Skip to main content

Description of research themes

Description for the general public


"What kind of computational machine is the brain?"

is the main theme of my research. I think that the brain is an input-output converter that produces output signals (motor commands) based on sensor (sensory) input. Therefore, I think that the brain is a "input-output converter".

How does the brain move the body?

I am researching with an emphasis on "How does the brain move the body? More specifically

  • How does the brain learn various types of exercise?
  • To find out the skills and tricks of exercises that the brain has learned.
  • Consider how the brain learns and chooses actions according to the situation.
  • Other things that each person wants to do (but which Nishii recognizes as interesting and worthwhile as research).

The following is a summary of the program. As a byproduct of the main theme, research on machine learning methods such as neural networks and reinforcement learning will be It may also be a part of the main theme.

Explanation of research theme

When practicing sports, we can

"You learn exercises with your body.

but this is a little bit wrong. The truth is

The truth is that exercise is learned by the brain.

The correct answer is that the brain is the one that gives the commands for physical exercise. The brain is the one that issues the commands for physical exercise, so the important thing is to train the brain that issues the commands to move the muscles. (The important thing is to train the brain that gives the commands to move the muscles.)

In order to understand how the brain issues commands for movement, we are studying the following two major approaches.

  • What kind of movement is the brain trying to achieve ...... Understanding the movement itself (understanding from outside the brain)
  • How does the brain achieve the desired movement ... Learning and control mechanisms (understanding from inside the brain)

For more specifics, please see the following. (Please also visit Nishii's homepage.)

Research on Animal Walking

Dogs and cats move their legs in a different order depending on their walking speed (if you don't know this, take a good look at dogs and cats). Six-legged insects such as ants and cockroaches also change the order in which they move their legs according to their walking speed. (Actually, you can't tell even if you look closely. ) They are too small and move too fast, and cockroaches jump at you. By the way, cockroaches are said to use only two hind legs to run when they are running as fast as they can...) (The cockroaches are said to use only two hind legs when they are running as fast as they can. They should always move their legs in the same order.

Therefore, we have studied why animals change their walking style depending on the speed of walking by using robotics, and have shown that animals need to use several different walking styles when they walk in "pursuit of ease".

Research on the knack of locomotion

Even if you build a walking robot, it is quite difficult to make it walk well. Simply moving the left and right legs alternately will make the robot fall down and become like a spoiled child.

On the other hand, many multipedal animals can walk not only on flat surfaces but also on bumpy paths and slopes, and humans can even kick a ball or swing a racket. We are tackling the question of what is the secret of humans and animals being able to walk across various types of exercise, and what are the tricks of exercise that expert sportsmen know. Specifically, we are trying to find out the tricks and skills of exercise through movement analysis.

Research on motor learning

Target cup acquisition. Visual coordinate position confirmation. Immediately determine the target length of each muscle in the arm. Upper arm muscle is deployed. 0.1 second elapses. Immediately activate the lower arm muscles. Increased output of thumb extensor muscle. Insufficient output of the extensor output of the little finger. At this rate, he will hit the target. Arrrrrrrrrr. Cup collapsed. Switch to escape action immediately.

When you walk, it's even more serious. I am working on the theoretical elucidation of how the brain learns and realizes various movements by controlling the entire motor (muscle). In my thesis and master's thesis, I have been working on the construction of a learning model for a biped robot.

Research on action selection

When you are walking in the forest and see a delicious-looking fruit, do you call out to your friends and eat it together? Or would you keep it to yourself? According to Darwin's theory of evolution, it is advantageous to keep the fruit to yourself because you can survive longer and produce more offspring than other individuals, but humans do not always act selfishly. The reasons for this are discussed using a virtual agent simulation.


So, although we have a main theme, we sometimes do other things depending on Nishii's whims or the wishes of the residents of the laboratory. Please refer to thesis/graduate-report/.