This repository contains code that demonstrates the use of a variety of machine learning strategies for low temperature plasma systems, as presented in the paper on the Foundations of machine learning for low-temperature plasmas: methods and case studies.

This code trains and implements a stochastic physics-informed neural ordinary differential equation (SPINODE) framework on a directed colloidal self-assembly test case, as presented in the paper on Stochastic physics-informed neural ordinary differential equations.

This repository contains codes for several of Diogo Rodrigues’ works with our group, including:

1. code that simulates a multivariable control strategy for a reactor system. The strategy deals explicitly with the existence of incomplete models using feedback linearization and continuous-time derivative estimation, as presented in the paper on Multivariable control based on incomplete models via feedback linearization and continuous-time derivative estimation;

2. code that simulates an efficient global solution method for a reaction system and rocket. The method approximates an optimal control problem as a set of polynomial optimization problems, which can be solved using sum-of-squares polynomials, as presented in the paper on Efficient global solutions to single-input optimal control problems via approximation by sum-of-squares polynomials;

3. code that simulates solution methods for two approximate formulations of the Bayesian optimal experiment design (OED) problem. The methods reformulate the Bayesian OED problem as an optimal control problem and integrate methods described in 2. to find tractable global solutions, as presented in the paper on Tractable global solutions to chance-constrained Bayesian optimal experiment design for arbitrary prior and noise distributions; and

4. code that simulates an optimal control approach for a cold plasma system. The approach uses data-driven modifier adaptation to deal with structural plant-model mismatch and integrates the strategy in 1. for online path constraint tracking, as presented in the paper on Data-driven adaptive optimal control under model uncertainty: An application to cold atmospheric plasmas.

This code trains and implements a characterization framework based on deep learning for characterizing structural states of colloidal self-assembly systems, as presented in the paper on Deep learning for characterizing the self-assembly of three-dimensional colloidal systems.

This code obtains closed-loop performance guarantees for automated controller tuning, which can be formulated as a black-box optimization problem under uncertainty, as presented in the paper on Data-Driven Scenario Optimization for Automated Controller Tuning With Probabilistic Performance Guarantees.

This code trains and implements a reinforcement learning framework for control of the thermal effects of an atmospheric pressure plasma jet, as presented in the paper on Sim-to-real transfer reinforcement learning for control of thermal effects of an atmospheric pressure plasma jet.

This code trains and implements machine learning models for real-time diagnostics of cold atmospheric plasma sources, as presented in the paper on Machine Learning for Real-Time Diagnostics of Cold Atmospheric Plasma Sources.