Press "Enter" to skip to content

Invited talk by Prof. Hyungbo Shim at ICROS 2021

Prof. Hyungbo Shim was invited to 36th ICROS (Institute of Control, Robotics and Systems, Korea) and held a lecture on “Blended dynamics theorem and its applications for networked dynamic systems” in June 24, 2021 at Yeosu, Korea.

Prof. Shim presented what is proper control theory that can let a large number of agents cooperate in each other to achieve a given goal.

출처: 로봇신문

He mentioned three key properties of the control system for multi-agent systems. First, such control systems should operate in a distributed manner, which means that each agent only communicates with its neighbors without a central unit. Second, each agent needs to be self-organized. That is, every agent must be able to design its control system in accordance with its environment. Other necessary features include plug-and-play operation, which means that agents are able to join or leave the network of agents during the operation without any reset process. Lastly, such control systems should be robust against some malfunctioning agents.

As a way to design a control system that satisfies these properties, blended dynamics theory was introduced. For multi-agent systems with different dynamics, if we apply diffusive coupling between the agents with a strong coupling gain, each agent exhibits emergent behavior towards the average dynamics of the system. A method for designing a distributed algorithm based on this theorem was also introduced. This is a way to design blended dynamics first and assign them to the dynamics of individual agents separately.

Subsequently, prof. Shim presented several useful examples which blended dynamics theorem can be applied to. One of the applications is economic power dispatch problem. It is an example of distributed optimization, solving the amount of power to be generated at each agent while minimizing the overall cost of generation given a power demand of each agent. This problem can be solved distributively, utilizing blended dynamics theorem. Other examples include network size estimation, distributed control of multi-channel plant, distributed p-quantile solver, and coupled oscillators.

출처: 로봇신문

Through this lecture, professor Shim emphasized the use of control theory in various applications and delivered the key idea of blended dynamics theorem.

This lecture was recorded in Korean and uploaded here.