当前位置: 首页  学术动态

数学学科现代分析及其应用研究所学术报告(Amol Yerudkar 意大利萨尼奥大学)

发布者:戴 情   发布时间:2021-07-19  浏览次数:232

报告题目1Reinforcement Learning for Control of Probabilistic Boolean Control Networks

 

报告时间:2021720日(周二)下午15:00-16:00

报告地点:zoom会议,会议ID93995271989

报告人:Amol Yerudkar 意大利萨尼奥大学 博士后研究员

摘要:We study the control of probabilistic Boolean control networks (PBCNs) by leveraging a model-free reinforcement learning (RL) technique. In particular, we propose a Q-learning (QL) based approach to address the feedback stabilization problem of PBCNs, and we design optimal state feedback controllers such that the PBCN is stabilized at a given equilibrium point. The optimal controllers are designed for both finite-time stability and asymptotic stability of PBCNs. In order to verify the convergence of the proposed QL algorithm, the obtained optimal policy is compared with the optimal solutions of model-based techniques, namely value iteration (VI) and semi-tensor product (STP) methods. Finally, some PBCN models of gene regulatory networks (GRNs) are considered to verify the obtained results.

 

 

 

报告题目2Model-Free Self-Triggered Control Co-Design for Probabilistic Boolean Control Networks

 

报告时间:2021720日(周二)下午16:00-17:00

报告地点:zoom会议,会议ID93995271989

报告人:Amol Yerudkar 意大利萨尼奥大学 博士后研究员

摘要:A model-free co-design scheme of triggering-driven controller is proposed for probabilistic Boolean control networks (PBCNs) in order to achieve feedback stabilization with minimum controller efforts. Specifically, Q-learning (QL) algorithm is exploited to devise a self-triggered strategy wherein the controller update time is computed in advance by using the current state information. A new self-triggered QL (STQL) algorithm is presented to achieve the co-design of feedback controller and self-triggered scheme rendering the closed-loop system stable at a given equilibrium point. Finally, some examples are presented to demonstrate the effectiveness of the proposed method.

 

 

 

报告人简介:Amol Yerudkar received the Bachelor’s degree in Electronics Engineering and the Master’s degree in Electrical Engineering (specialization in Control Systems) from the University of Mumbai, India, in 2009 and 2012, respectively. From January 2016 to January 2020, he was a PhD student at the University of Sannio, Benevento, Italy, where he is currently a post-doc researcher with the GRACE (Group of Research on Automatic Control Engineering). His current research interests include systems biology, control of logical networks and learning for control.

 

邀请人:刘洋