ABB Brings Legitimacy to Emerging AI in Robotics Practices with OmniCore EyeMotion

The adaptive Artificial Intelligence (AI) used to power the next generation of robotics will yield significant returns—and meet the acceptance criteria of end adopters and System Integrators (SIs) alike—when deployed on existing, well-understood robot form factors, chiefly collaborative and industrial robots.

A critical challenge in advanced robotics AI, particularly with reinforcement learning, is bridging the simulation-to-real gap, ensuring that policies learned in simulation translate effectively to dynamic real-world conditions. Few solutions master this alignment. We call this capability Dynamic Policy Adjustment (DPA), allowing robots to robustly track moving objects and adapt complex pick-and-place policies on the fly to bridge the gap between training data and real-world variables. Pioneering this methodology are innovative Software-as-a-Service (SaaS) startups such as Micropsi, Inbolt, Apera AI, Cambrian AI, V-Sim, and Summer Robotics, which offer solutions that retrofit existing robot deployments to extend their capabilities for more intricate tasks. These companies present impressive, cutting-edge solutions, yet their complexity often necessitates significant initial and ongoing support from their specialized staff. The intricate nature of these products means they are typically manageable only by their developers, hindering their hand-off to SIs or on-site engineers, impeding widespread scalability.

The industrial manufacturing sector, especially automotive, has become a key proving ground for such innovators. Manufacturing engineers often possess higher technical proficiencies, requiring less ongoing support, which enables these pioneering firms to refine and scale their DPA offerings more effectively. This contrasts sharply with the logistics and warehousing sectors, where the technical support burden for such complex AI solutions remains a significant barrier to adoption. Recognizing this critical market need, established players like YASKAWA, KUKA, and FANUC have also begun integrating AI augmentation into their robot controllers (with some significant launches at events like June 2025’s Automatica), signifying a broader industry movement toward embedded, adaptive intelligence in robotics. ABB's OmniCore EyeMotion now formally joins this cohort, aiming to deliver robust DPA directly within its established ecosystem.

The entry of a major player like ABB is a significant boon for the standing of AI in robotics (such technologies are called Embodied AI or Physical AI depending on the vendor, region, or investor). To date, the promise of adaptive AI in robotics has been tempered by deployment challenges. While agile SaaS firms have showcased impressive capabilities, their business model often relies on deep engagement and continuous support, hindering their ability to scale solutions through established channels. SIs, which are crucial for mass adoption, require robust, self-serviceable tools that they can deploy, maintain, and support independently. The current level of product maturity from many pure-play AI vendors means they are effectively selling a service, not a fully productized solution, which creates friction in a market accustomed to (relatively) turnkey or easily integrated components.

While DPA technology opens opportunities for certain complex, dynamic, or dexterous niche applications within existing brownfield automation, the overall Total Addressable Market (TAM) for retrofitting remains relatively small. Many established manufacturing processes are sufficiently simple, static, and controlled, making traditional, rigid programming perfectly adequate. Furthermore, many existing lines and cells have been optimally designed for specific tasks, where adding AI offers no further advantages. Consequently, many innovators are targeting under-automated markets requiring heterogeneous dexterous manipulation, such as life sciences, recycling, and niche manufacturing industries. However, the largest greenfield opportunity lies within warehousing and logistics. Here, vendors like OnRobot, Zebra, Solomon 3D, SICK, Keyence, Mech-Mind, and numerous Universal Robots partners offer intuitive automation, often without advanced DPA, capable of dexterous manipulation in controlled environments. Adoption of these products, despite their efficiencies, has been limited, primarily due to pervasive economic uncertainty and an ongoing aversion to significant Capital Expenditure (CAPEX) (or even Operational Expenditure (OPEX)) outlays, rather than technical shortcomings.

The advancements in DPA and the broader field of “embodied AI” are truly impressive, representing a significant leap in robot autonomy and capability, while aligning physical capabilities with simulated and reinforcement learned policies—a boon to NVIDIA’s overall robotics strategy. However, the main issue now shifts from technical feasibility to demonstrating a clear and viable Return on Investment (ROI) that can be used to seed self-sufficient support and deployment ecosystems. The next critical complexity lies in convincing SIs and end users to fully embrace and adopt these advanced AI solutions in their operations. The key to overcoming this adoption barrier will be a focus on usability, transparency, and simplicity—a seemingly paradoxical goal when dealing with complex algorithms for real-time robot control. Product strategies must abstract away the underlying AI complexities, providing intuitive interfaces and predictable performance (while maintaining generality!) that SIs and engineers can easily understand, deploy, and troubleshoot without deep AI expertise.

Adding to this challenge is the prevailing economic uncertainty, which acts as a significant dampener of new CAPEX and OPEX outlays. While adopters hesitate to pull the trigger on proven DPA solutions and other forms of “Physical AI,” the pervasive over-promising from certain highly visible, yet still nascent, sectors of embodied AI—particularly humanoid robotics—risks tarnishing the reputation of all advanced robotics technologies. This could severely damage the adoption prospects for genuinely impactful DPA solutions and other practical applications of physical AI. Therefore, innovators in this space, from SaaS providers to established robot manufacturers, must go beyond simply showcasing technical prowess. They must meticulously articulate and quantify the revenue potential, cost savings, and operational efficiencies that DPA and other physical AI applications bring, ensuring that SIs and end users can clearly see and validate a compelling business case. This demands robust case studies, clear metrics, and a concerted effort to build trust through reliable performance and accessible support, counteracting market skepticism and building a foundation for scalable growth across the entire robotics landscape.