In this video, I show the effects of tuning the PD algorithm that this ball balancing robot uses. Part 1: • Ball Balancing Robot pt.1 Instructable: www.instructables.com/Ball-Ba...
A logical development would be to now have balls of variable mass (or other physical properties?) It would be nice to see plots of the error with time. I think you could make the algorithm tune itself. Fuzzy control might also be possible. Great work!
Excellent explanations and very well done! Best I've seen on PID systems! Great for my students. Keep up the excellent work; you've earned my subscription!
I agree. I might have learned more; understood more, if they had this back when I needed it some 50 years ago. Also, I can't believe how few subscribers you have. Well now I'm *SUBSCRIBED!*
Wouldnt the integral value would help him being constantly learning from his incorrect movements and tuning himself all the time ?? Im not an engineer but i did a electromechanic formation some 12 years ago i did not worked with pid very much since school and even then it was simplified
A long time ago (in the BasicX days) I tried to develop a self-balancing robot. It had a single DOF: simply, just a T-square ruler placed upside-down on the floor with a single servo on top operating a single "arm". Sort of like a human keeping her balance standing with her legs apart and being only able to swing one arm. The robot had a single sensor - reporting the level angle. You may think of it as an upside-down Segway problem. I failed because the equations of angular momentum and acceleration got too hairy and BasicX was too slow. This problem is much simpler than the one that you have solved. Can you think (in your spare time) of a simple solution?
Thank you for doing this! I've always wanted to understand PID control and inverse kinematics. This second video explains PID control and your approach with the software very well. Is there any chance a third video is coming to explain the inverse kinematics?
Unfortunately no. The brief overview is that I used vector calculus to derive the equations and it took a really long time lol! The equations essentially allow you to input a vector and have the platform point in the direction of that vector. Knowing the distances between the servos and how long each linkage is, you can break the entire system into a bunch of 3D lines and can then solve for the positons which each servo must face in order to get the platform to point in the direction of the aformetntioned vector.