Comment by hakfoo
16 hours ago
Humanoid robots are a lot of sizzle-- they promise all sorts of flexibility, at the cost of hugely higher cost/complexity/unreliability.
If you can scope your problem to some degree, you can probably make some purpose-built automation that won't look like a human, but will do the job competently and cheaply.
I see the demos with the robots carrying boxes and think "okay, why not just use a conveyer belt?"
Because, again, for home use we don't want a laundry robot, a dish washing robot, a cleaning robot, etc. We kind of have those (laundry machine, dishwasher, Roomba-types) but they all have big limitations. What people want is something that can do everything a human can do, so it can put away those dishes, wash a pan, clean the table, counters, etc. We've already scoped the problem and a humanoid-ish robot is probably the best option to do those things.
Well for home use you probably also want a robot that won’t accidentally murder your pet, injure your children, break itself and/or your prized possessions by doing the wrong thing, etc etc etc.
These are unsolved problems for robotics. There is a reason that most industrial robots work behind guards or in very constrained areas with use cases that are 100% on rails and stringently tested.
The idea that if you just make a robot in a human shape all these cease to be problems is magical thinking. We are fare from knowing that a humanoid-ish robot is the best option to do any of these things because we have no idea what it would take for it to do these things safely other than to say it would take technology that we currently don’t have.
What? You don't think I should give my humanoid robot a chainsaw and tell it to clear the field?
WCGW?
2 replies →
The laundry bot would probably be a box with some some 6DOF chopstick like positioners doing “cloth origami” to fold clothes. No need for an overkill 2kW humanoid.