TCDM / PGDM
Meta research / TCDM team
A manipulation task suite and learning framework using example object motions and pre-grasps.
What you can explore
Study how an object trajectory and a starting grasp can guide a dexterous policy.
Before you use it
The documented environment and logging dependencies are required. This is a benchmark implementation, not proof that a policy will work on a different real hand.
Original contributors
Sudeep Dasari, Abhinav Gupta and Vikash Kumar
License / access: Code: MIT · object_sim and other dependencies retain their licenses
Keep exploring
DAPG
Demonstration-augmented policy-gradient learning for dexterous hand manipulation.
DexArt
A simulated benchmark for learning to manipulate articulated objects and testing generalization to unseen objects.
Editorial reference · checked 2026-09-22