List view
[ Updated observations: now include object quaternions, gripper status, and frame stacking ] [ Updated reward functions: fully markovian, unified scaling [0, 10] and much smoother gradients via fuzzy logic ] [ Updated textures and object models ] [ Resiliency to MuJoCo physics instabilities (so that long training runs don't crash) ] Made with ❤️ in Quarantine
Overdue by 4 year(s)•Due by February 14, 2021•9/9 issues closedFixes and enhancements which make the benchmark more feasible, reliable, appropriate, etc.
Overdue by 5 year(s)•Due by April 30, 2020•4/4 issues closedBug fixes to the original release
Overdue by 5 year(s)•Due by March 31, 2020•13/13 issues closed