A study conducted by Tatsuya Miyano was selected for presentation at The 2025 American Control Conference (ACC).
In control applications such as robotics and energy systems, it is often difficult to describe the internal dynamics of a system using an accurate mathematical model due to complexity and uncertainty. Nevertheless, the system’s responses to inputs and states can be observed. In such practical settings, including real machines and high-fidelity simulators, control theories that can treat the system as a black box without relying on explicit mathematical models are highly desirable. Passivity, which characterizes systems that do not generate more energy than they receive from the outside, has been widely used as a powerful concept for guaranteeing stability. However, most conventional passivity-based control methods assume detailed knowledge of the system’s state equations, making them difficult to apply to unknown nonlinear systems.
In this study, we focus on unknown nonlinear systems whose dynamics are nonlinear in the state but affine in the input. Using information on inputs, states, and outputs, we construct a passive reference system that is closest to the observed behavior of the original system at each moment. An adaptive control law is then designed to drive the real system toward this reference system. The proposed approach does not require precise prior modeling and provides stability guarantees suitable for real-world applications. It is expected to serve as a foundational technology for safely interconnecting complex systems in robotics, energy systems, and beyond.
This research was published in IEEE Control Systems Letters prior to its presentation at ACC (https://doi.org/10.1109/LCSYS.2024.3520031).
Title: Adaptive Passification of Unknown Input-Affine Nonlinear Systems
Authors: Miyano, T., Shima, R., Ito, Y.
Appears in: The 2025 American Control Conference
Presented: July 10, 2025