Micropsi raises $34M to retrain industrial robots using human demonstrations

As the pandemic places an increasing strain on the supply chain, manufacturers and warehouse operators are piloting — or upping their existing investments in — robotics technologies. According to a 2020 Automation World survey, 44.9% of operators reported that their assembly and manufacturing facilities currently use robots as an “integral part of their [workflows].” Global X’s Jay Jacobs told CNBC in a recent interview that industrial robots could grow from 16 billion to 37 billion over the next 10 years, with 2022 being the key inflection point.

But there’s many hurdles on the road to deploying robotics, as a 2021 study from MIT highlights. Among them are the high costs of integrating robot hardware with the existing supply chain and the absence of standardization, as well as inflexibility; current robotics technologies can’t always be quickly repurposed to perform new tasks. That’s perhaps why only 23% of the companies responding to Automation World’s poll said that they planned to add robots to their operations in the coming months.

New startups claim to simplify the implementation process with products that address some of the more challenging technical aspects of robotics. One is Micropsi Industries, a Berlin, Germany-based company that trains industrial robots to perform actions through human demonstration. Reflecting investors’ enthusiasm for the robotics sector, Micropsi today announced that it raised €30 million (roughly $34.31 million) at an approximately $93 million valuation, bringing its total raised to around $45 million.

Learning from demonstration

Founded in 2014, Micropsi — which has an office in Brooklyn — develops systems for controlling industrial robots. Its first product, MIRAI, is designed to be attached to existing robotic arms, allowing them to “perceive [their] workspace” through cameras and correct their movements as they perform tasks.

“We and everybody around us — that’s us, who provide the AI, but also the makers of robots and tools, as well as the system integrators — are hard at work lowering the barriers of entry for robotics,” CEO Ronnie Vuine told VentureBeat via email. “Robots are today mainly deployed in large corporations. They can afford to build the know-how, they can afford to pay the system integrators, they can afford to fail and learn. Small- and medium-sized businesses need to be much more cautious, and yet the potential is so much larger there. Robots are generic machines, AIs that can learn what to do are generic machines, so this combination is a great fit for a fragmented landscape of small manufacturers making a million different things.”

Micropsi claims that, using AI, MIRAI can generate robot movements in real time — dealing with variations in position, color, and lighting conditions. It can also be trained and retrained for various process steps, the company says, including detecting leaks in machinery, screwing screws and inserting cables into products, and sorting objects on the assembly line.

“Data is generated by giving demonstrations, the rest is done by MIRAI in the background — it takes [about] 30 minutes of demonstrations and between one and three hours of number-crunching time in the cloud to create a new skill from scratch,” Vuine explained. “Users learn how to create good datasets by iterating their skills: Record some data, see the robot perform, add some more data to address weaknesses, and after three or four iterations, a very robots skill has been created. No one at Micropsi Industries needs to understand the use case, and no one on the customer side needs to understand the machine learning.”

Micropsi’s approach is a form of “imitation learning,” which aims to mimic human behavior in a given task. In imitation learning, a system is trained to perform a task from demonstrations be learning a “mapping” between observations and actions.

For one customer, Siemens Energy, Micropsi — which claims to have sold 50 systems to date to about 30 customers — automated a process step in the refurbishment of gas turbine vanes. For ZF, a German tech company, Micropsi helped to program robots to pick up metal rings from a crate and place them on a conveyor belt.

Challenges ahead

While the pandemic hasn’t accelerated robotics adoption in the way that some analysts predicted it would — 75.6% of companies responding to the Automation World survey said that the pandemic didn’t drive any purchase of new robots within their facilities — that hasn’t stopped investors from pouring capital into the industry. From March 2020 to March 2021, venture firms invested $6.3 billion into robotics companies, up nearly 50% from the $4.3 billion they invested in the 12-month period a year earlier, according to a Forbes analysis of Pitchbook data.

Beyond Micropsi, there’s automation-as-a-service startup Rapid Robotics, which provides out-of-the-box automation solutions for manufacturing companies. In a similar category is Symbio Robotics, a startup developing technology that automates repetitive assembly tasks. Other direct — and indirect — Microspi competitors include industrial robotics company Elementary Robotics, robotics deployment firm SVT Robotics, and robotics safety system developer Fort Robotics.

Vuine asserts that MIRAI has an advantage over rivals in that it’s a “proven” and “independent” technology that’s “working 24/7 in the factories” of customers. “Most customers will start with a single system and then add more as they validate its reliability and performance in a real-life production setting,” he added. “Our product, MIRAI, isn’t just a tool for setting up a production line — it becomes part of he factory and controls robots 24/7. [C]ustomers need us to be as reliable as any other industrial equipment they use.”

Metaplanet led 45-employee Micropsi’s latest funding round with participation from Vsquared, Ahren Innovation Capital, and Project A. Vuine says it’ll be put toward expanding Micropsi’s operations in the U.S., ramping up sales efforts, expanding to more robotics platforms, and “developing features that will use AI to reduce integration cost[s].”