Oops, I forgot to make a feature branch and a second commit was included as well when I committed to my fork (d306d19). I think you should consider this one as well, so I will leave this PR as it is.
This second change migrates the integer conversation later in the calculation pipeline into a single cast, reducing the error produced by the previous two integer conversions (which were both before and after inputToReal).
My perception of the resulting movement is that it is quite a bit smoother. In (almost) all cases, it improves the accuracy of the movement as well, sometimes significantly.
Thank you for considering these changes. I am really appreciative of the work you are doing here and I will be happy if I can help!
I believe this to be a good general improvement, certainly for the constraints as low -Y values. Accuracy is improved, but some new things to look at have been uncovered.
I believe that this change (ace1702) will address issue #5 I've tested this by analysis and in use with the arm. Results are quite good!
Movement consistent with: https://s3.amazonaws.com/corboy-images/result2B.png https://s3.amazonaws.com/corboy-images/result5B.png
Oops, I forgot to make a feature branch and a second commit was included as well when I committed to my fork (d306d19). I think you should consider this one as well, so I will leave this PR as it is.
This second change migrates the integer conversation later in the calculation pipeline into a single cast, reducing the error produced by the previous two integer conversions (which were both before and after inputToReal).
My perception of the resulting movement is that it is quite a bit smoother. In (almost) all cases, it improves the accuracy of the movement as well, sometimes significantly.
Thank you for considering these changes. I am really appreciative of the work you are doing here and I will be happy if I can help!