dsfb-robotics 0.1.0

DSFB Structural Semiotics Engine for Robotics Health Monitoring — a deterministic, non-intrusive observer layer that reads the residuals existing robot control, kinematic identification, and whole-body balance observers already compute, and structures them into a human-readable grammar of typed episodes. Read-only augmentation, not replacement, of incumbent PHM / FDD / MPC / observer pipelines.
Documentation 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32

# SO-ARM100 (SO-100) pick-and-place — oracle protocol

Dataset: **HuggingFace LeRobot SO-ARM100 (SO-100)** pick-and-place corpus.

## Provenance

- **Public archive:** HuggingFace dataset [`lerobot/so100`](https://huggingface.co/datasets/lerobot/so100).
- **Platform:** SO-ARM100 (SO-100) — a low-cost 5-DoF DIY arm built on commodity Dynamixel servos. Open-hardware design.
- **Sampling rate:** ~30 Hz logged.

## Licence and redistribution

- **Licence:** Apache-2.0.
- **This crate ships:** preprocessed scalar residual stream at [`data/processed/so100.csv`](../data/processed/so100.csv) (19 631 frames).

## Fetch path

1. The preprocessor downloads the LeRobot parquet into `data/so100/`.
2. Residual = Euclidean norm of joint-state deviation from 20% early-window nominal per timestep.
3. Run `paper-lock so100`.

## Residual construction

`r(k) = ‖q(k) − nominal‖₂` for the 5-DoF arm + gripper joint state.

## DSFB bounded claim

Surfaces commodity-servo pick-and-place residual structure as grammar episodes. The high peak ‖r‖² is consistent with low-cost-servo amplitude characteristics — DSFB structures rather than corrects this. No claim about SO-100 servo accuracy.

## Reproducibility

- `paper-lock so100`. Smoke-test: `paper-lock so100 --fixture`.