ABB and Psyonic Partner to Develop Dexterous Robots

Robotics manufacturer ABB Robotics has joined forces with US bionic technology company Psyonic to create industrial robots with human-like dexterity and accelerate the deployment of ‘physical AI’ in factories and warehouses. Their goal? To solve a key challenge in robotics: getting robots to learn how to pick up, grasp and manipulate objects like humans do, intuitively.

How? By training robots to do so with data collected by a prosthetic hand designed by Psyonic. Called the Ability Hand, this artificial limb is equipped with sensors that detect touch, vibrations and flexing fingers. Humans use it to carry out tasks and in doing so, record detailed information about how they interact with objects, allowing the data to be used to train robotic applications to perform the same tasks. “Dexterous manipulation is both a hardware problem and a data problem,” explained Aadeel Akhtar, founder and CEO of Psyonic. “By enabling people to use the Ability Hand and subsequently allowing robots to use the Ability Hand, we can generate high-fidelity data from the physical world… which can be used to train robotic applications.”

1From pre-programmed arms to physical AI

Physical AI refers to machines that sense, think and act in the physical world, allowing them to adapt to their environment rather than execute a set program. “One of the biggest challenges in industrial robotics is replicating human dexterity and the intuitive knowledge of how to manipulate various objects, but it’s a critical requirement for building fully autonomous and flexible robots,” said Marc Segura, president of ABB Robotics. They hope to incorporate the technology into ABB’s GoFa collaborative robot and put it to the test in the automotive, aerospace, packaging, logistics and life sciences sectors. In a market where generative AI models were trained on internet-scale text and image data, the need for physical data is clear, but so is the challenge of gathering it.

2Why this matters for you

In recent years, AI has made itself known through words on our screens, whether it’s chatting to us or helping us write code or create images. This collaboration marks an important step forward into the future of AI, when it starts to use its hands. Whereas previously, the problem has been that we don’t have a smart enough model, now we need more data on the messy physical world. And the solution here, of having humans wear the same hand as the robots so they can learn from them, is pretty cool. In my humble opinion, physical AI is currently less advanced and slower than its chat window counterpart, and one collaboration is not going to transform a factory overnight. But if we can get robots reliable hands in places like warehouses, packing facilities and assembly lines, where we wonder whether they could ever work, then improvements in dexterity are what will make that happen.