19
A team of humanoid robots is working collaboratively at a car factory in China
(peertube.mesnumeriques.fr)
this post was submitted on 05 Mar 2025
19 points (100.0% liked)
Technology
1059 readers
67 users here now
A tech news sub for communists
founded 2 years ago
MODERATORS
you are viewing a single comment's thread
view the rest of the comments
view the rest of the comments
I always find it silly how humanoid robots almost never turn in a way that looks easy. They do a small shuffle and it takes multiple steps to turn 90 degrees.
I guessing that it's not a trivial problem to solve. Or maybe there are hardware limitations that don't allow human like movements. Like, maybe the hip and leg sockets are not able to make certain motions. Otherwise, it's something that could easily be solved through reinforcement learning. Maybe it's just never been a priority either.
I know that from an economic perspective, having a robot that can do a human task slowly but for less than it takes to hire a human for a proportional amount of time makes sense. And if we want to reduce mundane human working hours under socialism, it makes sense to build these sorts of robots. Especially as their abilities increase (more dexterity, better sensors, better software, etc.). Right now they feel super gimicky, but I can see the potential.
That bit at the end where the one bot charged the other bot was kind of cute.
I notice that as well, some of the movements just seem awkward. Interestingly, this has been the case for a long time too now. Once in a while you see very fluid motion from Boston Dynamics or UniTree, but it's always in demos where it's mostly scripted. Whenever you see robots operating in the wild it's always very slow and stuttered motion. I'm guessing the control systems are just not reliable enough to guarantee that the robot recover from losing balance, and it makes sense to just have them work slow and steady with minimal risk of accidents. I also expect that as the use of such robots becomes more prevalent, control systems will improve as well and they will be able to move similarly to humans.
Another aspect of it is sensors. Humans have skin, and we can immediately tell when we come in contact with an object. It's a trickier problem for a robot though since they mostly rely on visual or lidar data. I expect artificial skin and proprioception likely to be the next big innovations in robotics. If a robot is able to know the position of its body relative to the environment with a high degree of confidence, and able to sense the environment with its body, then it can move a lot more safely and reliably.