Dr. Atanu Kundu

Dr. Atanu Kundu

Dr. Atanu Kundu

I am a Postdoctoral Researcher in the Graduate School of Informatics at Kyoto University, where I am a member of the Computer Software Group and work with Professor Masaki Waga. I received my Ph.D. from the School of Mathematical and Computational Sciences, Indian Association for the Cultivation of Science (IACS), under the supervision of Professor Rajarshi Ray. My research focuses on the verification and falsification of Cyber-Physical Systems using formal methods and machine learning.

Gmail DBLP Google Scholar ORCID LinkedIn GitLab

News & Updates

🎉 Joined Kyoto University as a Postdoctoral Researcher

August 2026

Joined the Graduate School of Informatics, Kyoto University, Japan, as a Postdoctoral Researcher, where I work with Professor Masaki Waga on formal methods for AI enabled CPS.

🌍 Contributions at CPS-IoT Week 2026

Saint-Malo, France · 2026

📄
Paper Presentation

Presented our paper “Falsification of Cyber-Physical Systems using Causation-aware Reinforcement Learning” at 29th ACM International Conference on Hybrid Systems: Computation and Control.

🖼
Poster Presentation

Presented a poster titled “Interpretable and Learning-Driven Falsification of Cyber-Physical Systems” at the Workshop on Computation-Aware Algorithmic Design for Cyber-Physical Systems (CAADCPS).

🎓
PhD Forum

Participated in the PhD Forum by presenting my PhD research titled “Bounded Model Checking and Data-Driven Falsification Strategies for Cyber-Physical Systems” during CPS-IoT Week 2026.

🏆 ACM SIGBED International Travel Grant

2026

I have been awarded an ACM SIGBED International Travel Grant to attend CPS-IoT Week 2026, held in Saint Malo, France.

📢 Paper Accepted at HSCC 2026

2026

Our paper “Falsification of Cyber-Physical Systems using Causation-aware Reinforcement Learning” has been accepted at the 29th ACM International Conference on Hybrid Systems: Computation and Control (HSCC 2026), part of HSCC/ICCPS, to be held during CPS-IoT Week 2026 in Saint Malo, France.

🖼 Poster Presentation at ARCS 2026

February 2026

I presented a poster titled “FlexiFal: A Surrogate-Assisted Falsification Framework for Cyber-Physical Systems” at ARCS 2026, co-located with the ACM India Annual Event, held at IIT Hyderabad.

🎤 Talk

December 2025

I delivered a talk titled “Data-Driven Falsification of Cyber-Physical Systems” and presented a poster titled “A CEGAR-centric Bounded Reachability Analysis for Compositional Affine Hybrid Systems” at the RHPL Workshop, held as part of FSTTCS 2025 at BITS Pilani Goa.

📢 Paper Accepted at IEEE TCAD

September 2025

Our paper “Data-Driven Falsification of Cyber-Physical Systems” has been accepted for publication in IEEE Transactions on Computer-Aided Design of Integrated Circuits and Systems (TCAD).

🛠 Tool Update: FlexiFal Released

July 2025

FlexiFal — a surrogate-assisted falsification framework for CPS with improved surrogate training.

Research Interests

My current research focuses on the verification and safety assurance of AI-enabled CPS, where VLM and VLA models serve as intelligent controllers. Additionally, I am working on surrogate model based falsification frameworks, where several machine learning models utilized as surrogate models.

Cyber-Physical Systems Bounded Model Checking Falsification Machine Learning for Falsification

For more details,

Download CV

Research Projects

SAT-Assisted Graph Encoding

I formulate the discrete structure of hybrid automata as quantifier-free logical constraints to capture paths from initial to forbidden locations. By solving these constraints using the Z3 solver, the approach extracts truth assignments that guide reachability analysis toward unsafe regions. This encoding scales naturally to compositional hybrid systems, enabling the analysis of complex system interactions.

Building FNN Models for CPS

In this project, I develop feed-forward neural network (FNN) models for CPS, including hybrid automata and Simulink models. The constructed networks capture systems with linear and nonlinear dynamics, as well as time-varying input signals. These FNNs serve as surrogate models of the original systems and are leveraged to support applications such as analysis, verification, and falsification.

Publications

ARCH-COMP25 category report: Falsification

Authors: C. Menghi, T. Khandait, A. Kundu, et al.

International Workshop on Applied Verification of Continuous and Hybrid Systems (ARCH25), 2025

ARCH-COMP25 category report: Hybrid systems with piecewise constant dynamics and bounded model checking

Authors: L. Bu, A. Kundu , R. Ray, and Y. Shi

International Workshop on Applied Verification of Continuous and Hybrid Systems (ARCH25), 2025

A Cegar-centric Bounded Reachability Analysis for Compositional Affine Hybrid Systems

Authors: A. Kundu , P. Sarkar, and R. Ray

arXiv preprint arXiv:2509.03560, 2025

Data-Driven Falsification of Cyber-Physical Systems

Authors: Atanu Kundu, Sauvik Gon, and Rajarshi Ray

IEEE Transactions on Computer-Aided Design of Integrated Circuits and Systems (TCAD), 2025

Data-Driven Falsification of Cyber-Physical Systems

Authors: Atanu Kundu, Sauvik Gon, and Rajarshi Ray

17th Innovations in Software Engineering Conference, ISEC, 2024

ARCH-COMP24 category report: Falsification

Authors: C. Menghi, T. Khandait, A. Kundu, et al.

International Workshop on Applied Verification of Continuous and Hybrid Systems (ARCH24), 2024

ARCH-COMP24 category report: Hybrid systems with piecewise constant dynamics and bounded model checking

Authors: L. Bu, A. Kundu , R. Ray, and Y. Shi

International Workshop on Applied Verification of Continuous and Hybrid Systems (ARCH24), 2024

SAT-Reach : A Bounded Model Checker for Affine Hybrid Systems

Authors: Atanu Kundu, Sarthak Das, and Rajarshi Ray

ACM Transactions on Embedded Computing Systems (TECS), 2023

ARCH-COMP23 category report: Falsification

Authors: C. Menghi, P. Arcaini, A. Kundu, et al.

International Workshop on Applied Verification of Continuous and Hybrid Systems (ARCH23), 2023

ARCH-COMP22 category report: Hybrid systems with piecewise constant dynamics and bounded model checking

Authors: L. Bu, G. Frehse, A. Kundu, R. Ray, Y. Shi, and E. Zaffanella

International Workshop on Applied Verification of Continuous and Hybrid Systems (ARCH22), 2022

Selected Talks

Skills & Tools

Technical Skills

SAT Solving Bounded Model Checking Reachability Analysis Machine Learning Robustness Measure of a trace

Tools Developed

FlexiFal

Surrogate-Assisted Falsification Framework for CPS.
NNFal

FNN-based Falsification Framework for CPS.
SAT-Reach

Bounded model checker for affine hybrid systems.

Teaching Experience

Artificial Intelligence (Theory & Lab)

Teaching Assistant — Spring 2022

Object-Oriented Programming with C++ Lab

Teaching Assistant — Autumn 2023

Contact

Interested in collaboration, mentoring, or discussion? Get in touch!

Email Me

Computer Software Group

Graduate School of Informatics

Kyoto University

Kyoto 606-8501 JAPAN