With increased uses of robots, the technology for controlling them is also improving. Moving from simple joysticks, we can now control robots with our body movements via wearable devices. Sensors aree of the most important parts of such devices. In Korea and overseas, sensors that are highly accurate and comfortable to wear are being developed.

◇ One Sensor Detecting Multiple Stimuli 

Sensors are supposed to be thin, light, and soft. A research team led by Professor Yong-Lae Park of SNU Dept. of Mechanical Engineering developed a soft sensor that is both flexible and highly accurate. Their research was published the last month’s issues of Science Robotics.

The sensor developed by the team has a length of 7cm and width of 6mm. A single sensor can detect multiple deformations including extension, bending, and compression. Previously, sensors were able to detecte type of deformation, requiring multiple sensors for multiple operations.
The newly developed sensor is made out of silicon rubber which can be easily extended and curved. There is a thin pipe contained within a slim bar-like structure. The thin pipe is filled with transparent, conductible ionic solution.
Flexing and compressing the sensor causes the light permeability and conductivity of the ionic solution to change; measuring this makes it possible to detect different movements. The research team explained that by attaching the sensor to the wrist, we can remotely send various commands to robots or computers with body movements. “With this technology, we can develop easier and more intuitive methods to remotely control robots,” said Prof. Park.
An important aspect in developing sensors is how easily it can be attached to the body. Last August, a research team from National University of Singapore developed smart gloves for video games with sensors. The glove was made out of sensitive microfiber fabric, making it lighter and flexible. The principles are similar to those used by the SNU team; thely difference is that it is made into a glove.
This sensor is also extremely thin, with the width of a single hair, and filled with conductible liquid. The differences in electric signal makes it possible to detect movements. The prototype weightsly 40g.

The gaming smart glove can detect 11 different movements of the player. For instance, the player can pull their pointer finger to pull the trigger of a gun, then twist their wrist clockwise to move forward.

◇ Overcoming Limitations of Sensor Material with AI 

The sensor must be flexible to be attached to the human body, which is why it is composed of high molecular substance. However, using the sensor multiple times causes the substance to transform, lowering its sensitivity. Recently, researchers have developed a way to overcome this limitation in material with AI. AI tracks the changing signal over time and calibrates it.
Last month, a research team from UC Berkeley succeeded in combining AI with the sensor. First, they created a flexible band that can detect electric signals from 64 points from a human arm. The band collects electric signals that come from movements of the arm and the hand.
When someone tries to move, their brain sends electric signals to muscle fibers in the arm and the hand via neurons. The research team made the AI learn the characteristics and differences of these signals. It learned to distinguish between 21 different hand movements such as putting a thumb up, making a fist, and counting with fingers.

During this process, the sensor can calibrate the information learned via AI. For example, even though the signals coming from the same motion of making a fist become different over time, it recognizes the motion as identical. It calibrates the signals according to how they weaken over time.

Also, even when the user makes sudden movements like raising their arm above their head, the sensor does not perceive it as a new motion signal. The research team explained: “The signals coming from the user will change over time, affecting how the sensor operates. We were able to greatly improve the accuracy of categorizing different movements by updating the signal analysis model with AI.”