3 ms·
As someone who is on record saying this wouldn't be possible, I'm pretty amazed (I believe the technical term is gobsmacked) :-). That said, the ability to mana
by ChuckMcM 6y ago
As someone who is on record saying this wouldn't be possible, I'm pretty amazed (I believe the technical term is gobsmacked) :-). That said, the ability to manage this with essentially "shadow mode" is somewhat less impressive than having the kinematics "built in" so to speak. As I understand it (and correct me if you have more recent data!) the robots are programmed by using a person wearing an articulated sensor network to capture the motion[1]. The robots repeat what ever the person does.
There are lots of interesting domestic uses for this kind of capability. In California it would be interesting to see if they could be used to fight fires. Even if the kinematics where programmed and they were launched with a 'follow' drone, I could imagine a 3rd Person game type UI for fire fighters who could take over control and "drive them" like they would a video game. That would enable a much larger fire fighting crew to be deployed.
[1] FWIW 'training mode' for workstation robots has been a thing for ever where an operator manually moves the robot through the steps it needs to do, then tweaks the program after running it. Fixed problems like "pick up a part in any orientation, then put it into this orientation" are part of the robots built in moves.
- jedimastert 6y agoI'm going to take an educated guess that using mo-cap data wouldn't work, simply because the weight distributions of people and the BD robots are completely difference. If I had to take a guess, I'd say that they were given a timeline of locations of extremities and chest/trunk and the magic sauce took care of how to distribute the weight to make it work, like a footprint chart
- dagmx 6y agoYou can run a solve on animation to accommodate physical artifacts like vibrations and other weights. Disney have a paper on this for their robotics: https://youtu.be/Z1jgaEO9aRs https://youtu.be/Z1jgaEO9aRs Though of course Disney have very different constraints, however it does illustrate that you can map from mocap to robotics motions by running solves and reversing out the artifacts. Those can be encoded into timed poses and combine that with their real time compensation to achieve this motion
- kolinko 6y agoIt won’t work for standing robots because those need to adjust for dynamic variables or they will fall down with the first tiny gust of wind (or floor micro-shake) Disney animatronics don’t have this problem because they are attached to the ground.
- dagmx 6y agoMy last paragraph mentions that aspect already. You can run some amount of the solve before hand based on what you know, and combine that with the real-time physics adjustment that these robots already do to accommodate for any differences at "runtime".
- hliyan 6y agoHope you're picking up the terms from Real Steel (2011). I think BD now has the technology to build a device similar to the ones depicted in that movie, if they wanted to.