Could AI Make a Robot ‘See’ Like a Human?

The AI-enabled MIRAI software could help a robot automate highly variable tasks.

Daphne Allen, Editor-in-Chief

June 12, 2024

2 Min Read
AI vision software for robots
MIRAI AI-enabled vision software is compatible with systems from Universal Robots (pictured above) as well as FANUC and KUKA Robots (pictured below).Micropsi

In a way, MIRAI gives a robot human-like skills. The AI-enabled vision software from Micropsi can adjust a robot’s movements in real time to enable it to pick up and manipulate items with varying features and placement. The software’s latest update, MIRAI 2, allows robots to handle more variances than ever before. 

“We are really trying to give a robot hand-eye coordination,” Matt Jones, VP of US sales and operations, told Design News while exhibiting at IME South 2024. “Robots that are controlled by vision need more than just robotic programming. Variances make it impossible to fully automate without vision control.” MIRAI is essentially add-on vision control software that communicates with a robot through an Ethernet cable. 

“Real-time vision control is a game changer and allows us to tackle variances,” Jones explained. “The real-time aspect is pretty unique.” 

For instance, Jones described how MIRAI could guide a robot during automotive seat assembly operations that utilize heavy torque tools. “Our system helps the robot find the bolt, which could be in varying positions, and ensures that the robot is in the right position for bolt torquing.”  

MIRAI enables a robot to learn through imitation, Jones said. Similar training approaches may have taken one to two days, he said, but “MIRAI can train the robot with just 20 minutes of video, which contains about 100,000 reference images for that instruction.” 

AI-Robot-Micropsi-MIRAI-iff-DIM-wide-angle[59].jpg

MIRAI also overcomes some of the typical challenges with vision systems, such as reflection. “A lot of traditional vision systems struggle with reflective parts,” Jones said. “The pictures can be so bad that the robot can’t proceed.” 

MIRAI 2 offers five new features that make it even easier to enable robots to reliably perform automation tasks with variance in position, shape, color, lighting, or background. It utilizes a new means of collecting training data, provides a new option to run on industrial-grade PCs, and allows users to share skills across a fleet of robots.  

In addition to offering MIRAI software to robot users, Micropsi can provide cameras and lighting for integration into a robotic system. MIRAI is compatible with Universal Robots, FANUC, and KUKA robots and can guide them through different types of tasks, including picking, insertion, screwdriving, and more.

AI-Robot-Micropsi-MIRAI-kuka-dimm-landscape[83].jpg

During IME South 2024, Micropsi demonstrated MIRAI-guided robotic cable plugging and wire harnessing. See how a robot can learn in real time in the below video. 

View post on X

About the Author

Daphne Allen

Editor-in-Chief, Design News

Daphne Allen is editor-in-chief of Design News. She previously served as editor-in-chief of MD+DI and of Pharmaceutical & Medical Packaging News and also served as an editor for Packaging Digest. Daphne has covered design, manufacturing, materials, packaging, labeling, and regulatory issues for more than 20 years. She has also presented on these topics in several webinars and conferences, most recently discussing design and engineering trends at MD&M West 2024 and leading an Industry ShopTalk discussion during the show on artificial intelligence. She will be moderating the upcoming webinar, Best Practices in Medical Device Engineering and will be leading an Automation Tour at Advanced Manufacturing Minneapolis. She will also be attending DesignCon and MD&M West 2025.

Daphne has previously participated in meetings of the IoPP Medical Device Packaging Technical Committee and served as a judge in awards programs held by The Tube Council and the Healthcare Compliance Packaging Council. She also received the Bert Moore Excellence in Journalism Award in the AIM Awards in 2012.

Follow Daphne on X at @daphneallen and reach her at [email protected].

Sign up for Design News newsletters

You May Also Like