Pytorch code for CoRL 2020 paper Action-Conditional Recurrent Kalman Networks For Forward and Inverse Dynamics Learning
- torch==1.3.1
- python 3.*
Forward Dynamics - With action-conditional-rkn
as the working directory execute the python script
python experiments/PamForward/pneumaticArm.py
Inverse Dynamics - With action-conditional-rkn
as the working directory execute the python script
python experiments/Franka/frankaArm.py
Forward dynamics learning uses subset of muscular robot table tennis data from here. Inverse dynamics learning uses inhouse data collected from Franka Emika Panda