This paper discusses optimal batch-to-batch (B2B) control problems and presents a gradient descent method solution for unknown linear batch process systems. Using historical process data, we design a model-free method for B2B optimization that eliminates the need for model information about the system. By using quadratic programming (QP) to formulate the optimal controller design, we first present the optimal iterative learning control (ILC) results. Next, using the gradient descent method, we replace the uncertain term with the actual measurements and develop a new ILC approach based on convex hull representations of uncertain realizations. As compared to the norm-optimal ILC, our proposed ILC can guarantee superior performance with reasonably selected parameters. Finally, we demonstrate our design with an illustrative numerical example.
|Title of host publication||Proceeding of 2023 3rd International Symposium on Instrumentation, Control, Artificial Intelligence, and Robotics, ICA-SYMP 2023|
|Publisher||Institute of Electrical and Electronics Engineers Inc.|
|Number of pages||4|
|Publication status||Published - 2023|
|Event||3rd International Symposium on Instrumentation, Control, Artificial Intelligence, and Robotics, ICA-SYMP 2023 - Bangkok, Thailand|
Duration: 18 Jan 2023 → 20 Jan 2023
|Conference||3rd International Symposium on Instrumentation, Control, Artificial Intelligence, and Robotics, ICA-SYMP 2023|
|Period||18/01/23 → 20/01/23|
Bibliographical noteFunding Information:
This work was supported in part by the Hong Kong Research Grant Council under grant 16208520; in part by the Foshan-HKUST Project under grant FSUST19-FYTRI01; in part by the Guangdong scientific and technological project (202002030323 and 2014B050505002); and in part by the Hong Kong RGC Postdoctoral Fellowship Scheme (PDFS2122-6S06).
© 2023 IEEE.
- data-driven control
- iterative learning control (ILC)
- Process control