Your gateway into the world of robotics

N-aiR
Industry and automation
News / Robot vision

ABB: teaching robots to see with AI Robot Trainer

ABB completes OmniCore EyeMotion with AI Robot Trainer to facilitate the transition from workshop images to robot movements. The challenge: adapting a vision application to variations in parts and production.

From image to gesture

According to ABB, the solution brings together image acquisition, model training, deployment and continuous improvement. It combines real and synthetic data and works with automatic trajectory planning. The announced applications include part picking, sorting and quality control.

UNDERSTANDING ROBOT VISION

  1. Acquire images representative of the workshop.
  2. Train and then evaluate the perception model.
  3. Locate the part and prepare the movement.
  4. Execute the pick, measure failures and improve.
Original NewsRobotique educational diagram: simplified steps, without reproduction of the ABB interface.

Announced gains to be verified in the field

The manufacturer announces up to 90% reduction in commissioning time and up to 50% cycle time with automatic planning. These commercial maximums do not constitute guaranteed results for every workshop. The press release announces a presentation at the CIIF in Shanghai, from October 12 to 16, 2026.

NewsRobotique’s technical perspective

A useful evaluation will need to compare the success rate of picks, the complete cycle time, and human interventions, on parts and lighting conditions not present in training. Good recognition is not enough: the grip must also remain mechanically stable. These criteria are our analysis framework, not measured performance of the product.

Source: official ABB press release, consulted on October 19, 2026. No independent testing was carried out by NewsRobotique.

FREN