inbolt, making factories autonomous

Rudy Cohen, Albane Dersy and Louis Dumas met in 2019 at the X-HEC Entrepreneurs Master's degree after studying at ENS Paris Saclay, HEC and Supaero respectively. "As part of the course, we had to produce a business plan in six weeks for a dummy company. We really got into it, visiting factories and contacting automotive and aerospace manufacturers to find out what their problems were. We realised that their difficulties were linked to reindustrialisation. A lot of new sites are being built, requiring a large number of operators. But the job is tough, and skilled employees are becoming increasingly rare. Lacking personnel, factories have no choice but to automate their processes with tools that are as flexible as operators," explains Rudy Cohen.

While they were still students, the trio decided to set up their own company, inbolt, to address the shortages being experienced by manufacturers. They did so by developing AI vision solutions to be installed on industrial robotic arms, making them autonomous. The start-up designed its first prototypes at the Drahi X-Novation Center at the École Polytechnique. In 2019, it won the Jean-Louis Gerondeau - Safran award, which gave it 20,000 euros. "It was this award that really launched us and took us from a student project to a real company. This sum, albeit modest, enabled us to hire an intern and buy equipment. The award also put us on the map," adds the co-founder. 

Spotted by the American investment fund SOSV (HAX), inbolt spent two months at its incubator in Shenzhen, China. After this unique experience, the trio spent two years at the Parisian incubator, Agoranov. In 2020, it won the I-lab innovation competition, which awarded it a €250,000 grant. Finally, in the spring of 2022, the start-up raised three million euros in seed capital from international investment funds and Laurent Dassault.
 

Automating robots

The start-up, inbolt, which now has around twenty employees, is dedicated to making robotic arms autonomous using its three-dimensional vision. "We install 3D cameras directly onto the robots already present in factories or those waiting to be installed. We give them vision so that they can adapt to their environment and become more flexible," adds Rudy Cohen. In the past, automating processes meant constraining the environment, sometimes at a very high price. With its solution, inbolt sells the cameras and provides a license for accessing its software.

The robot equipped with the vision system is completely autonomous. "Our vision systems are 99.99% reliable, which avoids production line stoppages, which are very costly for manufacturers," explains the co-founder. At the moment, the only limit is the camera's resolution: if it's poor then the robot can't work on very small parts. "To counter this problem, we adapt and use different types of camera. This enables us to guide arms to accuracies of 0.01 mm," explains Rudy Cohen.
 

Hybrid technology

The young company's expertise lies in its ability to develop software that covers a wide range of use cases. "The question is how to use the robot's vision to modify its trajectory in real time. We are faced with two types of case: the first is a dynamic environment, i.e. where the part to be handled is constantly moving, as on an assembly line for example. We can automate these jobs, because our software works in real time. The second scenario is that of a static environment. Our technology enables the robot to work on the component regardless of where it is in its field of vision. Once the algorithm has been trained, it is automatically tested by programming a test trajectory for the robot. This is how we validate the algorithm's performance," explains Rudy Cohen.

inbolt's technology can be installed on all major brands of robotic arms used in industry. To deduce the position of the part, the software must first study the three-dimensional CAD assembly plans. In 15 to 20 minutes, the algorithm is trained and will track the component's position.

For its technology, the start-up deploys an approach that uses deep learning algorithms with logical constraints. This hybrid artificial intelligence combines reasoning and data to design safe, high-performance systems. "We chose this solution because using deep learning alone would have given us a much less reliable probabilistic output model," adds Rudy Cohen.
 

Capturing new markets

The start-up is targeting the manufacturing, automotive and electronics markets. It has already deployed its technology in some fifteen factories across Europe, in France, Italy, Hungary and Sweden, particularly in the automotive sector. "We haven't set ourselves any geographical constraints. Our customers have installations to be automated all over the world and we follow them. We're planning to open an office in the United States by the end of the year," explains the co-founder.

While inbolt is able to install its vision systems on most arms, rllr wants to conquer the entire market by adapting to all brands, thus expanding the frontiers of its technology. The company plans to supply its solution to the logistics and electronics industries, research teams and the medical robotics sector. "When we perform an operation, the patient's skin moves because they're breathing. The robot must therefore be able to anticipate this movement. Medical applications are a very long-term prospect, but one that underlines the technology's huge potential. We haven't yet set up a product roadmap for this kind of use case. Our ultimate goal is for anyone to be able to use our technology," explains Rudy Cohen. " These are huge markets with very strong potential," he continues, "inbolt works hand in hand with manufacturers and integrators, who are targeting the same markets." "They are our partners, so it's a win-win for both parties. They help us to find new customers."

In order to finance these future projects, the start-up is finalising a fundraising campaign, the amount of which is still being kept secret, so as to become profitable in the coming years and free itself from these financing rounds. This will enable inbolt to recruit fifteen people in 2024 and twenty the following year.