Kavin60606/so101_isr_193

SO-101 Teleop — ISR-Standardized (193 episodes)

Training arm B of the ISR vs Partner-sampling SmolVLA experiment. This is the client's SO-101 teleop set with ISR (Information-Standardized Trajectory Resampling) applied — pauses removed, motion kept at uniform information spacing — for comparison against the raw baseline and the partner company's length-normalization method.

TL;DR

	value
Source	makermods/200ep_blue_cube_orange_box (SO-101, real teleop)
Method	ISR — keep frames at ~uniform information gaps (arXiv:2606.22907)
Episodes	193 (holdout [96,97,98,196,197,198] excluded — never trained)
Frames	54,275 → 32,175 (59.3% kept)
fps	30 (uniform grid) · format LeRobot v3.0 · h264 video
State/action	float32[6] = [pan, lift, elbow, wrist_flex, wrist_roll, gripper], degrees
Cameras	observation.images.{front,wrist} 480×640

What ISR does

ISR resamples a trajectory by keeping one frame per fixed amount of information — instead of one frame per fixed amount of time:

information(step) = λ_dist · (joint distance moved) + λ_acc · (accumulated acceleration)

• Pauses / slow drift → ~0 information → dropped (removes the "same observation, contradictory stay-vs-move label" that makes policies hesitant).
• Fast motion + deceleration-before-contact → information-dense → sampled densely.
• Gripper open/close → force-kept (the grasp happens while the arm is nearly still, so it would otherwise be dropped).

No interpolation — the kept frames are the real recorded frames (pixels, state, action untouched); they are only re-indexed onto a uniform 30 fps grid.

Knobs (calibrated for SO-101, degrees)

Calibration measured on the 193 train episodes: median per-frame joint step 1.43°, mean acceleration 271, gripper span 40.9°/episode.

Knob	Value	Rationale
d_target	4.0	information stride ≈ 2.8× median step → 60% kept (production target; keep most of a small dataset, drop only redundancy)
λ_acc	0.00185	tuned so the acceleration term is ~20% of the info budget (paper's intent; not copied — its 0.01 is in metres, ours in degrees)
λ_dist	1.0	fixed reference scale
gripper_threshold	2.05	≈ 5% of the 40.9° gripper range → flags real grasps, not jitter

d_target is the compression dial in the signal's own units. It is anchored to the median joint step, so the same recipe transfers across robots/units; the paper's metre-values do not apply directly to degrees.

fps / 30 Hz alignment

All arms (baseline, ISR, partner) are stamped at 30 fps on a uniform grid. ISR keeps fewer frames, so an episode's dataset-time duration shrinks (its real-time non-uniform spacing is approximated as a uniform 30 fps grid). This keeps the SmolVLA action-chunk horizon (chunk_size=50) and the eval fps identical to the baseline, so the frame selection is the only variable.

Split

Holdout = last 3 episodes of each task string, [96,97,98,196,197,198] (1,622 frames), produced by lerobot eval_split=0.025. These are excluded from this dataset and reserved for evaluation — no arm ever trains on a holdout frame.

How it was built

data/teleop_std_poc pipeline:

1. fetch_trajs.py — read source states/actions, slice SO-101 layout (5 joints + gripper).
2. isr_resample.py — the ISR DP (derived from the official Apache-2.0 repo D-Robotics-AI-Lab/ISR).
3. build_dataset.py / build_parallel.py — materialize a LeRobot v3.0 video dataset:
   • sequential torchcodec range-decode of the source video (verified pixel-identical to lerobot's per-frame reader, ~40–90× faster), then h264 re-encode of the kept frames;
   • parallel sharded across CPU cores + aggregate_datasets merge (build in ~15 min).

Training / experiment

• Policy: SmolVLA (lerobot 0.6.0), byte-identical to the client recipe except the dataset.
• 20,000 steps, batch 16/GPU × 4, bf16, 4× RTX 3090.
• Compared against: raw baseline angkul07/mm_SO101_teleop (teleop100) and the partner-sampled arm Kavin60606/so101_partner_193.
• Eval: the 6-episode holdout — action MSE/MAE, per-joint MAE, gripper accuracy, and CI-MSE (grasp/place-interval error, where ISR's contact-preservation should show).

Status — trained + evaluated

Dataset built + verified (193 eps, 32,175 frames, video). SmolVLA trained 20k steps (Kavin60606/smolvla_so101_isr). Eval on the 6 holdout eps (1622 frames, seed 1000):

metric	baseline teleop100	ISR	partner
action MSE (deg²)	146.83	266.43	149.28
action MAE (deg)	5.840	8.93	5.81
gripper acc	n/a	0.852	0.932
CI-MSE mean	50.28	77.69	46.72
CI-MSE grasp	60.43	155.05	63.92
CI-MSE release	38.09	37.33	29.94

Result: ISR underperformed the raw baseline here. Global MSE ~1.8× worse and grasp-interval CI-MSE 2.5× worse — the interval ISR aimed to protect. Release CI-MSE held (≈ baseline). Read: at d_target=4.0 ISR stripped the low-motion contact-approach/settle frames around the grasp (gripper force-keep saves the close event but not the near-still approach), leaving a sparse, jumpy contact phase for the policy. A gentler d_target (keep ~80%) + wider gripper-window force-keep is the obvious next calibration. Not a verdict on ISR generally — a calibration result on this task.