SIMPLE
Collection
SIMulation-based Policy Learning and Evaluation • 3 items • Updated • 1
video video 0.02 0.03 | label class label 6
classes |
|---|---|
0observation.images.ego_view | |
0observation.images.ego_view | |
0observation.images.ego_view | |
0observation.images.ego_view | |
0observation.images.ego_view | |
0observation.images.ego_view | |
0observation.images.ego_view | |
0observation.images.ego_view | |
0observation.images.ego_view | |
0observation.images.ego_view | |
0observation.images.ego_view | |
0observation.images.ego_view | |
0observation.images.ego_view | |
0observation.images.ego_view | |
0observation.images.ego_view | |
0observation.images.ego_view | |
0observation.images.ego_view | |
0observation.images.ego_view | |
0observation.images.ego_view | |
0observation.images.ego_view | |
0observation.images.ego_view | |
0observation.images.ego_view | |
0observation.images.ego_view | |
0observation.images.ego_view | |
0observation.images.ego_view | |
0observation.images.ego_view | |
0observation.images.ego_view | |
0observation.images.ego_view | |
0observation.images.ego_view | |
0observation.images.ego_view | |
0observation.images.ego_view | |
0observation.images.ego_view | |
0observation.images.ego_view | |
0observation.images.ego_view | |
0observation.images.ego_view | |
0observation.images.ego_view | |
0observation.images.ego_view | |
0observation.images.ego_view | |
0observation.images.ego_view | |
0observation.images.ego_view | |
1observation.rgb_head_stereo_left | |
1observation.rgb_head_stereo_left | |
1observation.rgb_head_stereo_left | |
1observation.rgb_head_stereo_left | |
1observation.rgb_head_stereo_left | |
1observation.rgb_head_stereo_left | |
1observation.rgb_head_stereo_left | |
1observation.rgb_head_stereo_left | |
1observation.rgb_head_stereo_left | |
1observation.rgb_head_stereo_left | |
2observation.rgb_head_stereo_right | |
2observation.rgb_head_stereo_right | |
2observation.rgb_head_stereo_right | |
2observation.rgb_head_stereo_right | |
2observation.rgb_head_stereo_right | |
2observation.rgb_head_stereo_right | |
2observation.rgb_head_stereo_right | |
2observation.rgb_head_stereo_right | |
2observation.rgb_head_stereo_right | |
2observation.rgb_head_stereo_right | |
1observation.rgb_head_stereo_left | |
1observation.rgb_head_stereo_left | |
1observation.rgb_head_stereo_left | |
1observation.rgb_head_stereo_left | |
1observation.rgb_head_stereo_left | |
1observation.rgb_head_stereo_left | |
1observation.rgb_head_stereo_left | |
1observation.rgb_head_stereo_left | |
1observation.rgb_head_stereo_left | |
1observation.rgb_head_stereo_left | |
2observation.rgb_head_stereo_right | |
2observation.rgb_head_stereo_right | |
2observation.rgb_head_stereo_right | |
2observation.rgb_head_stereo_right | |
2observation.rgb_head_stereo_right | |
2observation.rgb_head_stereo_right | |
2observation.rgb_head_stereo_right | |
2observation.rgb_head_stereo_right | |
2observation.rgb_head_stereo_right | |
2observation.rgb_head_stereo_right | |
1observation.rgb_head_stereo_left | |
1observation.rgb_head_stereo_left | |
1observation.rgb_head_stereo_left | |
1observation.rgb_head_stereo_left | |
1observation.rgb_head_stereo_left | |
1observation.rgb_head_stereo_left | |
1observation.rgb_head_stereo_left | |
1observation.rgb_head_stereo_left | |
1observation.rgb_head_stereo_left | |
1observation.rgb_head_stereo_left | |
2observation.rgb_head_stereo_right | |
2observation.rgb_head_stereo_right | |
2observation.rgb_head_stereo_right | |
2observation.rgb_head_stereo_right | |
2observation.rgb_head_stereo_right | |
2observation.rgb_head_stereo_right | |
2observation.rgb_head_stereo_right | |
2observation.rgb_head_stereo_right | |
2observation.rgb_head_stereo_right | |
2observation.rgb_head_stereo_right |
Eval episodes for 17 SIMPLE G1 whole-body tasks, unzipped from SIMPLE-org/data eval-fixed/.
Layout: <Task>/dr-level-{0,1,2,hard}/.
Four tasks have a new success check. The other 13 are unchanged.
| Task | Change |
|---|---|
| G1WholebodyBendHandoverTeleop-v0 | The box's center of mass must now be inside the basket's xy footprint, not just touching the basket. A box resting against the outside of the basket no longer counts. |
| G1WholebodyBendPickAndPlaceTeleop-v0 | Same center-of-mass-in-container check as BendHandoverTeleop. |
| G1WholebodyBendPickAndPlaceOnSofaMP-v0 | Success now counts only while the box is on the sofa and released (no hand link touching it). The per-step reward is rounded so that 9 steps × 0.1 reach the 0.9 threshold exactly. |
| G1WholebodyPickAndBendPlaceMP-v0 | The container is now scene19's dog bed. Success means the box touches the bed and does not touch world; the old check counted contact with the floor. |
| All other 13 tasks | No change to the success check. |