June 24, 2021
Micropsi Industries has announced the next-generation of its MIRAI software, which uses artificial intelligence to help industrial and collaborative robots learn to perform camera-guided movements more quickly. Through the software, robots can flexibly react to variances in their tasks in real time by learning from humans. The company said MIRAI customers will notice quicker setup times, down from 2-3 days per skill to about three hours, along with increased robot speeds.
Variances in position, shape, surface properties and lighting conditions are common challenges when using robot automation in machine tenging, assembly, or test applications, the company said. The new MIRAI “positioning skills” feature lets a human give examples of quality movements to the robot, and the robot will generalize and understand what to do much more quickly.
“Our customers deployed the first generation of MIRAI with great success and were able to make their small-scale automation projects commercially viable for the first time,” said Ronnie Vuine, CEO of Micropsi Industries. “But we realized a decision for automating a workstation often isn’t based on a good business case alone. The project also needs to be the lowest-hanging automation fruit for the engineer who would be working on it, so we decided to lower these fruits in the second generation of MIRAI. Our new positioning skills meet the market’s need for making robotic applications flexible quickly and with little effort.”
Companies looking to use robots to perform precise and complex skills – such as gripping and inserting a bendable or soft component, like a cable, into differently arranged sockets – would primarily use the MIRAI controller at the first and last decisive centimeters of a manufacturing step.
Preparing robots to perform tasks that include variances requires that human workers guide the robot arm several times through typically occurring scenarios to show the robot to its destination, such as sockets in which freely hanging cables need to be inserted. A machine learning process then derives a motion intuition for the robot from the given examples. For a robot that is not required to follow specific paths to perform its task, MIRAI users can deploy the new positioning skills to teach the robot to find the destination even faster, because a human worker needs only to show MIRAI the surroundings of the target with the camera, Micropsi said. The robot then independently searches for the shortest path to the object.
The Germany-based company said robots can learn almost any tasks through demonstration, including precision tasks such as tracing lines, bolting differently placed screws, or even checking solder joints at varying positions. The new skills would enable automation of production steps in industries such as assembly or quality inspection, which could previously only be performed manually by human workers.
Earlier this year, Vuine (pictured at right) spoke with Robotics World about their MIRAI software and benefits for customers. “Our typical customer has had some experience with cobots,” said Vuine. “Maybe they bought one and put it up somewhere and went through all the trouble of making everything extremely precise, and they realized that materials supply is hard to do for robots. Because many of them are not coming from classical automation backgrounds, they tell us that the robot is nice, but for many applications that they could do, there’s a gap where the materials provided are not always the same, or the cable sticks out in a different direction every time. Or they need to insert it into a machine that doesn’t stop in the same place every time. There’s a gap where they need a camera, and the easiest way for them to solve that and bridge the gap is to use MIRAI.”
Training robots through pure demonstration can be beneficial to manufacturers that have variant-rich production processes, Micropsi said. “Not only will the robots be able to start performing their tasks sooner, but the manufacturers will save on both engineering and hardware costs,” the company added. Furthermore, MIRAI-supported robots have a return-on-investment of less than 12 months, making them affordable for small to midsize enterprises, the company said.
For additional details on the MIRAI software, visit the Micropsi Industries website.