Arka Pal

PROJECT DESCRIPTION

Neural differential controller on PAM driven muscle

May 2023-July 2023   Friedrich Alexander University, Germany


During the summer of 2023, I had the opportunity to perform a research internship at the Friedrich Alexander University of Erlangen Nuremberg(Germany) as a DAAD-WISE intern.

     

In this time I have worked on a new data-driven controller ANODEC. ANODEC or Automatic Neural-Ordinary Differential Equation Control, is a a data-efficient automatic design of output feedback controllers for finite-time reference tracking in systems with unknown nonlinear dynamics. First I have gatherd few seconds of data to develop the model of the system as a NDE(Neural Differential Equation) model. After some tuning and playing with the parameters, when the dynamics was captured, I used the chain_control package developed by Simon Bachhuber to get a controller based on another NDE. After that, we have compared the performance of the controller on some known and unknown trajectories with a conventional PID controller. The real-time integration of chain_control was done using ROS2 framework and I really every bit of this process!