Hugging Face's logo

danielharkin21
/
ath-h2-policies

Reinforcement Learning
ONNX
robotics
mujoco
humanoid
tennis
Model card Files
xet
 Community
ath-h2-policies / sim /ref_obs.json
danielharkin21's picture
danielharkin21
sim/: exact training MJCF, canonical feeds + spec, reference 137-dim obs
ea98a5d verified
Raw History Blame Contribute Delete
5.45 kB
{
"note": "Reference observation from the f64 oracle/screener obs builder (identical formula to body.py H2Ball._obs), at step 0 of screener episode seed [777,0]. obs is float32. R = pelvis xmat (world<-body); p = pelvis qpos[:3]. Ball history initialised by tiling the current ball pos/vel NH times, oldest first.",
"obs_dim": 137,
"NU": 29,
"NH": 5,
"EP": 172,
"feed_index": 7644,
"feed": [
-11.623903274536133,
-1.1025949716567993,
0.7991899847984314,
33.843265533447266,
5.603490352630615,
3.9499590396881104,
-24.566015243530273,
148.3707733154297,
122.46565246582031,
7.946280479431152,
2.8357927799224854,
0.7139999866485596,
1.3660000562667847
],
"qpos": [
11.400282101796009,
3.708462939512078,
1.03,
0.058558186426560764,
0.0,
0.0,
0.998283997068185,
-0.2803837437573452,
-0.01629585781334563,
0.006847498201466108,
0.5039592103431634,
0.023604842612876722,
-0.2569181198940864,
-0.2942473089334982,
-0.03339415989785495,
0.025740676597470244,
0.5179938602846471,
0.020152897120703298,
-0.28947396761690847,
-0.04535885224963976,
0.043190593305236905,
-0.007832063045376723,
0.3033845493439266,
0.15198306949036436,
-0.0010203408950045026,
0.8523134248569663,
-0.039676644568906794,
0.03265618218307775,
-0.037200897055508976,
0.3200018724894495,
-0.21299300357433093,
0.03920143663993371,
0.8938692872738688,
0.02779018824420379,
-0.04256148121370526,
-0.03679482832124549,
-11.623903274536133,
-1.1025949716567993,
0.7991899847984314,
1.0,
0.0,
0.0,
0.0
],
"qvel": [
0.0,
0.0,
0.0,
0.0,
0.0,
0.0,
0.0,
0.0,
0.0,
0.0,
0.0,
0.0,
0.0,
0.0,
0.0,
0.0,
0.0,
0.0,
0.0,
0.0,
0.0,
0.0,
0.0,
0.0,
0.0,
0.0,
0.0,
0.0,
0.0,
0.0,
0.0,
0.0,
0.0,
0.0,
0.0,
33.843265533447266,
5.603490352630615,
3.9499590396881104,
-24.566015243530273,
148.3707733154297,
122.46565246582031
],
"Q0": [
-0.25,
0.0,
0.0,
0.5,
0.0,
-0.25,
-0.25,
0.0,
0.0,
0.5,
0.0,
-0.25,
0.0,
0.0,
0.0,
0.35,
0.18,
0.0,
0.87,
0.0,
0.0,
0.0,
0.35,
-0.18,
0.0,
0.87,
0.0,
0.0,
0.0
],
"layout": [
{
"name": "gravity_in_pelvis",
"start": 0,
"len": 3
},
{
"name": "pelvis_angvel_local",
"start": 3,
"len": 3
},
{
"name": "joint_pos_minus_Q0",
"start": 6,
"len": 29
},
{
"name": "0.1*joint_vel",
"start": 35,
"len": 29
},
{
"name": "last_action",
"start": 64,
"len": 29
},
{
"name": "ball_pos_rel_pelvis_local",
"start": 93,
"len": 3
},
{
"name": "0.1*ball_vel_local",
"start": 96,
"len": 3
},
{
"name": "racket_center_minus_ball_local",
"start": 99,
"len": 3
},
{
"name": "pelvis_linvel_local(R.T@qvel[0:3])",
"start": 102,
"len": 3
},
{
"name": "pelvis_height",
"start": 105,
"len": 1
},
{
"name": "step/EP",
"start": 106,
"len": 1
},
{
"name": "ball_history_pos_rel_pelvis(NH x3, @R)",
"start": 107,
"len": 15
},
{
"name": "ball_history_0.1*vel(NH x3, @R)",
"start": 122,
"len": 15
}
],
"obs": [
0.0,
0.0,
-1.0,
0.0,
0.0,
0.0,
-0.030383743345737457,
-0.01629585772752762,
0.006847498007118702,
0.0039592101238667965,
0.02360484190285206,
-0.006918119732290506,
-0.04424731060862541,
-0.033394161611795425,
0.02574067749083042,
0.017993859946727753,
0.020152896642684937,
-0.039473965764045715,
-0.04535885155200958,
0.043190594762563705,
-0.007832063362002373,
-0.04661545157432556,
-0.02801693044602871,
-0.0010203409474343061,
-0.017686575651168823,
-0.039676643908023834,
0.03265618160367012,
-0.03720089793205261,
-0.02999812737107277,
-0.032993003726005554,
0.039201438426971436,
0.023869287222623825,
0.027790188789367676,
-0.04256148263812065,
-0.03679483011364937,
0.0,
0.0,
0.0,
0.0,
0.0,
0.0,
0.0,
0.0,
0.0,
0.0,
0.0,
0.0,
0.0,
0.0,
0.0,
0.0,
0.0,
0.0,
0.0,
0.0,
0.0,
0.0,
0.0,
0.0,
0.0,
0.0,
0.0,
0.0,
0.0,
0.0,
0.0,
0.0,
0.0,
0.0,
0.0,
0.0,
0.0,
0.0,
0.0,
0.0,
0.0,
0.0,
0.0,
0.0,
0.0,
0.0,
0.0,
0.0,
0.0,
0.0,
0.0,
0.0,
0.0,
0.0,
0.0,
0.0,
0.0,
0.0,
22.303796768188477,
7.469944953918457,
-0.2308100163936615,
-3.295603036880493,
-0.9521859884262085,
0.39499589800834656,
-21.908817291259766,
-7.758368968963623,
0.30706900358200073,
0.0,
0.0,
0.0,
1.0299999713897705,
0.0,
22.303796768188477,
7.469944953918457,
-0.2308100163936615,
22.303796768188477,
7.469944953918457,
-0.2308100163936615,
22.303796768188477,
7.469944953918457,
-0.2308100163936615,
22.303796768188477,
7.469944953918457,
-0.2308100163936615,
22.303796768188477,
7.469944953918457,
-0.2308100163936615,
-3.295603036880493,
-0.9521859884262085,
0.39499589800834656,
-3.295603036880493,
-0.9521859884262085,
0.39499589800834656,
-3.295603036880493,
-0.9521859884262085,
0.39499589800834656,
-3.295603036880493,
-0.9521859884262085,
0.39499589800834656,
-3.295603036880493,
-0.9521859884262085,
0.39499589800834656
],
"action_map": "ctrl = clip(Q0 + 0.3*tanh(policy_mean), ctrlrange); 10 physics substeps of 2 ms per 20 ms control step; aero drag+Magnus force applied to ball each substep"
}