Mirai for cobots

Handling of variances through an AI based control

Artificial intelligence on a cobot to implement new applications and to deal with different variations in production

MIRAI from Micropsi Industries is a computer vision-based robot controller that enables cobots to deal with variances in position, shape, movement and changing lighting conditions. With the help of machine learning, MIRAI generates robot movements directly and in real time.

FAuDE Tec uses the MIRAI system for variances in assembly, quality testing, machine loading and contour tracking. The big advantage is that you can retrain for new tasks without much effort, since no fixed, preprogrammed movements, but skills are learned. No AI knowledge is required for learning, it is comparable to observing human movements in a few days.

Advantages of the MIRAI system on a cobot

  • Easy handling of variances, especially where it is impossible to implement the programming by hand.
  • Easy to use as robot movements can be trained in a few days.
  • Simple cobot programming skills are sufficient to implement a stable system
  • Flexibility in use, as cobots equipped with MIRAI can easily “learn” tasks.
  • Cost-efficient overall solution, as no costly special systems are required for complex tasks.


The MIRAI system extends the control of the cobot and thus enables the work environment to be perceived with its variances. Thanks to artificial intelligence, MIRAI-controlled robots can observe and imitate the movements of human workers. All observations are recorded by a camera attached to the joint of the robot. The robot is trained by a human performing a task repeatedly while guiding the robot manually on the tool. The recordings are then converted into a computer vision-based real-time robot control scheme.

Application example test tube

In this application, the system should always grab only the blue test tube and ignore all others. Interesting to see how the cobot follows the blue test tube, even after the position is changed. An application can be taught to the system in a very short time. The AI has then “learned” the information and follows it.

Thus, variances of light incidence, color, position, etc., can be ignored, as the intelligence is from what is learned.