Prof. Toshiyuki Ohtsuka

Department of Systems Science, Graduate School of Informatics
Kyoto University, Japan


Real-Time Optimization Algorithm for Nonlinear Model Predictive Control


The aim of this talk is to share the idea of a real-time optimization (RTO) algorithm tailored for nonlinear model predictive control (NMPC) and to introduce some cutting edge applications. Model predictive control (MPC) is one of the most successful control techniques in industrial processes. At each sampling time of MPC, the response of a plant over a finite future is predicted by using a model of the plant and is optimized by minimizing a given performance index subject to some constraints. MPC can handle various types of control objectives and constraints and achieve the best possible performance as long as the optimal control problem over the finite future can be solved at each sampling time in real time, which is an appealing feature of MPC as a general framework of feedback control for nonlinear systems and for realizing any sort of intelligent systems. Although NMPC, MPC for nonlinear systems, had long been considered unrealistic owing to the heavy computational burden for optimization, advances in computation methods and computers have been expanding the frontiers of NMPC. In particular, it is possible to develop an efficient RTO algorithm on the basis of the assumption that the sampling period is sufficiently short. The RTO algorithm is obtained as an initial-value problem of an ordinary differential equation for an unknown quantity to be optimized. That is, the RTO algorithm can be viewed as a dynamical system to generate a time-dependent optimal solution with no iterative search. Through the use of the non-iterative RTO algorithm, NMPC with a sampling period on the order of milliseconds is becoming more and more common. Application examples of NMPC nowadays include a robotic manipulator, a radio-controlled hovercraft, the autopilot of a ship, path generation for an automobile, and a steel making process. NMPC of distributed parameter systems such as thermofluid systems is one of the ongoing research challenges. The idea of the RTO algorithm can also be adopted in differential games, estimation, and adaptive control with moving horizons. Moreover, a program of the RTO algorithm for a particular application can be automatically generated by such symbolic computation languages as Maple and Mathematica, which is quite useful for dealing with complex nonlinear systems.

Brief Biography:

Toshiyuki Ohtsuka was born in 1967 in Tokyo, Japan. He received his Bachelor, Master and Doctor Degrees in Engineering from Tokyo Metropolitan Institute of Technology, Japan, in 1990, 1992 and 1995, respectively. From 1995 to 1999, he worked as an Assistant Professor at the Institute of Engineering Mechanics, the University of Tsukuba. In 1999, he joined Osaka University as an Associate Professor at the Department of Mechanical Engineering, the Graduate School of Engineering, and he was a Professor at the Department of Systems Innovation, the Graduate School of Engineering Science from 2007 to 2013. In 2013, he joined Kyoto University as a Professor at the Department of Systems Science, the Graduate School of Informatics. His research interests include nonlinear control theory and real-time optimization with applications to mechanical systems and environmental systems. He is a member of SICE, IEEE, ISCIE, AIAA, JSASS, and JSME. He was the recipient of the SICE Control Division Pioneer Award in 2006, SICE Best Paper Awards in 2004 and 2013, SICE Best Writing Award in 2012 for his book, “Introduction to Nonlinear Optimal Control” (in Japanese), among other academic awards.