The human sense of touch is an intricate and highly developed system, allowing us to perceive and interact with the world in a nuanced way. Our skin is equipped with an array of sensors, including mechanoreceptors, that detect pressure, vibration, and more. This intricate network enables us to experience a vast spectrum of tactile sensations, from the gentle brush of a breeze to the firm handshake of a friend.
Reproducing this capability in robots is a complex challenge that has proven difficult to achieve with anywhere near the same level of sophistication. One key limitation is the inability of existing sensors to simultaneously achieve high sensitivity and a rapid response to changes in texture. These capabilities must coexist and be exquisitely sensitive to rival human-like sensing performance.
Due to technical limitations, current sensing devices often prioritize one aspect over the other, leading to a trade-off between sensitivity and responsiveness. Achieving a delicate balance that mirrors the human sense of touch is essential for enabling robots to navigate and interact with their environments with the same level of dexterity and intuition as humans.
Until researchers can overcome these challenges, robotic touch sensors will fall short of the needs of applications involving more sophisticated and versatile robotic interactions, such as those in fields ranging from manufacturing to healthcare. A step in the right direction has recently been announced by a team led by researchers at the Southern University of Science and Technology. They have developed a robotic sensory system that can detect both static and dynamic stimuli with a high level of accuracy. They have shown that this system is useful in distinguishing even very fine details of surface features.
The key to the team’s success lies in the tunable electric double layers, with a nanoscale charge separation for capacitive signals, that make up their sensors. In conjunction with the use of a low-viscosity ionic material that was leveraged to create a unique microstructural design, these innovations allow the sensor to be both very accurate and capable of sensing high speed vibrations. When combined with a machine learning classifier, it was found that this system could learn to recognize some very complex features.
The soft, flexible sensor was attached to the fingertip of a prosthetic human hand as part of a test rig designed to assess the performance of the system. The fingertip was then slid over 20 different types of textiles, and the sensor readings were forwarded into a random forest classification algorithm. Classifications were provided in real-time, and an average accuracy rate of 100% was observed. While this is clearly a very impressive result, it is important to note that only 20 textures were included in the study. Larger scale experiments would need to be conducted to give a better idea of how the sensing system would perform under real-world conditions.
Looking ahead, the researchers believe that with some refinement of their techniques, there may be a number of potential commercial applications for their technology. These applications could be in areas as diverse as virtual reality and artificial prosthetic limbs.