Bringing new intelligence to industrial robots
The more than 4 million industrial robots currently deployed in factories around the world are not that smart. They do not walk or talk, or pack artificial general intelligence into a humanoid form factor. Even so, they are the indispensable workhorses of advanced manufacturing. Across a broad range of industries, they have proven their durability, reliability, and productivity—especially with precision work. Yet much of their potential remains unrealized, stifled by the technical complexity of their integration and operation. “The standard operating procedure at a factory today is for a highly trained operator to maintain constant oversight over a robot that has no intelligence of its own,” says Asad Tirmizi, CEO and co-founder of T-Robotics. “To make it work, they have to program it. And if something goes wrong, they have to fix it. As a result, even the most sophisticated robots are vastly underutilized. We need robots that can do more with less.”
T-robotics, a startup founded by Asad Tirmizi along with CTO Lars Tingelstad, is valorizing industrial robots with a revolutionary software platform. Supplanting fifty years of legacy programming techniques, they have leveraged the power and flexibility of AI to develop operating software that brings new capabilities to both robots and the humans who manage them. Called ActGPT, its application-level innovation unlocks two fundamental advantages from the existing control system of any brand of industrial robot. On the human interface side, ActGPT incorporates natural language instructions, no-code programming, and a flexible digital twin, to make commissioning a robotic manufacturing line more accessible and efficient than ever before. On the control side, its pioneering AI-driven model is pre-trained with an expanding library of known skills, augmented by the rapid problem-solving adaptability of neural networks. The combination makes even older robots dramatically more performant. “ActGPT provides a step-change in a robot’s capability, enabling functionality that has been unrealized for decades,” says Tirmizi. “Existing industrial robots are vastly underutilized,” adds Tingelstad. “Our platform allows any human to program a robot, and makes any robot more robust and intelligent.”
T-robotics’ technology has its roots in Tirmizi and Tingelstad’s doctoral collaboration, undertaken at the University of Siena and the Norwegian University of Science and Technology. At its heart is a fundamental re-thinking of how industrial robots are programmed. Since their earliest implementations in the 1960s, they have primarily relied upon “point to point” programming, in which an engineer describes where the robot has to be to complete each step of its task, and the robot draws the motion between those points. But point to point is time-consuming to implement and brittle to operate. “We knew that if we wanted to bring intelligence to the application layer, we had to change the mathematics that are running the robot,” says Tirmizi.
The ActGPT platform is designed to address both the challenge of programming new manufacturing cells, as well as making existing ones more performant. Its core innovation is a model that can see, touch, understand and act. This proprietary T-robotics technology allows the robot to make sense of the visuals of the task; to understand its haptics; to incorporate language from a human operator; and to rely on an innovative combination of known skills and AI-based problem-solving to efficiently work through any actions. “It is always the case in robotics that no matter how much you model the world, you will encounter unknown situations,” explains Tirmizi. ActGPT’s next-generation architecture has the ability to quickly and accurately overcome any process malfunctions, while also improving its skills over time, and sharing that data across the platform. “Our software is trained on real world industrial data, and continually learns from real world industrial tasks,” says Tingelstad.
T-robotics is partnering with major robot manufacturers and integrators to bring the benefits of AI to the millions of robot arms already in use. The commercial mission is a product that can be dropped-in to the full breadth of industrial robot operations, serving as an end-to-end solution not only for existing arms, but for manufacturing execution systems more broadly. “There will be a day when we arrive at humanoid robots, but single-purpose robots will always be the workhorses of the industry—and that’s the opportunity we can empower with our software,” says Tirmizi. “Our aim is to be not only a platform for programming robots, but the platform for driving industrial automation.”