---

Hi, my name is Yoshinari Takayama. I work on making formal guarantees hold up on real systems — taking control specifications that are expressive on paper and making them tractable to compute, real to deploy, and sound to trust.

I’m a 3rd-year Ph.D. student in control system engineering at the Laboratory of Signals and Systems (L2S), CentraleSupélec, University of Paris-Saclay in France, funded by the French National Research Agency (ANR). I completed my master’s degree at the Graduate School of Informatics, Kyoto University, advised by Prof. Toshiyuki Ohtsuka, and my bachelor’s degree at the School of Informatics and Mathematical Science, Faculty of Engineering, also at Kyoto University.

---

My work follows one question: do formal guarantees survive contact with reality? Each thread removes a different obstacle between an expressive specification and a system you can actually trust.

• Tractable optimization for control: turning expressive but intractable specifications — like the mixed-integer encoding of Signal Temporal Logic — into convex programs that scale to long horizons.
• Formal methods for cyber-physical systems: assume-guarantee contracts and runtime verification that certify safety and liveness on continuous-time hardware, not only in the discrete-time model that planned it.
• Safe learning and control: equipping reinforcement learning (RL) and model predictive control (MPC) with guarantees that hold under disturbance, delay, and model error — so learned policies inherit the same certificates as designed ones.
• Power systems as the proving ground: where the theory meets a real grid — battery cycling, multi-level overload tolerance, and hybrid energy storage under genuine operating rules.

Most recently, I’ve been turning the same question on learning systems themselves — the theoretical foundations of fine-tuning RL for foundation models — asking what adaptation and generalization mean when the object being controlled is the learner.

---

Journal Reviewer

• Systems & Control Letters (L-CSS)
• IEEE Robotics and Automation Letters (IEEE-RAL)

---

• 2023 — Encouragement Award, Institute of Systems, Control and Information Engineers (ISCIE)
• 2023 — SICE International Young Authors Award (IEEE CDC 2022)
• 2022 — Best Student Presentation Award, SCI 2022