报告题目：Data-Driven Robust Model Predictive Control
报告时间： 2025. 09. 22（周一）上午10：00
报告地点： 民主楼118
报告人：邓梨 (博士后)，University of Alberta
摘要: Model predictive control (MPC) has become an advanced control approach due to its capability of receding horizon control and handling physical constraints. Typically, a prediction model is required to be accurate enough to capture the true dynamics of a real system, thus ensuring the closed-loop performance. However, with the increase of the complexity of controlled systems in engineering, deriving accurate models is becoming more and more challenging, and even impossible. To address this challenge, this talk introduces two data-driven MPC approaches for unknown systems. First, for linear time-invariant (LTI) systems, we present a data-driven robust MPC method that employs a terminal inequality constraint, eliminating the need for explicit system identification or online estimation. Second, for unknown linear time-varying (LTV) systems where the time-varying system matrices are assumed to lie within an unknown polytope, we propose a robust MPC scheme with event-triggered learning. In this framework, model estimation and polytope learning are activated only when necessary, thereby ensuring both robustness and efficiency.
报告人简介：邓梨，加拿大阿尔伯塔大学电气与计算机工程系博士后。目前主要从事数据驱动与基于学习的鲁棒模型预测控制研究。2023年8月获阿尔伯塔大学电气与计算机工程系（控制系统专业）博士学位，导师是Tongwen Chen教授（加拿大皇家科学院院士、加拿大工程院院士、加拿大智能监测与控制领域首席科学家、IEEE Fellow、IFAC Fellow）。在国际重要期刊和会议累计发表论文24篇。担任《Automatica》、《IEEE Transactions on Automatic Control》等多个权威期刊审稿人。曾担任American Control Conference等国际会议分会/专题主席。
