Mujin’s motion planning AI
High speed analytical inverse kinematics for any robot
Before now, each robot required a program to calculate the forward kinematics for each task, which always took a long period of time to develop, but Mujin’s kinematics calculation library is able to create an optimal program for calculating reverse kinematics equations for every robot considering all the edge cases. At this point, it has the distinction of being used by more than 1000 robots throughout the world.
Motion planning that takes the real world into account
In the real world, a teachless robot is not good enough if it is merely able to move. To be useful, it must reach its destination while considering the obstacles, joint limits, singularities and dynamics that are always present in the real world. Manual teaching work, which involved manually moving a robot while recording its trajectory, is not required when using Mujin technology.
It looks like a small step, but this is a big, innovative step for the robot world
What is Mujin’s teachless technology, motion planning AI?
Before now, robots could not move without teaching (programming); on the contrary, a robot would, by definition, faithfully repeat whatever it was taught.
Compared to these robots that could not move unless taught, robots equipped with Mujin’s teachless technology, the motion planning AI can think through movements themselves. This ability comes natural to human beings, so we will explain why it is a big step.
How are humans and robots different?
Movement is one of the variety of things that we humans think of as ordinary, but actually requires us to do very sophisticated processing.
For example, imagine that I ask you to pick up a cup in front of you. Of course, you are able to grab the cup. Even if there was a different cup in front of it, you would be able to avoid that cup and grab the correct one.
But when you grabbed the cup, do you remember how you moved your arm to reach for it? Raise your shoulder joint by 20 degrees, lower the elbow by 30 degrees, hold that pose while moving forward 20 centimeters, grab! There’s really no need to think about it. Movement is done completely unconsciously, with no thought at all. Humans just need to see the target; essentially, if they know the destination, their bodies will unconsciously move towards it.
Robots cannot move unconsciously
However, robots are different. Basically, robots cannot move without prior instructions. To further the analogy, if it sees a cup, even if you specify the target location, it will not be able to move there unless motion instructions are supplied beforehand.
If only a simple motion is needed, it is fine to teach it once so the movement is remembered, but to rely on 3D sensors and the like to move unregistered objects without providing enormous quantities of movement instructions in advance to account for all of the possibilities is extremely difficult in reality.
In short, teaching a robot is not a matter of instructing it to “Please pick up that cup,” as the robot will not move without movement instructions, which need to be supplied by a robot teaching professional.
Motion planning technology to make robots more like humans
When Mujin motion planning technology is installed in a robot, if the object can be seen or the movement destination specified, the robot will automatically move towards it like a human, while avoiding obstacles of its own accord, making it possible to “become closer to a human” in ability.
Although this looks like a small step at first glance, being able to use robots in such a human-like way is an innovative breakthrough, a large step away from the world of teaching.
There is huge demand in the picking area of logistics, food and automotives that could not be done by robots before; now that robots will be able to see and move by themselves, they can fill these roles.
Behaviors that cannot be taught by human teaching are now possible with motion planning technology
It is almost impossible for humans to program a robot to transport liquid at this speed without spilling it.