Topic : Autonomous control for Radio Frequency control system

  

 Scope  

                        Radio frequency (RF) is one of the systems in a synchrotron radiation facility of the Siam Photon Source (SPS). This dynamical system is highly complex due to its high-dimensional parameter space and dynamic nature. Instability of the RF system can easily affect the operation of the machine if improper control algorithm is not carefully designed and implemented. In addition, to achieve desired controlled behavior of the machine, accelerator parameters tuning for optimal performance can be challenging. These characteristics makes it potentially suited for a reinforcement learning (RL) approach that learns to make sequences of optimal decisions under parameter uncertainty in this dynamical system. Recent developments in RL have shown promising result in control system applications. The PhD candidates will be working by designing an AI-based agent using Deep RL approach (RL coupled with deep neural networks for continuous state and action-space representation) that can teach itself how to learn optimal control policies and implementing controller for the RF system. The goal of the controller implementation is to reduce the machine tuning time without any input or supervision from a human and ultimately achieve a near-autonomous control scheme.
                        Tools to be used: High-performance computer, FPGA controller boards, microcontrollers, single-board computers (for direct implementation of the control, signal processing, and machine learning algorithms).
                        Skill required: Programming/coding skill. Knowledge of control system engineering/machine learning is preferable.

  

 Supervisor  

                        Dr. Roengrut Rujanakraikarn
                        and SLRI researchers