google-deepmind / mujoco_mpc

Real-time behaviour synthesis with MuJoCo, using Predictive Control
https://github.com/deepmind/mujoco_mpc
Apache License 2.0
913 stars 135 forks source link

Cube solve task: Approximate remaining unscramble cost #267

Closed thowell closed 6 months ago

thowell commented 6 months ago

@erez-tom

Added a cost term Remaining that depends on goal_index_ in order to show total cost to solve state. Currently uses a fixed factor of 12. Note this factor is based on the current settings and weights. Should be sufficient for now.