How about the ATE(Absolute Trajectory Error) of RPBA?

Hi, helmayer, thanks to open source the RPBA library. I have implemented point consensus bundle adjust using ADMM, as you said in the paper that it can not converge very well enven with a lot of iteration and the ATE error is huge. Have you ever compare the ATE error of RPBA? 