Conversation

Replying to
The basic idea is not complex. You do it when driving. Straight arc segments and steady turns are what are known as trim trajectories. You patch them with discontinuous but not unstable maneuvers (turning steering wheel to new positions while braking/accelerating to a profile).
1
14
In a car, the 3 inputs get translated to 2 steering angles, and differential turn rates with fixed formulas, so 3 inputs control like a dozen variables. Here it is more complex. At least 12 joint angles controlled at once. But still the same principle.
2
15
The stable trim trajectory segments would be parametrized gaits. Like jog, run, skip, etc. Then the patching. Probably the 12 angles couple and correlate down to 3-4 most of the time.
1
11
The hardest bits would be unstable transitions (one foot landing on an inclined plane is basically a hard inverted pendulum problem) but if it’s in a human-validated mocap sequence, you’d only need to solve it for a fraction of a second to transition to next maneuver.
1
16
If this is what they’ve done, it’s extremely impressive. Though known in the literature, these are hardly routine techniques like PID and LQG (which mostly won’t work here due to nonlinearity). Much more sophisticated. Still, not defensible.
2
14
A general “tell” of control strategies is beauty. The more elegant it looks, the more symmetry there is to the behavior. Which means many variables are being reduced to a few via symmetry-exploiting maneuvers.
1
36
Contrast with the awkward flailing of trying to learn *any* workable functional behavior. Babies and injured pets often learn weird looking maneuvers this way. But more elegant looking ones are either strong attractors everybody will stumble into, or explicitly trained.
1
20
Hardware — solved Basic control — solved Maneuver control — solved Localized deep learning: 2022 Deep learning entire behavior envelope from humans: 2024 Doing it without human training examples: 2026 Learning the unstable/high-risk bits in simulation: 2028 Akira entity: 2030
7
39
Replying to
My predictions regarding robotics advance is roughly similar timeline. Do you think independent self-funded research with cheap bots (spot knockoff, robotic arms) could be a valid alternative to big labs and academic path?
1
Replying to
I'm at fork in the road so to speak. Leverage money from a well paying job towards funding this type of experiments. Or go for a research masters in germany, then figure things out by getting into their lab. I would really appreciate your thoughts here if any.
2
Show replies