What is it about?

The research explores advancements in optimal control techniques of systems governed by differential equations, specifically comparing established methods like Direct-Adjoint Looping with newer approaches such as Physics-Informed Neural Networks and Differentiable Programming.

Featured Image

Why is it important?

This research holds significance as it navigates the evolving landscape of optimal control methods, shedding light on the effectiveness of emerging techniques like Differentiable Programming in comparison to established approaches. By providing practical insights and benchmarks, the study equips practitioners with a guide to leverage the strengths of traditional and modern methods, fostering a more informed and effective approach to optimal control.

Perspectives

Coming from an applied mathematics background, I am used to traditional solvers with measurable accuracy for my problems. As I joined the AI world, I realised how data-driven approaches were taking over. Not satisfied with how uncertain these newer models could be, I decided to conduct a comparison; hence this work.

Roussel Nzoyem
University of Bristol

Read the Original

This page is a summary of: A Comparison of Mesh-Free Differentiable Programming and Data-Driven Strategies for Optimal Control under PDE Constraints, November 2023, ACM (Association for Computing Machinery),
DOI: 10.1145/3624062.3626078.
You can read the full text:

Read

Contributors

The following have contributed to this page