Present Robots and Future Potentials Robotics Society of Japan Seminar "Unsolved Problems in Robotics"

Present Robots and Future Potentials Robotics Society of Japan Seminar "Unsolved Problems in Robotics"

Environmental intelligence and structuring for robots

It also presents the future market forecast of the robot industry and introduces the current state of initiatives such as the government's "Robot Revolution Realization Conference" and "Robot New Strategy" as there are personal services and public services in the service field where growth is expected. did. In addition to technology, it is important to improve the environment in a broad sense, and the concept of "RT (robot technology)" is introduced. "Robot technology" has technological elements of sensors, intelligence/control systems, and drive systems. He stressed that it should be regarded as an intelligent “mechanical system”. It is important to give solutions to missions and systemize them. He said that it is the true value of robot technology to integrate elements to realize functions, not just elements, but it is difficult to understand.

He said that it is important to design the environment itself, including the environmental intelligence and structuring to make it easier for the robot to move. He introduced the research concept of Professor Asama so far. In addition, automation, unmanned, and autonomy have different meanings. Automation is to operate a machine by performing some operation on the environment. Unmanned operation also includes remote control. And autonomy is a technology in which the environment is not limited, and actions can be generated and moved by themselves. Mobile robots can also be captured in different frameworks. Autonomy is less necessary when environmental factors are limited, such as on-orbit movement. Nowadays, highly accurate self-localization system can be used by utilizing satellites and AI, but the current robots cannot move in unlimited environment. There is no doubt that this is a research theme that will be required in the future.

Keidanren's “Industrial Competitiveness Roundtable” is also making efforts to make recommendations to the country regarding “human coexistence robotics”. Robot technology for direct interaction between a user and a robot. Information technology and safety technology for coexistence with users and coexistence with third parties are required there. For that purpose, technology for knowing people and understanding/estimating behaviors and intentions is also important. On the other hand, there is a strong need for labor saving due to the lack of manpower. However, he said that not only economic perspectives, but also new value creation and humanitarian perspectives are extremely important.

Safety and security are important in human coexistence robots. Mr. Asama introduced the idea of   "ALARP (As Low As Reasonably Practicable)". In a word, we set a standard value that is not acceptable if it cannot be achieved and a standard value that is considered to be sufficiently safe, in the sense that risk should be reduced as much as reasonably practicable, and the risk The idea is to judge and respond accordingly.

He introduced the research on the Kakenhi "Mobile Knowledge". This is a study that aims to scientifically elucidate the "adaptive behavior mechanism" that is the intellectual ability of animals to appropriately act according to the environment. Movement of the body results in physical interaction between the environment and the body. That information enters the brain through the body. And adaptive behavior is born. If the mechanism can be elucidated, robots that can act adaptively can be realized. First of all, it is different from the conventional way of thinking that there is action. By applying this research, we are also developing a stand-up support system.

In addition, Mr. Asama introduced topics such as the introduction of robots related to the decommissioning of the Fukushima Daiichi Nuclear Power Plant and the intervention of machines that are important in remote control.

Fusion of deep learning and robotics, structure of intelligence, future of science

Yutaka Matsuo, a professor of the Graduate School of Engineering, the University of Tokyo, gave a lecture entitled "The Possibility of Fusion of Deep Learning and Robotics". Mr. Matsuo first introduced cases such as "Alpha Zero". "Alpha Zero" takes a strategy different from the conventional theoretical framework, such as moving the king general to the middle stage in the beginning of shogi. This shows that machine learning technology can find good strategies that are not bound by human thinking habits. In addition, applications such as "AlphaFold", which can predict the three-dimensional structure of a protein from an amino acid sequence, are emerging.

Mr. Matsuo's development of machine learning is that the deep learning revolution advances from recognition to learning of physicality and movement to understanding of the meaning of language. In 2014, we forecast that the application will go through crime prevention and autonomous driving, housework and care, and through true language processing to translate and support white-collar, so far, roughly as expected That.

Since 2012, the application of image recognition technology started around 2015, and now there are more and more commercialized products, such as "Yolo" for object detection by the University of Washington.

Research on exercise proficiency is being carried out using deep reinforcement learning technology. This has also begun around 2015, and much research is underway. Matsuo gave examples of research on UC Berkeley's "BRETT", Google's distributed learning, and robotic hands using OpenAI. In addition, DeepX, one of the ventures from Matsuo Laboratory, is working on automatic control of construction equipment using image recognition in joint research with Fujita.