MIRAI: Vision-Based Controller for Handling Variance
MIRAI (by Micropsi Industries) is a vision-based control system that enables industrial robots to deal with variance in position, shape, or movement. Using artificial intelligence, the MIRAI controller generates robot movements directly and in real-time, as opposed to outputting individual measurements. Deploy MIRAI in assembly, testing and measuring, machine tending, gripping, or contour following (e.g., for dispersing). MIRAI can be used for multiple steps within a task (e.g., gripping from variable position and assembly with varying shape). It can also be flexibly retrained for new tasks. Skills are not programmed but trained through human demonstration in a few days — no knowledge of programming or AI necessary.
There are four key advantages of the MIRAI robot control system compared to more classical automation approaches. One, industrial robots powered by MIRAI are able to deal with variance in a workspace. They can perform tasks that would be very hard or impossible to hand-engineer. Two, MIRAI is easy to deploy. Customers can quickly and easily train or retrain robot movements by themselves in just a few days. No knowledge of programming or AI is required. Three, because a MIRAI-enabled robot can learn many different tasks, it can be used for various process steps. Four, MIRAI is a cost-effective solution for highly complex tasks. The cameras delivered with the system are sufficient for solving various sub-tasks. Expensive, special-purpose systems are not necessary anymore.
How does it work?
MIRAI is a control system that augments the robot’s native controller, giving machines the capacity to perceive their work environment and deal with variance. Through deep learning, MIRAI-enabled robots solve complex tasks by observing actions performed by a human operator and then imitating those actions (not by replaying the actions but generalizing from them). Observations are recorded by a camera attached to the robot’s wrist and/or situated at a fixed position proximate to the workspace. To train the robot, a human user performs and records repeated demonstrations of a task by manually guiding the robot by the robot’s wrist. The recordings are then transformed into a vision-based, real-time robot control schema, called a MIRAI skill, which shows up as a simple node in a Polyscope program. Trained skills allow the MIRAI-controlled robot to deal with all the dynamics the user trained for.
Software version requiredMin. Polyscope 5.3.0
Certifications and standardsCE, FCC, RoHS
DimensionsDimensions MIRAI controller box: 210mm x 177mm x 45mm
What's in the box?1 x MIRAI ux-1 robot controller
1 x L-mount for controller box
1 x universal AC power supply DC19V/120W
1 x AC power cord for EU and US region
2 x wireless antennas
1 x Ximea xiQ USB3 camera
1 x USB3 cable
1 x Fujinon lense
1 x Camera fixtures for UR robot arms
4 x mounting screws for camera fixture
1 x USB pendrive with MIRAI URCaps plugin and product documentation
License typeOne-time purchase