1. there has been lots of good recent work on closely related topics like parallel distributed MPC, hierarchical MPC, and parallel surrogate-enabled multi-objective optimization. One of the best papers I've found, from the standpoint of bringing many of these ideas together is this one from 2024: "Cooperative nonlinear distributed model predictive control with dissimilar control horizons" https://arxiv.org/html/2410.10428v1
2. For some useful kinda-sorta-related mindstretching, consider...
a. Michael Levin's work on more general notions of "intelligence", nested agency, etc.
b. Tim Lenton's "survival of systems"
c. Lynn Margulis' work on "endosymbiosis"
d. Donella Meadows, Thinking in Systems
e. Max Bennett's "five breakthroughs" model of the evolution of human intelligence (#3 is "simulating in early mammals", #4 is "mentalizing (i.e. model-based decision-making using models that include the internal mental state of the decision-maker and other agents present) in early primates")
Genuinely valuable framing on PID's underappreciation. The cross-domain applicability to ML optimization and biologcal systems is something I hadn't deeply considered until seeing it articulated this way. One thing I've struggled with in practice is knowing when model misspecification tips from "negative feedback compensates" to catastrophic failure mode, thats always felt more art than science.
I'm not a "control guy" myself, but Control Theory is embedded, even if implicitly, in the work I've read (and still read) on electricity market design. When discussing spot pricing, "Spot Pricing of Electricity" by Schweppe, Caramanis, Tabors, and Bohn (1988) is inescapable. A lot has been done since then, and the language of "optimization meets microecon" has certainly taken over the introductory books on the topic. But in Schweppe et al.'s "Acknowledgments" section, the authors highlight that the work on the book is the culmination of the "informal Homeostatic Control workshops" that took place at MIT in 1977-78; a fact that I think is often underappreciated.
"Optimal Spot Pricing: practice and theory" (Caramanis, Bohn, Schweppe, 1982)[DOI: 10.1109/TPAS.1982.317507] explains nearly 10 years before _the_ book how homeostatic control came into play when developing the concept of "spot pricing".
Another piece of work that deserves attention when it comes to Homeostatic Control applied to "energy management" is "Homestatic Control: the utility/customer marketplace for electric power", by Schweppe, Tabors, and Kirtley (1981). In 1981, the authors were already talking about the profound changes "uncertainty" was causing to power-systems operations. Search "In summary, the utility [...]" to read a 40-year-old paragraph that remains relevant today. Finally, on the same topic, Breslau's "Redistributing Agency: The Control Roots of Spot Pricing of Electricity" [https://doi.org/10.1215/00182702-8718023] might be a good resource for tracking this history. I think all of this can give an "energy flavor" to the topic of stability as "theory of homeostasis".
Taking control theory (vaguely) as "the theory of using feedback/newly available data to make sequential decisions" (I apologize if this is too vague. In fact, it probably is🙃), I believe that "Power Systems" as a field provides fertile ground for rich discussions. For instance, y'all can discuss the implications of adopting a pricing policy derived from a static, single-shot optimization problem when, in reality, the system updates frequently, incorporating new information to make decisions sequentially under uncertainty over a given (finite) operational horizon. I myself took a shot at this problem here: https://ieeexplore.ieee.org/document/11359506. (Self-citing, I know, I know...). But my paper belongs to a large body of work that explores the "art" of making decisions sequentially in the face of substantial uncertainty, and the cost of neglecting this "art".
I find Control Theory extremely cool and enjoyed a couple of classes I've taken on "stochastic control". I'm excited to see what this course is gonna shape up to be.
To end, I second the comment that mentioned "Respect the Unstable" by Gunter Stein.
This looks like a banger. Exciting!
I like "Respect the Unstable" by Gunter Stein
Hooboy, I am really excited to follow along with this course!
I have a soft spot for this essay by Alistair McFarlane: https://link.springer.com/chapter/10.1007/978-1-4612-0313-1_1
1. there has been lots of good recent work on closely related topics like parallel distributed MPC, hierarchical MPC, and parallel surrogate-enabled multi-objective optimization. One of the best papers I've found, from the standpoint of bringing many of these ideas together is this one from 2024: "Cooperative nonlinear distributed model predictive control with dissimilar control horizons" https://arxiv.org/html/2410.10428v1
2. For some useful kinda-sorta-related mindstretching, consider...
a. Michael Levin's work on more general notions of "intelligence", nested agency, etc.
b. Tim Lenton's "survival of systems"
c. Lynn Margulis' work on "endosymbiosis"
d. Donella Meadows, Thinking in Systems
e. Max Bennett's "five breakthroughs" model of the evolution of human intelligence (#3 is "simulating in early mammals", #4 is "mentalizing (i.e. model-based decision-making using models that include the internal mental state of the decision-maker and other agents present) in early primates")
Best regards,
Steve Coy
CEO/CTO, TimeLike Systems
www.timelike.systems
Genuinely valuable framing on PID's underappreciation. The cross-domain applicability to ML optimization and biologcal systems is something I hadn't deeply considered until seeing it articulated this way. One thing I've struggled with in practice is knowing when model misspecification tips from "negative feedback compensates" to catastrophic failure mode, thats always felt more art than science.
I'm not a "control guy" myself, but Control Theory is embedded, even if implicitly, in the work I've read (and still read) on electricity market design. When discussing spot pricing, "Spot Pricing of Electricity" by Schweppe, Caramanis, Tabors, and Bohn (1988) is inescapable. A lot has been done since then, and the language of "optimization meets microecon" has certainly taken over the introductory books on the topic. But in Schweppe et al.'s "Acknowledgments" section, the authors highlight that the work on the book is the culmination of the "informal Homeostatic Control workshops" that took place at MIT in 1977-78; a fact that I think is often underappreciated.
"Optimal Spot Pricing: practice and theory" (Caramanis, Bohn, Schweppe, 1982)[DOI: 10.1109/TPAS.1982.317507] explains nearly 10 years before _the_ book how homeostatic control came into play when developing the concept of "spot pricing".
Another piece of work that deserves attention when it comes to Homeostatic Control applied to "energy management" is "Homestatic Control: the utility/customer marketplace for electric power", by Schweppe, Tabors, and Kirtley (1981). In 1981, the authors were already talking about the profound changes "uncertainty" was causing to power-systems operations. Search "In summary, the utility [...]" to read a 40-year-old paragraph that remains relevant today. Finally, on the same topic, Breslau's "Redistributing Agency: The Control Roots of Spot Pricing of Electricity" [https://doi.org/10.1215/00182702-8718023] might be a good resource for tracking this history. I think all of this can give an "energy flavor" to the topic of stability as "theory of homeostasis".
Taking control theory (vaguely) as "the theory of using feedback/newly available data to make sequential decisions" (I apologize if this is too vague. In fact, it probably is🙃), I believe that "Power Systems" as a field provides fertile ground for rich discussions. For instance, y'all can discuss the implications of adopting a pricing policy derived from a static, single-shot optimization problem when, in reality, the system updates frequently, incorporating new information to make decisions sequentially under uncertainty over a given (finite) operational horizon. I myself took a shot at this problem here: https://ieeexplore.ieee.org/document/11359506. (Self-citing, I know, I know...). But my paper belongs to a large body of work that explores the "art" of making decisions sequentially in the face of substantial uncertainty, and the cost of neglecting this "art".
I find Control Theory extremely cool and enjoyed a couple of classes I've taken on "stochastic control". I'm excited to see what this course is gonna shape up to be.
To end, I second the comment that mentioned "Respect the Unstable" by Gunter Stein.