Faculty of Engineering
Developing “Brains” for Robots to Work in Complex Environments
In this edition of “Tell Us Teacher,” we interviewed Professor Fukushima Hiroaki of the Department of Mechanical and Electrical Systems Engineering in the Faculty of Engineering, who specializes in the development of something called “control algorithms,” which act as the “brains” of a robot.
Q：Please tell us about how you started your current research.
I started my current research when I was assigned to a “Control Engineering” laboratory for my senior year graduation project. I had always been interested in robotics, so I went to a department where I would be assigned to a robotics lab, but by the time I was a senior, the professor and assistant professor of the lab had moved to another department, and the remaining assistant professor was more interested in control engineering than robotics, so the lab was essentially a control engineering lab. In effect, it became a control engineering laboratory. Although control engineering is not a discipline that focuses only on robots, it is important for robot control, and I was originally more interested in software than robot hardware, so I requested to be assigned to the laboratory and started research on “control engineering.” When I started, I found that the field required not only computer knowledge such as programming but also a wide range of knowledge in mathematics and physics. I had never really liked mathematics and physics, but as I realized that they were actually useful for real-world problems in control engineering, I began to find them interesting.
Q：What kind of research does control engineering involve?
Control engineering is the study of how to make various systems behave as desired. This includes but is not limited to robots. The main objective is to develop control algorithms that determine the amount of system operation depending on the situation. For example, in the case of a robot, the amount of operation is the voltage of the motors attached to the joints and wheels, and it is essential to have a control algorithm that determines the appropriate amount of operation to make the robot behave as desired. To develop this control algorithm, a mathematical model is used, which is a mathematical expression of the properties of the control target, but since it does not necessarily represent the actual system accurately, the designer has to modify the values in the algorithm by trial and error while making observations during experimentation. In my thesis, I studied how to support these designers’ trial-and-error efforts. That is, how to mathematically derive a modified value for the algorithm that would improve the response of a system based on experimental data.
I realized that the methods derived were actually useful in the real world, and I became interested in control engineering and the various mathematics on which it is based. I continued to work in the same laboratory until I completed my doctorate, and then I also worked as a researcher, so I was there for about 10 years in total. I was able to conduct research in various fields of control engineering and even studied with a prominent professor in San Diego. After I had become rather satisfied with my accomplishments, I was given the opportunity to work in a robotics laboratory. I stayed there for 15 years until I moved to KUAS, where I now perform research on control for lots of different kinds of robots.
Q：Concretely speaking, what specific areas of research do you hope to emphasize?
In particular, we are currently researching two areas where control methods have yet to be established.
One of them is the control of robot swarms. The coordination of multiple robots is expected to have several advantages, such as the ability to perform tasks that would be difficult for individual robots, the ability to solve problems quickly through parallel work, and a minimization of impact on the entire system even if a few of the robots fail. On the other hand, as the number of robots increases, the system becomes more complex, and it becomes difficult to apply conventional centralized control methods. To solve this problem, we are researching a decentralized control method in which each robot decides how to move based on its own surroundings to achieve the overall control objective.
The second is the control of snake-like robots. These elongated snake-shaped robots are expected to be used for information gathering in confined spaces. Snake robots, which do not move using wheels or crawlers but rather by changing the shape of their own body, are interesting because they pose different characteristics from those of conventional control targets, and a control method for them has not yet been established. Although snakes have a simple shape, they can move in a variety of ways with potential application to a variety of environments. Currently, I am mainly working on the movement control of snake robots in water.
Q：How would you like to contribute to society in the future through your current research?
In the future, robots are expected to assist humans not only in factories but also in dangerous environments that are difficult for humans to enter as well as in daily life. To achieve this, not only their hardware but also their software, such as their control algorithms, needs to be further refined. While incorporating the rapidly developing technology of artificial intelligence, I would like to continue my research by focusing on subjects that cannot be handled by existing control methods, such as those described above, and hope that the knowledge gained through this research will lead to the development of new control methods for robots to perform advanced tasks in complex environments in the future.
Q：Can you tell us about your hobbies and interests, and what you were into during university?
Until my third year of undergraduate school, I had relatively more time to pursue my hobbies. Especially during the summer and spring breaks, I went on a homestay in Los Angeles, worked and lived at a ski resort, and drove all they way around Japan.
After being assigned to a laboratory in my fourth year of undergraduate studies, I spent my days immersed in research, but when I was about to start my doctoral research, I started surfing, which had been recommended by an acquaintance, and for a while I was so passionate about it that I did it almost every morning. After the birth of my child, I could no longer go surfing, and now I’m back to being hobbyless, but the unique sensation of riding a wave is so vivid in my memory that it still appears in my dreams, and I would like to do it again someday. When I talk about this kind of thing to my students, they often get interested, but I don’t recommend they do it because they might get too absorbed in it once they start (laughs).
