We were discussing about this during the last ICS-PFI-MCS telecon, have you started to work on this yet ?

I'm working on an way of flagging misidentifications automatically, but haven't got results yet. 

I've got a method for flagging potential misidentification, and am adding some diagnostic plotting routines to keep track of what's happening, and where things go wrong. We are getting cases here and there, and the effect of the misidentification on convergence is very similar to what was predicted by simulations. 

Please add a comment if you have have favourite misidentifications to contribute. 

97904: theta
97909: theta
97916: phi
97919: phi

We need to look at the column definitions in cobra_target.

There are pfi_nominal_x_mm and pfi_target_x_mm. Originally, pfi_target contained the final target postion for the convergence run, and pfi_nominal was supposed to contain the target position for each iteration. The latter took a while to implement, however, it looks like pfi_target was changed to contain the target for each iteration, while pfi_nominal contains the centre position of the patrol region. 

The cobra identification was using pfi_nominal, rather than pfi_target, which results in misidentifications in some cases. Changing either the table or the identification code will get rid of ~2/3 of the misidentifications in the regular convergence cases. 

Even with this problem, mis-identification are rare. Of 151 runs analyzed, 36 incidents were flagged.  

The code has been tested and merged to master, so I close this ticket.