Introduction to Physics-Generated AI-Driven Filter and Control Scheme of Nonlinear Stochastic Systems of Man-Made Machines
Bor-Sen Chen · 2026
This chapter discusses the limitations of big data-driven AI algorithms for nonlinear filter and control design in complex man-made machines. Then, H∞ physics-generated AI-driven filter and control schemes are introduced to overcome the limitations of big data-driven AI schemes. These schemes integrate system dynamic models, observers and estimation error models with worst-case external disturbance, measurement noise, attack signals and coupling to generate training data, which are used to train the weighting parameters of deep neural networks (DNNs) embedded in physics-generated AI to solve the Hamilton–Jacobi–Isaacs equation (HJIE) via the Adam learning algorithm. The aim is to achieve robust H∞ filter and control design strategies for nonlinear stochastic systems with potential applications in complex man-made machines. The chapter also provides a brief overview of the scope of each chapter in this book.