Founded only nine months ago, the French start-up UMA has now unveiled North Star, the design of its future humanoid robot, as well as Version 0, its first functional prototype. Its co-founder and CEO, Rémi Cadène, discusses the company’s strategy.
JDN. Can you present UMA and its ambition to us?
Rémy Cadène. UMA stands for Universal Mechanical Assistant. Our ambition is to develop general purpose robots capable of automating a very wide range of tasks. These are mobile robots, on wheels or on legs depending on use, equipped with hands to manipulate objects and tools, as well as artificial intelligence capable of learning and adapting to its environment. By the end of the year, we plan to deploy our first prototypes in logistics centers to meet the automation needs of the sector and validate our technology, which combines artificial intelligence, software and hardware.
Have you already established partnerships with logistics players?
We currently collaborate with around fifty companies. We have already carried out more than fifteen visits to factories and logistics centers in order to identify the tasks on which our robots can be deployed most effectively.
Why did you choose logistics as your first market?
Logistics is a great place to start. The environments there are relatively predictable while still relying heavily on human labor. Many tasks are repetitive, arduous and unrewarding, while requiring a certain ability to adapt. In certain positions, the turnover rate can reach 60% after just one year. There is therefore a real need for robots capable of assisting human teams. Furthermore, a single logistics site can accommodate several hundred robots, which allows rapid deployment on a large scale. The second stage of our roadmap will concern the assembly of lightweight products. These operations require more dexterity, in particular to manipulate tools and carry out more complex assembly sequences. In the longer term, we are also targeting the general public market with robots intended for individuals.
You have unveiled North Star, your future humanoid robot. Can you introduce it to us?
We chose a humanoid morphology because it allows the robot to interact naturally with a world designed for humans. In the same way that large language models have made artificial intelligence accessible to everyone, we want our robots to be able to evolve in the same environments as us and use the same tools. North Star is therefore equipped with legs to move around and hands capable of manipulating a wide variety of objects. The robot wears technical clothing and a luminous head that allows it to be immediately identified as a robot, even from a distance. We worked on calm, natural movements to inspire confidence while reflecting precision.
You also unveiled Version 0, your first working prototype.
It was designed and developed in our Parisian laboratory, and already integrates all the artificial intelligence necessary to control the robot. He knows how to walk, maintain his balance and react to disturbances. This prototype allows us to validate the essential technological building blocks of our platform before building the final version of North Star.
You insist on safety. How does this translate into design?
From the design stage, we retained three main principles. The first is a lightweight design to limit risks in the event of interaction with a human. The second is based on a redundant architecture, inspired by aeronautics. We multiply independent systems, both hardware and software, so that they control each other. Finally, we are developing our own world model, a model capable of predicting the consequences of actions planned by the main artificial intelligence. If a movement is not judged sufficiently safe, precise or if its outcome is too uncertain, it is simply not performed.
How do you envisage the scaling up and industrialization of your robots?
Our priority is to move forward as quickly as possible while demonstrating our capabilities in artificial intelligence applied to robotics. We have already demonstrated our ability to develop our AI, design our own hardware and define a coherent industrial strategy. Our objective is now to validate this approach by the end of the year by deploying our first prototypes to pilot customers. These test phases will last approximately four weeks and will be based on performance indicators defined with our partners. In the longer term, we believe that sufficiently dexterous robots will be able to participate themselves in theassembly of subsequent generations of robots.
The automotive industry appears to have many synergies with humanoid robotics, which can enable production at scale. Is this an avenue you are exploring?
Yes. The automotive industry is undergoing transformation and is seeking new outlets for industrial know-how developed over several decades. Partnerships with this sector could accelerate the development and industrialization of our robots.
But our priority today is elsewhere. As a very young company, our main challenge is to demonstrate the potential of our technology and attract the best global talent. We also want to contribute to structuring a true European ecosystem of humanoid robotics. It is with this in mind that we created Neon Noir, a place dedicated to hackathons, conferences and meetings around artificial intelligence, hardware and robotics.
What will be your business model: direct sales or Robot-as-a-Service?
We chose a Robot-as-a-Service model. In reality, what we offer is an automation solution. Our customers will rent the robots via a monthly subscription, possibly supplemented by an installation cost as needed. We have not yet communicated on prices.
You also develop your own world model. In your opinion, is this a key step to accelerate progress in robotics?
In our opinion, there are two main ways of learning. The first consists of observing a task and then reproducing it until you master it. This is the principle of our Real-Time Learning technology, which allows the robot to quickly acquire new skills through practice. The second consists of constantly anticipating the consequences of one’s actions. This is precisely the role of the world model. It allows the robot to simulate several scenarios, evaluate the consequences, then choose the safest and most effective action. This capability improves both safety and learning speed, two essential elements for developing robots capable of evolving in complex environments, including ultimately in homes.
What will be UMA’s next steps?
By the end of the year, we want to deploy our first robots with pilot customers in order to validate three essential elements: the technology, the product and its economic model. Then, we will continue our development by accelerating the growth of the company and continuing to recruit the best talents. This is, in our opinion, the main factor for success.
