The speedy rise of synthetic intelligence and robotics has targeted consideration on software program, massive language fashions, and more and more highly effective processors.
However for robots to function successfully within the bodily world, additionally they want one thing way more basic: the power to sense and perceive their atmosphere.
That problem is creating rising curiosity in superior sensing applied sciences able to offering machines with far richer details about the world round them.
Among the many corporations working on this space is Digid, a Germany-based expertise developer specializing in nanoscale sensors for power, temperature, and different measurements.
The corporate says its sensors are sufficiently small to be built-in straight onto buyer merchandise and parts, enabling measurements in places the place typical sensors can’t match.
Digid has industrialized its nanosensor expertise and produced a couple of million sensors to this point, with functions spanning robotics, medical units, wearables, industrial techniques, and AI infrastructure.
Based in 2019, Digid initially targeted on diagnostics and BioMEMS applied sciences earlier than increasing into broader industrial and robotics functions. Extra lately, the corporate attracted consideration at CES, the place it showcased its nanoscale sensing expertise and its potential function in next-generation robotics and clever techniques.
On this interview, Nils Könne, head of product, and Christian Kreil, head of enterprise improvement, talk about how nanosensors may assist tackle one in all robotics’ most persistent challenges: tactile sensing.
They clarify how high-density sensor arrays may allow robotic pores and skin and extra succesful robotic palms, why native processing will change into more and more necessary as sensor counts rise, and the way related applied sciences may additionally enhance medical units, wearable merchandise, and AI information middle infrastructure.
For Robotics & Automation Information readers, the dialogue gives an fascinating glimpse right into a expertise that sits a number of layers beneath the robotic itself, however which can in the end play an necessary function within the improvement of extra succesful autonomous machines and embodied AI techniques.
Interview with Nils Könne and Christian Kreil
Robotics & Automation Information: Might you introduce yourselves and clarify what Digid does?
Nils Könne: I joined Digid in 2020 and now function head of product. My function entails creating our nanosensor expertise and dealing with clients to grasp the way it can clear up real-world challenges.
Christian Kreil: I lead enterprise improvement. Digid was based in 2019 and initially targeted on BioMEMS functions. Right this moment we work with clients throughout robotics, wearables, human-machine interfaces, industrial functions, medical units, and AI infrastructure.
R&AN: What makes Digid’s nanosensors totally different from typical sensors?
NK: The most important distinction is dimension. We regularly say the second-smallest sensor on this planet is 4 orders of magnitude bigger than ours.
As a result of our sensors are so small, we are able to place full sensing buildings straight onto a buyer’s machine and create dense arrays that measure power, pressure, and temperature concurrently.
In sensible phrases, this implies we are able to accumulate info that was beforehand tough or inconceivable to acquire. We will additionally combine sensors into merchandise with out requiring main design modifications as a result of the sensors add only some micrometers to the floor.
For a lot of clients, the query is now not whether or not our sensors are smaller, however whether or not the measurement was attainable in any respect earlier than our expertise grew to become accessible.
R&AN: Humanoid robotics is attracting monumental funding. How can nanosensors contribute?
NK: Tactile sensing stays one of many main challenges in robotics.
Our sensors can be utilized to create robotic pores and skin and extremely delicate fingertips with sensor densities far past these present in human pores and skin. The purpose just isn’t essentially to match each organic operate, however to offer robots with a lot richer details about the objects and environments they work together with.
Right this moment’s robots rely closely on imaginative and prescient. However when a robotic picks up an object, it additionally must understand how a lot power it’s making use of, whether or not the thing is slipping, and the way situations are altering throughout the interplay.
The higher a robotic can understand the bodily world, the extra succesful it turns into.
R&AN: If robots are coated in hundreds of sensors, received’t that create an information drawback?
NK: Completely.
The answer is to not ship each sensor studying to the cloud. Knowledge must be processed regionally in order that solely related info reaches higher-level techniques.
For instance, when a robotic picks up a cup, the necessary query is whether or not the grip is safe, not the uncooked output from each sensor. We imagine native preprocessing can be crucial as tactile sensing turns into extra refined.
R&AN: What varieties of business functions are producing essentially the most curiosity?
CK: We work with everybody from startups to Fortune 500 corporations.
The strongest curiosity at present comes from robotics, wearables, medical units, industrial sensing, and human-machine interfaces. A typical theme is that clients want sensing capabilities in locations the place conventional sensors are too massive or too costly to deploy.
NK: Since exhibiting at CES, we’ve seen rising curiosity from massive corporations searching for power and temperature measurements in places the place current applied sciences merely don’t match.
R&AN: How has the corporate advanced because it was based?
NK: We started with a powerful deal with diagnostics and BioMEMS expertise.
Over the previous two years we’ve expanded into industrial sensing and robotics after realizing there was vital demand for terribly small power and temperature sensors.
Right this moment we’ve round 30 workers, are producing income, and are working towards turning into totally self-sustaining.
One venture allowed us to cut back a buyer’s pressure-sensing system from roughly 30 parts down to 6, lowering complexity, manufacturing effort, and price.
R&AN: You additionally talked about AI infrastructure. How does that match into the image?
CK: AI information facilities generate monumental quantities of warmth, and managing that warmth effectively is turning into more and more necessary.
Our sensors will be built-in very near heat-generating parts, offering extremely localized temperature measurements and permitting operators to grasp thermal conduct in a lot larger element.
NK: Whether or not it’s robotics or AI infrastructure, the precept is similar: higher sensing results in higher selections.
R&AN: Trying forward, what are the most important alternatives for Digid?
CK: We imagine we’re nonetheless firstly. Prospects frequently method us with functions we had by no means thought-about.
Demand for extra correct info from the bodily world continues to develop throughout robotics, healthcare, industrial automation, wearables, and AI infrastructure.
NK: Our purpose is to change into the usual answer every time clients want sensing capabilities at a scale that conventional applied sciences can’t obtain.
As merchandise change into smaller, smarter, and extra related, we imagine demand for nanoscale sensing will proceed to extend.
