System Identification: Koopman with Control

preview_player
Показать описание
This lecture provides an overview of the use of modern Koopman spectral theory for nonlinear control. In particular, we develop control in a coordinate system defined by eigenfunctions of the Koopman operator.

Data-driven discovery of {K}oopman eigenfunctions for control
E. Kaiser, J. N. Kutz, and S. L. Brunton, arxiv 2017.

This video was produced at the University of Washington
  1. Steve Brunton
  2. Control theory
  3. Machine learning
  4. Data science
  5. Dynamical systems
  6. Machine learning control
Рекомендации по теме
System Identification: Koopman 0:10:04

System Identification: Koopman with Control

System Identification: Full-State 0:10:19

System Identification: Full-State Models with Control

Koopman Operator Based 0:10:25

Koopman Operator Based Data Driven Identification of Tethered Subsatellite Deployment Dynamics

Nonlinear System Identification 0:02:29

Nonlinear System Identification of Soft Robot Dynamics Using Koopman Operator Theory - ICRA 2019

Koopman Operator-Based Data-Driven 0:10:25

Koopman Operator-Based Data-Driven Identification of Tethered Subsatellite Deployment Dynamics

An introduction to 0:14:27

An introduction to the koopman operator ds4ds 8 01

End-to-end Learning of 0:58:27

End-to-end Learning of Koopman Models for Control

Data-Driven Control: Linear 0:20:16

Data-Driven Control: Linear System Identification

Koopman-operator-based Attitude Dynamics 0:00:51

Koopman-operator-based Attitude Dynamics and Control on SO(3)

System Identification: Dynamic 0:11:38

System Identification: Dynamic Mode Decomposition with Control

System Identification: Sparse 0:08:25

System Identification: Sparse Nonlinear Models with Control

A2IR2 Seminar 2 2:10:49

A2IR2 Seminar 2 - Modal Description of Nonlinear Dynamical Systems with Koopman Operator Theory

Local Koopman Operators 0:01:23

Local Koopman Operators for Data-Driven Control of Robotic Systems

Data Driven Control 0:56:23

Data Driven Control using Dynamic Mode Decomposition and its extensions using Koopman operators

System Identification: Regression 0:05:58

System Identification: Regression Models

Koopman Spectral Analysis 0:15:26

Koopman Spectral Analysis (Control)

Autoware CoE Seminar 0:57:59

Autoware CoE Seminar - Koopman Operator for Modeling & Control of Autonomous Vehicles

System Identification: DMD 0:08:09

System Identification: DMD Control Example

Methods for System 0:44:18

Methods for System Identification (Prof. Steve L. Brunton)

Redundant Space Robot 0:00:24

Redundant Space Robot System Identification

Modeling and Control 0:02:13

Modeling and Control of Soft Robots Using the Koopman Operator and Model Predictive Control

Two seminars on 2:05:16

Two seminars on Data Science for Koopman Methods and Vice Versa by Alexandre Mauroy & Felix Diet...

Derivative-Based Koopman Operators 0:05:49

Derivative-Based Koopman Operators for Real-Time Control of Robotic Systems

The Koopman Generator 0:15:11

The Koopman Generator (DS4DS 8.13)

Комментарии
Автор

Really interesting and useful stuff. At 6:18 it even feels a little philosophical so to speak.

trejohnson
Автор

absolutely. The o/p jitter occurs when augmented control function shifts exponentially i.e( dx/dt to (d^2x/dt^2) or at points of inversion in non-linear component. Someone in linked-in spoke about reducing these jitters through a "stepped" approach by adding an extra (du/dt - where u represents the max delta feasible in o/p) balancing component @ both sides of equation.

ankeshcdac
Автор

I really enjoy your videos. A food for thought - How does this tie in with the input-state feedback linearization? The two conceps are somewhat related by finding a state-dependent transformation of our input to obtain linear dynamics of the system. Thanks.

matejrajchl
Автор

I don't understand the difference between d \phi(x)/dt and \phi dot.

lukistrela
join shbcf.ru