• Back
• Next

日本語

Elderly people, whose muscle strength weakens with age, may need to receive assistance in their daily lives, so robots are being developed to assist with especially frequent actions such as standing up and walking. Conventional robots have normally assisted in one of these actions, but since standing up, walking, and sitting down are done in sequence, it is desirable to have a single robot assist with them all.

Moreover, large-scale systems are often constructed to produce great force or recognize the user's state in detail, but systems for household use should ideally be small and have few sensors. If all movements are initiated by the robot, it may lead to dissatisfaction on two counts: muscular weakening and a reduced sense of independence on the part of the users. As such, they have developed a robot that allows the user to initiate movement and assists as appropriate by recognizing the user's situation and adding the force that is missing.

The research team focused on center of gravity as a key to estimating the robot user's state. Center of gravity is effective as an indicator of a person's state, cannot be accurately assessed with only a few sensors, so they developed a technique for calculating the range where center of gravity candidates can be found by considering data such as the motion range of joints. They produced a robot that uses these center of gravity candidates to estimate the user's state and detect irregularities through a type of machine learning called support vector machine and then move according to that information.

The research team believe that estimation of people's states using center of gravity candidates can be applied to welfare robots of various shapes and uses. Moreover, they want to make robots easier to use for elderly users by better understanding what movements the robots are expected to perform and how they should communicate with the users.