Skip to content

Recruit

Students

SOKENDAI?

NII faculty members are basically concurrently appointed as faculty members of SOKENDAI (The Graduate University for Advanced Studies), and can supervise SOKENDAI students. SOKENDAI is a graduate university that is a collection of several national institutes and accepts (working) students for the 5-year integrated program (master's course + doctoral course) or for the 3rd year transfer (doctoral course). Unfortunately, it is a very special university and not very well known, but it is attractive for the following reasons.

  1. Quality of instruction: Students can receive direct instruction from top-class faculty members who work at the NII and other national institutes. Moreover, the faculty members are able to devote themselves to research because they have few classes and other duties, and the number of students is small, so they can provide very close supervision.
  2. Financial support: Students are employed as RAs at SOKENDAI and can earn approximately 100,000 yen per month if they devote themselves to their research.
    In addition to this, our laboratory can provide other fees for participation in some projects.
  3. International exchange: In addition to a high ratio of international students, we regularly have short-term international students who stay in our laboratory for about six months through NII's internship program (see below).

Please check the official information about entrance examinations.

Before applying...

Since our laboratory conducts research that integrate robotics with machine learning, a lot of specialized knowledge and experience must be required for our research. Although you do not need to be a utility player who can do everything from the time you enter our laboratory, I recommend you to have either of the following.

  1. Experience in robot hardware design and development: Our research focuses on robot intelligence, which is inseparable from the robot's embodiement. A unique robot will have or need its own intelligence, and that alone can be a starting point for new research. However, I am a software engineer, so I would be very happy to have students who have experience in building robots from scratch.
  2. Knowledge of robot motion control: When introducing machine learning to a robot, we must carefully consider which layer to introduce. For example, a lower layer of motion controllers can be trained easily by exploiting existing control theory, or a better motion controller can be implemented and a higher layer can generate reference values to it. In either case, knowledge of the robot's motion control is very important, and neglecting this knowledge will not allow the robot to achieve the full potential of its motion performance.
  3. Knowledge of reinforcement learning: The intelligence we study in our laboratory is a kind of motor intelligence, and technically speaking, we often focus on reinforcement learning and imitation learning. In this context, we mainly improves the learning algorithms theoretically, which requires a solid understanding of the basic and latest theories. In particular, reinforcement learning is different from general machine learning, so if you want to work on a research topic related to reinforcement learning, I recommend that you prepare in advance.
  4. Original theme: Since we have several projects going on in parallel, it would be helpful if you could participate in one of them. On the other hand, if you have special experience, many of the problems and questions that you have noticed there can become excellent research topics. If you do not have sufficient knowledge and experience in the above aspects, but are enthusiastic about original themes, I welcome your suggestions.

On the other hand, even if you have these specialized knowledge and skills, I believe that personality is also important to promote and guide your research smoothly. Personally, I would like to work with the following people.

  1. A cooperative person: Since the entire system of recent robots tends to become more and more complex, it is often necessary to implement peripheral technologies even if you just want to verify your proposed method, and there is a limit to independent development. In addition, for safety and efficiency, robot experiments are often conducted by more than one person. Therefore, it is desirable to build a relationship in which we can cooperate with each other in various situations, such as implementation, and experimentation, without being selfish.
  2. A person who can work actively: NII is a research institute, and I am still an active researcher. Therefore, it is difficult to keep an eye on students' progress and follow up with them at all times, so time is wasted if you wait passively for help (until the regular reports/meetings). If you have any problems, please do not hesitate to ask questions (if I am really busy, I will designate a free time).
  3. A person who can manage his/her time (to some extent): In research activities, we often lose track of time and make drastic revisions to our plans, but on the other hand, there are also deadlines in various situations, such as paper submission and review. Missing these deadlines can be a source of trouble, so it is good if you can manage your time with an awareness of the deadlines yourself, or at the very least work according to the guidelines I set for you.

If you are interested in taking the exam, please contact us from Contact.

NII Internship Program

NII offers short-term internships (up to six months) for international students twice a year, as described in here. I also have topics related to robot learning available, so if you are interested in taking the SOKENDAI exam or doing research in Japan, please apply this program.

JSPS Postdoctoral Fellowships

JSPS offers several postdoctoral fellowships (see here). If you want to work in Japan, these should be one option. When considering the application, please contact us in advance from Contact.