Learn more about Dr. Hiroaki Fukushima
Q：Please tell us about how you started your current research.
I started my current research when I was assigned to a “Control Engineering” laboratory for my senior year graduation project. I had always been interested in robotics, so I went to a department where I would be assigned to a robotics lab, but by the time I was a senior, the professor and assistant professor of the lab had moved to another department, and the remaining assistant professor was more interested in control engineering than robotics, so the lab was essentially a control engineering lab. In effect, it became a control engineering laboratory. Although control engineering is not a discipline that focuses only on robots, it is important for robot control, and I was originally more interested in software than robot hardware, so I requested to be assigned to the laboratory and started research on “control engineering.” When I started, I found that the field required not only computer knowledge such as programming but also a wide range of knowledge in mathematics and physics. I had never really liked mathematics and physics, but as I realized that they were actually useful for real-world problems in control engineering, I began to find them interesting.
Q：What kind of research does control engineering involve?
Control engineering is the study of how to make various systems behave as desired. This includes but is not limited to robots. The main objective is to develop control algorithms that determine the amount of system operation depending on the situation. For example, in the case of a robot, the amount of operation is the voltage of the motors attached to the joints and wheels, and it is essential to have a control algorithm that determines the appropriate amount of operation to make the robot behave as desired. To develop this control algorithm, a mathematical model is used, which is a mathematical expression of the properties of the control target, but since it does not necessarily represent the actual system accurately, the designer has to modify the values in the algorithm by trial and error while making observations during experimentation. In my thesis, I studied how to support these designers’ trial-and-error efforts. That is, how to mathematically derive a modified value for the algorithm that would improve the response of a system based on experimental data.
I realized that the methods derived were actually useful in the real world, and I became interested in control engineering and the various mathematics on which it is based. I continued to work in the same laboratory until I completed my doctorate, and then I also worked as a researcher, so I was there for about 10 years in total. I was able to conduct research in various fields of control engineering and even studied with a prominent professor in San Diego. After I had become rather satisfied with my accomplishments, I was given the opportunity to work in a robotics laboratory. I stayed there for 15 years until I moved to KUAS, where I now perform research on control for lots of different kinds of robots.
Q：Concretely speaking, what specific areas of research do you hope to emphasize?
In particular, we are currently researching two areas where control methods have yet to be established.
One of them is the control of robot swarms. The coordination of multiple robots is expected to have several advantages, such as the ability to perform tasks that would be difficult for individual robots, the ability to solve problems quickly through parallel work, and a minimization of impact on the entire system even if a few of the robots fail. On the other hand, as the number of robots increases, the system becomes more complex, and it becomes difficult to apply conventional centralized control methods. To solve this problem, we are researching a decentralized control method in which each robot decides how to move based on its own surroundings to achieve the overall control objective.
The second is the control of snake-like robots. These elongated snake-shaped robots are expected to be used for information gathering in confined spaces. Snake robots, which do not move using wheels or crawlers but rather by changing the shape of their own body, are interesting because they pose different characteristics from those of conventional control targets, and a control method for them has not yet been established. Although snakes have a simple shape, they can move in a variety of ways with potential application to a variety of environments. Currently, I am mainly working on the movement control of snake robots in water.
Q：How would you like to contribute to society in the future through your current research?
In the future, robots are expected to assist humans not only in factories but also in dangerous environments that are difficult for humans to enter as well as in daily life. To achieve this, not only their hardware but also their software, such as their control algorithms, needs to be further refined. While incorporating the rapidly developing technology of artificial intelligence, I would like to continue my research by focusing on subjects that cannot be handled by existing control methods, such as those described above, and hope that the knowledge gained through this research will lead to the development of new control methods for robots to perform advanced tasks in complex environments in the future.
Q：Can you tell us about your hobbies and interests, and what you were into during university?
Until my third year of undergraduate school, I had relatively more time to pursue my hobbies. Especially during the summer and spring breaks, I went on a homestay in Los Angeles, worked and lived at a ski resort, and drove all they way around Japan.
After being assigned to a laboratory in my fourth year of undergraduate studies, I spent my days immersed in research, but when I was about to start my doctoral research, I started surfing, which had been recommended by an acquaintance, and for a while I was so passionate about it that I did it almost every morning. After the birth of my child, I could no longer go surfing, and now I’m back to being hobbyless, but the unique sensation of riding a wave is so vivid in my memory that it still appears in my dreams, and I would like to do it again someday. When I talk about this kind of thing to my students, they often get interested, but I don’t recommend they do it because they might get too absorbed in it once they start (laughs).
Learn more about Dr. Hiroaki Fukushima