Automate 2026

Most factory vision systems force a choice: cloud intelligence with full management visibility, or local inference with data sovereignty. iENSO and Ambarella demonstrated at Booth #13018 that this is a false choice.

The platform combines iENSO vision AI cameras with the Ambarella N1-655 CVflow Neural Vector Processor and BrinqAI AcuSight software to deliver distributed vision AI inspection across multiple camera nodes, centralized production dashboards and alert management, and edge VLM-powered operator safety monitoring and co-pilot capabilities. All fully on-premise. All without a packet leaving the facility.

Depth perception built into the eyes of a robot head. Infineon's Real3 ToF sensor was embedded directly in the eyes of a robot head demo, displaying a live VGA image alongside a real-time heat map showing the precise distance of every surrounding object in the field of view.

This is the depth sensing layer that enables obstacle avoidance and SLAM navigation at a fraction of the cost and size of stereo camera or structured light alternatives.

Computer vision, robotics control, on-device Speech-to-Text, and large language model inference running simultaneously on a single compact platform, with no cloud connection. Silex demonstrated their EP-200Q-EVK running a live robotic arm use case alongside on-device STT and LLM capabilities.

This is what it looks like when the edge gets smart enough to replace the cloud.

Object detection, pose estimation, face recognition, and CLIP-based action recognition running simultaneously on a single M.2 module at 25 TOPS and under 5W. The DX-M1 NPU handles all AI inference independently. The host simply pipes data to the card, demonstrating just how little supporting infrastructure a real-world edge AI deployment actually needs.

STREAL was one of the most talked-about demos at the booth, generating significant interest from engineers who are accustomed to traditional foil strain gauges. The difference is significant: conventional strain gauges are analog, require external Wheatstone bridge circuits, instrumentation amplifiers, and separate ADCs, and are sensitive to temperature drift and installation variability.

STREAL integrates the entire signal chain into a single silicon chip that outputs clean digital data over SPI, placed directly inside joints, links, grippers, and AMR frames where mechanical events actually occur, not just at the wrist where conventional sensors miss most of them.