<ul><li>Integer-order calculus is limited in capturing long-range dependencies and memory effects in real-world processes, leading to challenges in system identification and optimal control.</li><li>Fractional calculus addresses these issues using fractional-order integrals and derivatives, particularly for fractional-order dynamical systems, which lack standard control methodologies.</li><li>A new framework has been developed that theoretically derives optimal control for fractional-order linear time-invariant (FOLTI) systems using linear quadratic regulator (LQR) and incorporates deep learning for data-driven optimal control.</li><li>The approach introduces an innovative system identification method, a data-driven learning framework called Fractional-Order Learning for Optimal Control (FOLOC), and provides a theoretical sample complexity analysis for accurate optimal control in real-world problems.</li></ul>

End-to-End Learning Framework for Solving Non-Markovian Optimal Control

Discover more