Uploaded June 2021 | Updated November 2026, 6 hours ago
This topic demonstrates how you can use observed controller and trajectory data from a control affine dynamical system to make predictions about a new feedback controller that you want to analyze!
I had a lot of fun doing this video, and I took a prompt from The Tim Traveler’s YouTube channel (International Staycation Challenge) to start the video off, and I still manage to tie it back into the problem at hand.
This theory might hold the record for how quickly it was developed in my group. We went from a vague desire to do DMD for control affine systems, and in two weeks, we developed two new operators (control Liouville Operators and a new Multiplication operator), a new kind of occupation kernel (control occupation kernels), and then we devised the theory that gave us a DMD algorithm that is “off policy.” And that ultimate manuscript that we wrote up in two weeks was accepted to MNTS 2020, but isn’t going to appear until this August (2021… COVID…)
arXiv link here: arxiv.org/abs/2101.02620
I had a lot of fun working on this project, publishing the results, and making this video. Come give it a watch! I am eager for feedback!
(One typo in the video I found. The feedback controller should be mapping to R^m, not R^n.)
The Tim Traveler's Channel youtube.com/c/TheTimTraveller
#InternationalStaycationChallenge
0:00 Intro
0:39 Staycations in Florida
3:32 SIR Models as Nonlinear Control Affine Systems
5:03 Control Affine Dynamical Systems and their Operators
8:04 The Data Driven Perspective
9:32 Complications with DMD and Control Affine Systems
10:38 The Two Operators
13:20 Vector Valued Reproducing Kernel Hilbert Spaces
15:33 Control Occupation Kernels
18:01 The Finite Rank Representation
27:19 Recap!
29:42 Closing Remarks
Music:
Come 2gether by Ooyy
Midnight Radio by Cacti
I Know Now by Cacti
Wilder by Gamma Skies
This was Fun by Gavin Look
Wrong by Dan Henig
Guardians + Tek by Craig Hardgrove
This topic demonstrates how you can use observed controller and trajectory data from a control affine dynamical system to make predictions about a new feedback controller that you want to analyze!
I had a lot of fun doing this video, and I took a prompt from The Tim Traveler’s YouTube channel (International Staycation Challenge) to start the video off, and I still manage to tie it back into the problem at hand.
This theory might hold the record for how quickly it was developed in my group. We went from a vague desire to do DMD for control affine systems, and in two weeks, we developed two new operators (control Liouville Operators and a new Multiplication operator), a new kind of occupation kernel (control occupation kernels), and then we devised the theory that gave us a DMD algorithm that is “off policy.” And that ultimate manuscript that we wrote up in two weeks was accepted to MNTS 2020, but isn’t going to appear until this August (2021… COVID…)
arXiv link here: arxiv.org/abs/2101.02620
I had a lot of fun working on this project, publishing the results, and making this video. Come give it a watch! I am eager for feedback!
(One typo in the video I found. The feedback controller should be mapping to R^m, not R^n.)
The Tim Traveler's Channel youtube.com/c/TheTimTraveller
#InternationalStaycationChallenge
0:00 Intro
0:39 Staycations in Florida
3:32 SIR Models as Nonlinear Control Affine Systems
5:03 Control Affine Dynamical Systems and their Operators
8:04 The Data Driven Perspective
9:32 Complications with DMD and Control Affine Systems
10:38 The Two Operators
13:20 Vector Valued Reproducing Kernel Hilbert Spaces
15:33 Control Occupation Kernels
18:01 The Finite Rank Representation
27:19 Recap!
29:42 Closing Remarks
Music:
Come 2gether by Ooyy
Midnight Radio by Cacti
I Know Now by Cacti
Wilder by Gamma Skies
This was Fun by Gavin Look
Wrong by Dan Henig
Guardians + Tek by Craig Hardgrove










