From b23db588d510995fc75e51034a6cba0860286e95 Mon Sep 17 00:00:00 2001 From: Dane Sabo Date: Fri, 8 Aug 2025 13:56:35 -0400 Subject: [PATCH] update --- .../Editing/thesis_ideas_notes.md | 3 ++ .../Fleeting Notes/Journal/2025_08_07.md | 11 ++++ Zettelkasten/Permanent Notes/thesis-ideas.md | 53 ++++++++++++------- 3 files changed, 47 insertions(+), 20 deletions(-) create mode 100644 Zettelkasten/Fleeting Notes/Editing/thesis_ideas_notes.md create mode 100644 Zettelkasten/Fleeting Notes/Journal/2025_08_07.md diff --git a/Zettelkasten/Fleeting Notes/Editing/thesis_ideas_notes.md b/Zettelkasten/Fleeting Notes/Editing/thesis_ideas_notes.md new file mode 100644 index 000000000..9f86c6aa9 --- /dev/null +++ b/Zettelkasten/Fleeting Notes/Editing/thesis_ideas_notes.md @@ -0,0 +1,3 @@ +# Notes on [[thesis-ideas-2025-07-30]] + + diff --git a/Zettelkasten/Fleeting Notes/Journal/2025_08_07.md b/Zettelkasten/Fleeting Notes/Journal/2025_08_07.md new file mode 100644 index 000000000..ccd6f9298 --- /dev/null +++ b/Zettelkasten/Fleeting Notes/Journal/2025_08_07.md @@ -0,0 +1,11 @@ +--- +--- + +# 2025-08-06 +Today I have not gotten too much done, but I did get here +early today. Today, I'll sell the truck and work on Sam's +car. I saw Patrick today too, which was nice. Robert's stuff +is all gone. + +My main focus for today is to finish the thesis ideas and +tune up the writing. That'll be enough. diff --git a/Zettelkasten/Permanent Notes/thesis-ideas.md b/Zettelkasten/Permanent Notes/thesis-ideas.md index a2c1a41f8..259fae878 100644 --- a/Zettelkasten/Permanent Notes/thesis-ideas.md +++ b/Zettelkasten/Permanent Notes/thesis-ideas.md @@ -15,32 +15,45 @@ ___________________________________________________________ ## **Integrating Shielding into Nuclear Power Control** ### Goal: -The goal of this research is to develop machine learning -control algorithms for nuclear power applications with -strict safety guarantees. +The goal of this research is develop machine learning +enabled control algorithims for nuclear power applications +that incoporate shielding: a formal guarantee of adherence +to system specifications without augmenting the machine +learning process. ### Outcomes: -If this research is successful, I will have accomplished the -following: +For this research to be successful, I will accomplish the +following: -1. Develop controller shielding methods for nuclear power - contexts +1. Identify key controllers in a nuclear power context with + the most benefit from using an ML-based controller -2. Provide concrete safety guarantees for autonomous control - of a nuclear asset. +2. Translate regulatory and system level requirements into a + formal specification to synthesize a controller 'shield'. + This shield monitors the ML controller and intervenes + whenever a requirement is predicted to be violated. -3. ??? +3. Evaluate performance of the ML controller with attached + shield, while assessing the amount of shield useage for + different operating scenarios (power up, shut down, regular + load following) ### Impact: -Machine learning based systems have been shown to be more -efficient than typical PID based controllers, and are able to -learn more complex objective functions than a typical controller -can. The problem with these controllers though is that they are -often unexplainable. This is not acceptable for high assurance -applications, where slight perturbations on inputs can yield -wildly different outputs. Shielding can solve this problem, -helping ensure safety of ML based controllers while not limiting -their development or construction. + +Machine learning controllers can outperform PID and +rule-based controllers by adapting to nonlinear dynamics, +optimizing over multi-objective cost functions, and changing +plant conditions. But, these ML controllers are often +*unexplainable*, meaning that their global behavior is not +easily understood.This unexplainability prevents ML based +controllers from being used in high-assurance usecases such +as nuclear power. Shielding can address this issue, by +providing a formal runtime assurance, allieviating the +burden of explainability away from the machine learning +algorithm. This work would further bring regulatory +requiremnts into the formal design of control systems and +help bridge the gap between high assurance systems and the +start of the art in control. ### Relevant Papers [[safe-reinforcement-learning-via-shielding]] @@ -118,7 +131,7 @@ ___________________________________________________________ ### Goals: If this research is successful, we will be able to generate autonomous controller shields that provably adhere to specifications -written with temporal logic automatically. +written with temporal logic. ### Outcomes: - Create an intermediary shield that mediates signals between an