Research semester project
Online Adaptive MPC on Tendon-Driven Soft Arm
A research project about online data-driven adaptive MPC done at the CREATE lab.
The goal was to control with an MPC a tendon-driven soft arm. 
But only model the arm in air and find a way for it to adapt by itself when entering another medium such as water.
To achieve this, our approach was to have add additional parameters to the model we optimize by minimizing the models past prediction with its actual recorded position.
The complex modelling of such arm made the simulation sensitive and sometimes unstables but our implementation worked even with disturbances, with the arm having a stable position (pointing downwards) or no stable position (pointing upwards).
Different trajectories have been tried, you can find some below.
Unfortunately the hardware was not working properly and we could not try the adaptive MPC in real conditions
Don’t hesitate to check out the code and report!
