IEEE CSS Hybrid Systems TC Fall 2025 Seminar Series

We are excited to announce the first Hybrid Systems Seminar Series this Fall! The series will start with a distinguished faculty seminar from Prof. Agung Julius on September 30. We will continue with three more seminars, each consisting of two 30-minute student presentations. The seminars will be recorded and uploaded to the Hybrid Systems TC YouTube channel for offline viewing: https://youtube.com/@HybridSystemsTC.

Dates: Every other Tuesday starting on Sept 30 until Nov 25 (skipping Oct 14)
Time: 15:00 - 16:00 UTC (11am - 12pm EST, 8am - 9am PST, GMT)
Where: Zoom (https://gatech.zoom.us/j/96168286788?pwd=j2tn4m72S8AJa8ZKH7Sd0zceDko56Q.1)

Schedule:

If you have any questions, please reach out to Akash Harapanahalli at aharapan [at] gatech.edu.

Distinguished Seminar I – Dr. Agung Julius

September 30, UTC 15:00 - 16:00 (11:00am - 12:00pm EST)

Dr. Agung Julius

Professor
Department of Electrical, Computer, and Systems Engineering
Rensselaer Polytechnic Institute

“Temporal Logics for Control Systems: Verification, Synthesis, and ML-driven Property Inference”

Abstract

Temporal logics provide an interpretable and unambiguous framework for reasoning about temporal behaviors in dynamical and control systems. Their formal semantics and resemblance to natural language make them useful tools for specifying, verifying, and synthesizing control systems, enabling rigorous analysis of system behaviors over time. This talk introduces temporal logics in these contexts and demonstrates their application through examples. Beyond these foundational aspects, we discuss the intersection of temporal logics and machine learning, focusing on how data-driven methods can infer system properties in scenarios where explicit specifications or behavioral descriptions are incomplete or unavailable. By integrating formal methods with machine learning, we explore new possibilities for developing interpretable and high-performance machine learning models.

Seminar II - Zhaoyong Liu and Frederik Baymler Mathiesen

October 28, UTC 15:00 - 16:00 (11:00am - 12:00pm EST)

Zhaoyong Liu

Zhaoyong Liu received the B.E. degree in automatic control from the Nanjing University of Science and Technology (NJUST), Nanjing, China, in 2019. He is currently working toward the Ph.D. degree in control science and engineering from the School of Automation, NJUST. He is now under the supervision of Prof. Haoping Wang. Besides, he is a visiting student at the University of L’Aquila from 2024.10-2025.10, whose co-supervisor is Prof. Elena De Santis. His current research interests include switched systems, observability, sampled-data observer, and adaptive observer.

Event-Triggered State Estimation for Networked Switched Systems: An Output Predictor Approach

Abstract

In this article, we investigate the event-triggered state estimation for networked switched systems with (a)synchronous switching. A new multi-mode output predictor-based sampled-data observer and an event-triggered mechanism are proposed. Between event-triggered sampling times, an output predictor is utilized to compensate for the impact of sampling, and at the sampling instants, the predicted output is reset with the updated sampled output. Two scenarios are considered: the observer’s switching signal is synchronous with the system’s switching signal, and the switching signals among them are asynchronous, and frequent switching is allowed during interevent intervals. Average dwell time and linear matrix inequalities techniques are employed to guarantee the global uniform exponential convergence of the proposed observer. Moreover, Zeno phenomenon can be ruled out by proving the existence of a positive lower bound of interevent intervals. Finally, the effectiveness of the proposed approach is illustrated by a circuit system and an academic example.

Frederik Baymler Mathiesen

Frederik Baymler Mathiesen is a PhD candidate at Delft University of Technology, working under the supervision of Luca Laurenti and Simeon C. Calvert. Frederik received his M.Sc. and B.Sc. in Computer Science from Aalborg University, Denmark. He was a visiting student at UC Berkeley in spring 2020. His work focuses on scalable verification methods, including synthesis methods for stochastic barrier functions, robust MDP-based abstraction methods, and bound propagation techniques for learning-based systems.

Formal Verification of Stochastic Systems: From Stochastic Barrier Functions to Abstraction-Based Methods

Abstract

Formal verification of stochastic systems is increasingly important with the growing prevalence of autonomous safety-critical systems. Two main approaches have emerged for this purpose: stochastic barrier functions (SBFs) and finite-state abstraction-based methods. SBFs rely on the construction of a value function to bound safety probabilities, while abstraction-based methods employ model checking techniques on simplified and carefully constructed finite-state models.

In this talk, I will present an overview of both approaches, including recent advances in SBF synthesis - such as data-driven and neural formulations - and developments in abstraction methods and target models. Despite their apparent differences, the methods can be unified within a dynamic programming framework, revealing the context where each method excels. The talk will conclude with a discussion of emerging opportunities for merging these techniques.

Seminar III - Emmanuel Junior Wafo Wembe and Mira Khalil

November 11, UTC 15:00 - 16:00 (11:00am - 12:00pm EST)

Emmanuel Junior Wafo Wembe

PhD Student at Mohammed VI Polytechnic University

“TBD”

Abstract

TBD

Mira Khalil

PhD Student at CRAN, CNRS, GREEN, Université de Lorraine

“Estimation of the minimum and maximum states of charge of lithium-ion battery packs: A hybrid approach”

Abstract

The safety and longevity of lithium-ion batteries depend on the battery management system’s (BMS) ability to accurately estimate internal states, including the state of charge (SOC). One major challenge is that batteries are typically implemented as packs made of a potentially large number of cells. As a result, we may not be able to duplicate SOC estimation algorithms developed for single cells due to the BMS’s limited computational capabilities. An alternative consists of only estimating the minimum and maximum SOCs of the battery pack. Given these quantities, we can ensure that the SOCs of all the cells are within the operating limits. The problem is that the limiting cells, i.e., the cells having the minimum and maximum SOCs, cannot be directly determined based solely on the available measurements. Moreover, these limiting cells may change with time. In this talk, we present a hybrid estimator of the minimum and maximum SOCs, whose dimension is independent of the number of cells and is thus attractive for large battery packs. For this purpose, we consider a battery pack consisting of heterogeneous cells interconnected in series, each being modeled by an equivalent circuit model. We then present the hybrid estimator, which relies on a mechanism that determines online two cells, which are candidates for having the minimum and maximum SOCs. Upon the selection of the cells, the estimator runs an observer designed to estimate their SOCs. We establish robust exponential stability properties for the estimation errors on the minimum and maximum SOCs. We also rule out the Zeno phenomenon. Finally, we illustrate the relevance of the proposed estimator with a numerical example.

Seminar IV - Othman Cherkaoui Dekkaki and Youssef Ait Si

November 25, UTC 15:00 - 16:00 (11:00am - 12:00pm EST)

Othman Cherkaoui Dekkaki

Postdoc at Mohammed VI Polytechnic University

“TBD”

Abstract

TBD

Youssef Ait Si

PhD Student at Mohammed VI Polytechnic University

“TBD”

Abstract

TBD