Adiabatic Spectral Submanifolds in Data-Driven Modeling and Control

This abstract has open access
Summary
We extend the current theory for spectral submanifolds (SSMs) to ones covering slowly varying dynamical systems. Restricting the dynamics onto these SSMs provides one with a mathematically rigorous model reduction technique. We also observe a justified approximation of adiabtic SSMs can be obtained from data. We demonstrate these capabilities for controlling a high- dimensional nonlinear finite element model describing a soft trunk robot. We formulate a novel model predictive control scheme using adiabatic SSMs constructed from data. We find our method outperforms the state-of-the-art techniques for trajectory tracking control tasks.
Abstract ID :
73
Doctoral Student
,
ETH Zurich
Stanford University
Stanford University
10 visits