News for hackers. Stuff that matters.
Re:Dobb·E: An open-source framework for learning household r... by GaggiX on Wednesday November 29, 2023 @09:06AM (#38459454)
Wow the model architecture seems extremely simple, very cool project.
Re:Dobb·E: An open-source framework for learning household r... by dobb-e on Wednesday November 29, 2023 @10:09AM (#38460276)
Thank you! We tried to keep things as simple as possible on the policy side, but definitely there is a lot more room for innovation to go from 81% to 99%, like using temporal information and global structure of the task.
Re:Dobb·E: An open-source framework for learning household r... by GaggiX on Wednesday November 29, 2023 @10:19AM (#38460454)
Another limitation here is that the model is regressive, so for example if a task was to pick up one bottle out of two and the demos showed 50/50 of picking up one than the other, the model would output the mean even though it is not meaningful.
Re:Dobb·E: An open-source framework for learning household r... by dobb-e on Wednesday November 29, 2023 @10:22AM (#38460493)
Indeed! In fact, I have a project [0] from last year that uses a GPT-style transformer to address that exact issue :) However, it’s hard to go far outside simulations in real home robotics without a good platform, out of which efforts came Dobb-E.

[0] https://mahis.life/bet/

Re:Dobb·E: An open-source framework for learning household r... by GaggiX on Wednesday November 29, 2023 @10:39AM (#38460766)
I've also seen the one that uses the diffusion process for planning, I imagine it's even slower, but maybe with a consistency loss something can be done about it.