A Fourier Series Approach to Modeling Nonlinear Damping Using L1 Regularized Regression

Abstract

MS Thesis Defense:

An accurate model of friction is an essential tool in developing high-performance control techniques in robotic systems. Current approaches typically focus on best fit models either based on first principles or general basis functions. However, such models can quickly become overly complex. It is also unclear to what degree such models generalize in the face of measurement uncertainty. To address these issues, this paper proposes a method to algorithmically construct a generalized nonlinear model of frictional force using Fourier series elements as model features. Our findings could facilitate more effective compensation of dynamics as a robotic system encounters unfamiliar environments.

Please join meeting via Zoom:

https://tufts.zoom.us/my/yanxi.piao

See colloquia email for password.

Dial-in not an option for this event.