91ÊÓƵÍøÕ¾ÎÛ

Collaborative robots are becoming increasingly common in manufacturing facilities, but there are still more opportunities for automation. With the United States ranking as the third-largest market for cobot maker Universal Robots globally, some of that opportunity is domestic. According to Jonathan Sbert, Universal Robots’ VP of Sales for the Americas, this is because the US market has more small- and medium-sized manufacturers with a heavier focus on R&D and small-batch production, making automation more difficult. “The U.S. market is just so fragmented and has so many small operations,” he says. “That's what makes the economy so great, but it's hard.” But new technologies such as AI and vision systems are increasing adoption by making cobots easier to implement for the small- and medium-sized manufacturers that dominate the US market.

Such technologies were the highlight of Universal Robots’ Collaborate 2025 event, which took place January 28 in Novi, Michigan. This was the first U.S. edition of the event, which has previously taken place in other regions. It featured presentations from Universal Robots representatives and partners, as well as technology displays by companies within the Universal Robots ecosystem.

Accessible AI

AI has the potential to break down some of the barriers that prevent small and medium-sized manufacturers from automating. According to Bernd Raithel, the director of product management and marketing for the Factory Automation business unit at Siemens USA and a keynote speaker at the event, one of the biggest hurdles to automation is that efforts eventually plateau. “You can gain so much efficiency, you can do so many improvements, but then maybe you could do more, but the ROI just doesn't justify it anymore,” he said. But AI will change this calculation. “With much less effort, you can solve problems you weren't able to solve before,” he continued. “There’s basically another leap, and you can improve your overall equipment efficiency.” This could take many forms, including reducing downtime through AI-aided predictive maintenance, AI-powered inspection systems that detect defects, and robots that are more flexible and easier to deploy thanks to AI programming. Siemens is working toward this with products such its Robot Pick AI software for picking unknown parts without programming. AI is also a priority for Universal Robots — it has partnered with AI GPU maker Nvidia on AI Accelerator, a hardware and software platform designed to enable Universal Robots users to implement AI.  

A Vision of the Future

Vision systems also emerged as a theme at the event. Historically, vision systems have been difficult to implement, requiring expensive cameras and lengthy training processes. But as 3D vision systems become more common, combined with the use of AI to train them, users can more easily automate complex tasks such as bin-picking, inspection and assembly.

Cambrian’s system includes a camera mounted on the robot’s wrist, keeping it in close proximity to the parts the robot is handling. Users upload a CAD file of the part to the system and define the pick points. Then the system uses simulated data to train an AI model to detect pick points on the part, eliminating the need to program or train the robot for bin picking, kitting or assembly tasks. Keyence was demonstrating a bin-picking task with four cameras mounted above a robot. The cameras provide a wide field of view, as well as information on the depth of the bin. Together, they take 136 images after every pick, which are compiled into a 3D image. This gives the system insight into “blind spots” such as shadows and glares, so it can locate all parts in the bin. Keyence’s path-planning packages enable the robot to pick parts without getting stuck. And if the system can’t find a part that’s properly oriented to be picked up, the robot can move the parts around until it finds a part that can. Rob Halasz, Keyence sales director for 3D vision-guided robotics, said the system can be implemented within half a day. InBolt’s path planning system enables a wrist-mounted live 3D camera to follow parts in real-time, even as they move, for example, down a conveyor belt. The company’s co-founder and CEO, Albane Dersy, pointed out that many manufacturing applications require a “constrained environment” where variables are kept as constant as possible, often requiring the environment to be adapted accordingly. However, vision systems enable manufacturing processes to adapt as the environment around them changes, opening up new possibilities for manufacturing.

Seeing the Benefits

According to Sbert, the best way for shops to automate is to start with one cobot. “We're seeing established customers growing, and they're finding more and more applications for the robot,” he says. “It's not necessarily doing the same thing multiple times. It's doing totally different things.”