Skip to content
marcinpohl edited this page Sep 12, 2010 · 5 revisions

TODO:

OBJECTIVE FUNCTION

  • figure out the objective function, is norm(CAM0-CAM2) enough or do we have to go to norm(AFR-AFRwb)? Is there a point to that without a full injector model?

ROBUST FIT

  • do we want OBJFUN to take low-robust values into account?
  • if we do, OBJFUN has to be changed from norm(Y-Yhat) to RMS by dividing it by sqrt(Nrobust)

ZONES and BOUNDARIES

  • putting in equations for zone borders as constraints
  • rearranging zone belonging for low-sample count zones/zones that make sense together (MAP>90, RPM>3750)
  • introduce flexible boundaries to get ready for optimizing them
  • can we just ‘optimize’ boundaries by looking at sample count?
  • adjust zones so they can be adjusted like the E38 tables (one-d table for rpm, but two-d for map)

MATHY:

  • fuzzy clustering aiding zone assignment?

GENERAL:

  • code cleanup
  • moving away from globals to param passing
  • abstraction of platform independent pieces (function pointers?)
Clone this wiki locally