Subtitles for the film shown in the Mister Machine exhibition.
– Maxime Labat
Robots have existed in one form or another in our collective imagination for centuries.
It is of course the myth of the perfect slave without free will who is completely slave to thank you and who nevertheless finds a way to rebel against his creators. But other imaginaries are available, for example in Japan where robots are rather benevolent, most certainly because there were very old myths in which objects or natural entities had souls.
Today these different imaginaries are mixed. We have the impression that robots will replace us in the factory chains, that some will go to war or while others will take care of dependent people. Except that robots, we do not see many of them and even less robots that look like us. So I wanted to know where we really were in the research laboratories beyond the prophecies and science fiction fantasies.
So I was welcomed to Toulouse at the Laboratory of Architectures and Systems Analysis to meet the Gepetto team who work on the movements of anthropomorphic systems. On humanoid robots what.
– Nicolas Mansard
I am Nicolas Mansard. I am a researcher in the Gepetto team of LAAS-CNRS and I am a human being. Until proven otherwise. I am going to make you play a game that is extremely difficult for robots.
– Maxime Labat
No, but a game of chess… I’m going to lose, you’ll see.
– Nicolas Mansard
It’s not what you think. The black king is missing. I hid it in the room and you will look for it, you will find it. This is something that is very difficult for a robot. So there already, you see, you are walking. You move around, you grab a book: that’s very difficult for our robot. Putting the knee on the ground, that’s fantastic. If I knew how to do that, I would have the Nobel Prize. You support yourself and you put yourself on one foot, that’s fantastic. He really is a beautiful chef. And here you recognize the black pawn even if it is in the dark. It’s very difficult for a robot, you can really consider yourself an extremely valuable robot.
– Maxime Labat
I got it buddy.
It’s funny, so your research team is called Gepetto. Is that really the plan? It’s to have the puppeteer but we cut the wires and then the puppet manages a bit on its own…
– Nicolas Mansard
We want to put the puppeteer back in the robot’s head to have Gepetto’s algorithms that make Pinocchio move.
– Maxime Labat
In fact, it’s completely banal actions that I do every day that are prodigious.
– Nicolas Mansard
Movement decisions are magical. There are a lot of decisions, very small decisions but cumulated together. That’s thousands of decisions that are made at the same time with reaction times that are much less than a second and the robot has to stand up, take a step, grab a chess piece, also make decisions at the level of perception, deciding to differentiate one object from another object… all of these are decisions that are really very complicated to make. And these decisions, we don’t associate them at all, or the public don’t associate them at all, with the intelligence of saying I want to do artificial intelligence and I want to do like my robot looks at itself climbing stairs. It’s not a smart task, a smart task is playing chess. That’s smart. But from the point of view of the number of decisions to be made, playing chess is very small, and climbing stairs involves tons of decisions, not to mention all the context around it. Getting out of bed to find your slippers, trudging over the walls, making yourself a coffee… it’s a dauntingly clever task.
– Pierre-Alexandre Leziard
So there is a small drift behind that we have to settle…
– Maxime Labat
So, that’s Pyrènes, 1m75 for 100 kg on the scale. 32 joints set in motion by electric motors. World perception level: it has two cameras for stereoscopic vision, a lidar and it is able to feel the force feedback exerted on these motors and on these soles. The pyrenes also exist in the form of a model, it is on this computer version that the movements of the robot are calculated. The goal of roboticists and computer scientists is to successfully link these two mechanical and computer entities to create a robot that is as autonomous and as less dangerous as possible, for itself and for others. In particular, it involves calibration work that you see there. To reduce the cases between planned movements and realized movements.
– Pierre-Alexandre Leziard
In addition to all that, one of the very interesting points of this robot is that it will have torque sensors in all these joints except at the level of the head and the wrists and that will make it possible to measure the forces that are applied at the robot level. So it’s going to be a lot softer control which is a lot less dangerous when operated close to the hand because if you go into it it’s not going to stay stiff but it’s going to absorb the impact.
– Maxime Labat
There is another field of study which is really the robot body. You are also working on this on the material part: which engine which material which weight in relation to the power…
– Nicolas Mansard
Designing the robot is really one of the questions. And for the quadruped robot it is a question that is well resolved. you really start to know how to design a quadruped course. For bipeds it is really a topical question.
– Maxime Labat
There you are working on a quadruped robot. Compared to Pyrene, are there things that you manage to do more?
– Pierre-Alexandre Leziard
The 2 robots implement the same control techniques. The point is to have a smaller control system, which is easier to repair and tune. Then we can put it on a humanoid robot.
We send an instruction with this joystick. The robot will therefore take this direction into account and predict what will happen half a second in the future in terms of the positioning of its feet of the forces that will apply to the ground. So he has a model of himself and he applies what is going to do and he sees the result and he does it. He was given the masses of each of these elements the masses of the legs the masses of the base of the robot. So this prediction, in half a second in the future, it does it 500 times per second and afterwards, at the engine calculation level, it does it 10,000 times per second.
– Maxime Labat
OK I feel like I’m in the future here
Making a humanoid robot is a bit like trying to calculate the decimals of Pi. The goal is to have a practical case, a kind of challenge around which researchers can come together to talk between disciplines which, without that, would have no reason to discuss.
– Nicolas Mansard
By nature robotics is a related science. Robotics has this property of connecting hard science, math and technology. We are really at the interface between the two.
– Maxime Labat
All the knowledge that you produce, it circulates within a global community freely. There is no patent here there is no open source and shared everything and public.
– Nicolas Mansard
The CNRS really aims to produce knowledge and make it available in the broadest sense. We publish under open license: our software, our solutions, our mechanical concepts or our licenses. It is really a wish in the team that we have to make available to the community at large.
– Maxime Labat
So what will be the place of humanoid robots in the future? who knows. But I tell myself that it would still be crazy if they succeeded in changing our lives without ever really existing. Without crossing the humanoid robots in the street. Never. So the moment when they will rebel against us to demand rights to the Apocalypse of killer robots is not for now.
– Nicolas Mansard
So, the joke in the field is to say: to avoid the Apocalypse of killer robots, remember to close the doors. Because we’re at the point where right now the robot they can’t walk to the door and open it. Already a door, that’s hard. So a variety of doors: a door that slides, a door that opens by pulling with a round handle… that’s playing around with the key, putting the key in the lock. So no, we are really far from having a robot that works for that. It is something that is too difficult.
– Maxime Labat
But maybe they will have changed the way we operate on cancers, make airplanes fly automatically, or make reverses work on Mars. In other words, research carried out on humanoid robots will have an impact on mathematics, computer science or even medicine.
