Embedded Vision Beyond Autonomous robots

Embedded vision is widely recognized in modern mobility applications such as autonomous driving, advanced driver-assistance systems (ADAS), AMRs, AGVs, drones, and humanoid robots. Besides these relatively new application areas, there is further potential in manufacturing.

The traditionally conservative industrial sector is slowly embracing new technologies, and AI and vision are no exception. This creates significant opportunities for embedded vision, empowered by AI and robotics, on the factory floor. However, some differences from more mature markets need to be taken into consideration.

At ImagingNext 2026, Maxim Soroka, AgileSense Team Lead at Agile Robots, presents “Embedded Vision Beyond Autonomous Robots” a session about AI, vision, and robotics in industrial applications.

About Maxim Soroka

Maxim Soroka is AgileSense Team Lead at Agile Robots. He holds a degree in Electrical Engineering from Saint Petersburg Electrotechnical University (LETI) and received his business education from the Stockholm School of Economics. Since 1995, he has led multinational engineering teams on international projects across Europe, North America, China, and India. His primary focus is the integration of AI, vision, and robotics on the factory floor for technologically demanding industrial applications.

Why this matters for industrial robotics

Robotics, as the most advanced part of industrial automation, can serve as a reference point. According to the International Federation of Robotics, more than 500,000 new industrial robots have been installed on factory floors worldwide each year over the past four year. The total number of robots in operation has reached five million.

There are no reliable statistics, but in contrast to AGVs, which are widely used for in-house logistics, only a small proportion of installed industrial robots are equipped with vision. Even fewer are equipped with both vision systems and AI. As a result, much of this advanced robotic hardware still operates in a fixed, “blind” mode.

What are the barriers to more flexible, perception-based, and widespread use of robotics? Besides cost and availability, several engineering challenges must be addressed.

AI, vision, and robotics need to be integrated through a relatively open software framework, similar to those developed for autonomous vehicles, SLAM, and mapping. Existing NVIDIA foundation models could be reused for industrial robot functions such as object grasping and robot-arm guidance, as well as for simulation and offline robot programming.

Advanced time synchronization between multiple cameras, light sources, and robot motion is also required. An embedded vision camera system should provide image-processing capabilities comparable to those of a standalone camera, while offering tight integration with the computational platform, compactness, and cost-effectiveness.

A practical example of a relatively simple industrial application will illustrate how the combination of AI, vision, and robotics can lower the barriers to adoption.

What you’ll take away

• The potential of embedded vision beyond autonomous mobility and humanoid robots.
• How open software frameworks and foundation models could support industrial robotic applications.
• Why advanced time synchronization between cameras and robot motion is essential.
• How tightly integrated vision and computational platforms can lower barriers to industrial adoption.