Drives and Controls Magazine
Digital twins help cobots to build more cobots faster
Published:  30 May, 2023

The Taiwanese cobot (collaborative robot) manufacturer Techman has turned to digital simulation to help it cut inspection times on its cobot assembly line by around 20%.

Techman’s cobots are unusual in having built-in automated optical inspection (AOI) capabilities, using machine vision systems mounted on the robot arms, combined with graphics processors in the robot controllers. This helps the cobots to identify defects and to inspect the quality of the products they are working on.

“The distinctive features of Techman’s robots – compared to other robot brands – lie in their built-in vision system and AI inference engine,” explains Scott Huang, the company’s chief operations officer.

Techman is using this capability to examine the cobots that it builds on its own production lines. It uses its own cobots to assemble and inspect the cobots. The AOI cameras on the inspection arms can reach areas that fixed cameras could not, as well as using AI (artificial intelligence) to reveal any defects immediately.

But programming the movement of these robots can be time-consuming. Developers need to determine the precise positions of the arms, as well as the most efficient sequence of movements to capture possibly hundreds of images as quickly as possible. It can take several days to explore the tens of thousands of possibilities to determine an optimal sequence.

Techman has therefore built digital twins of the inspection robots – as well as of the products they are inspecting. Programming the robots using these simulations can be more than 70% faster than relying on manual programming of physical robots. Using an accurate 3D model of the product, the application can be developed as a digital twin before the real product is manufactured, saving more time.

To create its digital twins, Techman is using Isaac Sim – a robotics simulation application based on Nvidia’s Onniverse open development platform for building and operating industrial metaverse applications. The application’s optimisation tools allow Techman to explore a large number of program options running in parallel on Nvidia GPUs. The result is that the Taiwanese robot-maker has cut inspection cycle times by 20%, with every second saved helping to reduce the cost of the cobots.

Techman is using the vision systems built into its cobot arms to inspect its own cobots as they are being assembled

Techman is also using synthetic data to improve the quality of its inspections. Gathering and labelling real-world images of defects is costly and time-consuming, so the company has turned to Nvidia’s Omniverse Replicator framework to generate high-quality synthetic datasets rapidly. These perfectly labelled images are used to train AI models in the cloud, enhancing their performance “dramatically”.

Dozens of AI models can be run at the edge, efficiently and with low latency, while inspecting particularly complicated products, some of which take more than 40 models to scrutinise their different aspects.

•  Techman has a strategic alliance with Omron under which the Japanese automation giant supplies Techman cobots under its own brand.

Techman Robot:  Twitter  LinkedIn  Facebook

Nvidia Omniverse:  Twitter  LinkedIn