidavide]

Hi,

first of all, fantastic work. I tested this and got very good results (much better than RF2O)

I wonder which are your thought about defining a covariance matrix, to allow a EKF algorithm to fuse this odometry with another one (wheels).

As it could be expected, this method fails in the case of a long corridor with no features. This is obvious and I don't think we can/should do anything about that. But it would be interesting to detect this at run-time and update the covariance matrix accordingly.

What do you think?

[artivis]

Hi,

If I'm not mistaken, the covariance matrix of the odometry increment is readily available (see cov_odo).