Tamam Abou Al Fadel

I have not had a chance to try it on microAutoBox yet. But to me its to be expected that any functions with timers and printing out will have issues with certain dSPACE models. I would first go through the ACADO generated source files looking for those structs and check what flags are used to activate/deactivate them. Also include the flags in my previous comment in this post regarding dSPACE hardware. Only way is to track the error back to the source files and check where its going wrong. Regards,...

Hi Alberto, In this rather long forum post, i described to other members how to implement the changes needed for autogenereating S-Functions with external constraint support: https://sourceforge.net/p/acado/discussion/general/thread/aafddff061/ Regards, Tamam

Hi Alberto, I think the issue in the other forum topic you linked is a different problem. I think the problem in your code is just a system formulation problem. I'll give you a few tips that you can try, they should fix the problem and optimize your formulation. To me it seems like you are using Matlab. 1- you use expressions likecos(delta )and sin(delta)and so on in multiple places in your system. Everytime this comes up, it's calculated again. Instead, define them as seperate intermediateStates,...

Hi Alberto, If you use the automatic S-Function generation, you will get the CPU time. However, by default you cannot use variable constraints. For a work around, search the forum for a post where I describe the how to adjust the code to allow automatic S-Function generation and variable constraints simultanuously. Regards, Tamam

Hi, Your reference function can have constants, states, controls and online data. Its all time-varying. It's best to use acado.IntermediateState(your expression here); Example: j = acado.IntermediateState(state^2 + OnlineData1); h = [j]; to minimize the expression j. The OnlineData is inputted to the acado in a time-varying matrix for the complete horizon. Regards, Tamam

Hi Lee, Sorry, i think i misread your original post. Do you mean that your discrete integration time step dt for the prediction horizon is different than the overall system sampling rate which is calling the solver? If that is the case, then the performance will depend on how big the difference is. It is just equivalent to a bad initialization. It will possibly make your solver slower. Whether you need to do something about it, depends on your system. Keep in mind, that your system is subjected to...

Hi Nadia, How did you generate the S-Function? did you use the automatic S-Function generation or the manual generation provided as an example from Rien? Regards, Tamam

Hi, Well providing the previous solution as initialization for the next solver with different time step is just equivalent to a bad initialization. You could do that, then call the new solver AGAIN with that solution without shifting, thus doing 2 sqp iterations. That would be a quick and easy thing to try. Regards, Tamam