Sheng Zhong (Johnson)

EDIT3: It's able to work somewhat reliably now after gathering some data (around ~500 steps). I had to normalize the angle before feeding into the network because the `env` doesn't...

OK what's really strange is that using the same model structure and random seed, my [pytorch implementation](https://github.com/LemonPi/pytorch_mppi) of the generic MPPI (generalized to other problem dimensions) applied to the pendulum...

Hi, I've used it in peg-in-hole tasks both in simulation as well as in the real world. Take a look at https://arxiv.org/pdf/2010.12516 with reproducible simulation code: https://github.com/UM-ARM-Lab/tampc It depends on...

Hi @mhubii, thanks for your interest; currently there is just so(3) -> SO(3) and SO(3) -> so(3). There are no immediate plans to add in the equivalent transforms for SE(3),...

Hi I think there was a refactor at some point where now there is an inconsistency in input parameters (especially for serial chain vs chain where end effector doesn't exist)....

Thanks for your interest in citing it! For now the best way is probably the repo itself with the bibtex in the readme. We might release a technical paper in...

That should be fine - IK can be done on each kinematic chain independently. I think there was a branch that someone in the lab was working on for hands...

Hi you'll need to specify the joint values (position) of each joint. You can then do something like: ```python # specify joint values (can do so in many forms) th...

That is quite odd - do you mind sharing code that can replicate this problem? EDIT: actually looking at the code more, the joint limits are currently only used in...