Factory automation has plowed ahead in no small part because of the controlled conditions, but achieving the same thing in the field is much more difficult.
On the flip side, navigating a vehicle is likely far simpler than navigating a bipedal robot with digits and making it do what you want because vehicles really only move along a two-dimensional plane and in an environment that is, at least in its locality, highly structured and standardized with specific controls. You go forward/back and left/right, motion is achieved simply by spinning motors up to your desired speed, you move based on signage and colored signals.
Factories are more structured and standardized than even roads are. Everything is marked with lines, signage and in some cases visual signals. Nearly everything is rectilinear. If proper 5S housekeeping practices are being followed, then everything is clean and organized. Human behavior is the main cause of variance and disarray in factories, because we constantly need to be reminded to apply discipline about cleaning up after ourselves and putting stuff where it belongs. A production cell with no humans and only Optimi won’t have this problem.
Moving legs around to walk across flat, level, uniform concrete flooring is not very difficult with today’s technology. Boston Dynamics already has this down pat.
A bipedal robot needs to be far more versatile and dexterous just to pull off what we'd consider simple maneuvers, like walking over to a box and picking it up. You need motors moving the legs correctly, bending the knees, bending the arms, making sure the robot lifts with its legs and doesn't strain the low back, grip the box with its hands, and stand again.
For many tasks yes, but I’m sure there are some tasks simple enough to start with.
Making a humanoid robot position its joint correctly for optimal leverage in picking up a box is probably a lot easier than teaching a human to do it. I would know because I like powerlifting and have taught quite a few humans how to squat and deadlift correctly. The robot has sensors detecting joint torque and aperture angles precisely and can run a simple mechanical simulation to get the job done perfectly. A leg is just a compound lever with three fulcrums, and the math for a vertical lift is thus straightforward trigonometry that any engineer should be able to program. Humans have to go by feel, have imperfect proprioception and balance, have shaky connections from their motor cortex to their muscles, and have reaction times of 300+ milliseconds which makes real-time correction of mistakes laggy. Humans also have to break bad habits that developed as a result of living a physically easy sedentary lifestyle.
Indeed, Boston Dynamics solved bending over and picking up boxes 6 years ago.
Also, I’m pretty sure the standard for factories for lifting is 50 lbs maximum and anything at that limit is supposed to be located at waist height for optimal ergonomics, and in this power zone barely any joint flexion is needed in the first place.
Can the robot install and fasten nuts and bolts, route and terminate wires/tubes/hoses, rip plastic wrap off pallets, open boxes and break down the cardboard for recycling, or click interior trim into place? We will have to wait and see.
And if we're to generalize this outside of controlled factories, the complexity explodes and there will be no standardized controls.
The challenge in pulling this off is monumental, the complexity is staggering
Right, but not even Tesla hyperbulls are anticipating that being achieved by Sep 30th. When asked on the Q4 2021 earning call about where Optimus will be used first, Elon said “Yeah. The first use of the Tesla Bot, Optimus, the Optimus name seems to be sticking at least internally, Optimus Subprime. Like if we can't find a use for it then we shouldn't expect that others would. So the first use of the Optimus robots would be at Tesla, kind of like moving parts around the factory or something like that.” So for now, the goal is actual usefulness for simple factory tasks. The broader goal is much harder and is a question for further in the future.
