The race for robotics has greatly accelerated in recent years, and more and more companies and institutions are advancing the sector by leaps and bounds. Figure and his multitasking humanoid robot, Tesla and his Optimus prototype (with a few mistakes all the same), Boston Dynamics with its acrobat Atlas and even Apple is starting to take an interest in it.
This time, the innovation comes to us from the corridors of Duke University (North Carolina, USA). Engineers have developed a four-fingered robotic hand, equipped with tactile microphones, which they will present at the Conference on Robot Learning in Munich. Equipped in this way, this hand can perceive its environment like a human being.
From vision to touch
The majority of modern robotic solutions essentially rely on vision for machines to understand their environment. SonicSense transcends this limitation by equipping robots with sophisticated tactile perception. Each finger of this robotic hand incorporates a contact microphone which captures the vibrations generated during physical interactions. This technology naturally filters ambient noise, allowing precise analysis of acoustic signals emitted by the objects handled.
“ We wanted to create a solution capable of working with complex and varied objects encountered in everyday life, giving robots a much richer ability to “feel” and understand the world,” explains Jiaxun Liu, co-author of this study and doctoral student in Mechanical Engineering who participated in the project.
The SonicSense system, in addition to these microphones, also integrates the latest advances in artificial intelligence to transform this vibrational data into concrete information on the nature and structure of objects.
Unprecedented sensory capabilities
The versatility of SonicSense completely pushes the boundaries of current robotic capabilities. By shaking a box of dice, the system precisely determines their number and shape. Vibration analysis allows it to evaluate the liquid level in a closed container. Even more remarkable, by simply tapping, the robotic hand reconstructs the three-dimensional structure of objects and identifies their constituent materials, including on reflective or transparent surfaces!
This performance requires only four interactions for the objects listed in its database, while twenty are enough for him to characterize an object that he has never seen. Boyuan Chen, professor of mechanical engineering and materials science at Duke, emphasizes: “ SonicSense gives robots a new way to hear and feel, just like humans, which can transform how they perceive and interact with objects. While vision is key, sound adds layers of information that can reveal what the eye might miss “.
An accessible and promising technology
The Duke team favors a pragmatic approach. By using standard contact microphones, similar to those fitted to electric guitars, and commercial or other 3D printed components, production costs remain under control. Building the entire system costs just $200.
This financial accessibility will necessarily facilitate its deployment in various industrial contexts and researchers are now perfecting object tracking algorithms to keep the hand effective in cluttered environments. The team is also working on the integration of pressure and temperature sensors, to further enrich the sensory capabilities of the system.
Chen is already planning a great future : “ This is just the beginning. SonicSense could be integrated into more advanced robotic hands with precise manipulation skills, allowing robots to perform tasks requiring a refined sense of touch. We look forward to exploring how this technology can be developed further (…) enabling even more complex interactions “.
The potential of SonicSense is immense : automotive or pharmaceutical industry, quality inspection or high-precision assembly, handling of fragile materials, surgery, etc. If the team continues its improvements, we could even envisage robots equipped with their technology developing a form tactile intelligencewhich could help them collaborate more closely with humans. A dream for some, a nightmare for others.
- Duke University has developed a robotic hand equipped with microphones that allows it to sense its surroundings through touch.
- SonicSense uses vibrations to analyze objects, reconstruct their structure and identify their materials
- This inexpensive technology could be applied to many industrial sectors.
