B-spline based repetitive neurocontroller

A non-local learning rule is employed in a repetitive neurocontroller based on B-spline network.

A numerical experiment described in http://dx.doi.org/10.1109/TIE.2007.909064 [*] is reproduced here. In [*] the forgetting mechanism is employed to robustify the control scheme. By contract, in http://dx.doi.org/10.1109/IECON.2013.6700120 [**] weight constraints are used instead of forgetting and that turns out to robustify the controller. Hence, the same idea has been tested also in the B-spline based repetitive neurocontroller proposed in [*]. To be clear, I haven't invented the controller introduced in this model. I've just modified the robustification mechanism used in [*]. You can play here with both mechanisms and decide for yourself which one of them is more suitable for your application. You can even combine them. It should be noted that [**] uses a global update rule and not necessarily the same robustification mechanisms are equally effective in both controllers. For more information please see m-files and our conference paper: M. Malkowski, B. Ufnalski and L. M. Grzesiak, B-spline based repetitive controller revisited: error shift, higher-order polynomials and smooth pass-to-pass transition, ICSTCC 2015, http://ufnalski.edu.pl/proceedings/icstcc2015/ .