|
Download README.md from devkumarhf/Engineering-Workbench: direct link, hf CLI and curl.
- Browser
- Download file 4.38 kB
-
https://huggingface.co/spaces/devkumarhf/Engineering-Workbench/resolve/main/README.md
- Command line
-
hf download hf://spaces/devkumarhf/Engineering-Workbench/README.md
-
curl -L -o README.md https://huggingface.co/spaces/devkumarhf/Engineering-Workbench/resolve/main/README.md
4.38 kB
| title: Engineering Workbench | |
| emoji: 🔬 | |
| colorFrom: green | |
| colorTo: gray | |
| sdk: static | |
| app_file: index.html | |
| license: mit | |
| short_description: Explore engineering projects, RST control and robot data. | |
| tags: | |
| - robotics | |
| - control-systems | |
| - computer-vision | |
| - scientific-computing | |
| - education | |
| # Engineering Workbench · Dev Kumar | |
| A browser-native technical companion to [Dev Kumar's portfolio](https://dev-kumar.com/), connecting engineering methods to pinned code, contributors, reports and reproducible analysis. All computation runs locally in the visitor's browser; no inference API, model download, user upload or paid hardware is used. | |
| ## What visitors can inspect | |
| - **12 project studies** covering robotics, control, geometric perception, metrology, semiconductor modelling, turbomachinery and software. Search by technique or discipline. Each study links to a dated Git commit, identifies available artifacts, retains attribution and describes the validation scope. | |
| - **Control Lab:** an independent JavaScript translation of the RST feedback/feedforward synthesis in [`rst_pendulum_synthesis.m`](https://github.com/devkumar-projects/RST-controller-for-an-inverted-pendulum/blob/0a8a383810060c9c4d79e0ba39a400f11a3acb09/matlab/rst_pendulum_synthesis.m). It recomputes R, S, T, the Bézout coefficient residual, assigned poles and the nominal closed-loop frequency response. The internal-model frequency is fixed at 1 rad/s. Every parameter change re-synthesizes the controller; it is not a frozen-controller robustness test. The reduced transfer function is not the full Simulink or physical cart-pendulum system. | |
| - **Run explorer:** descriptive plots of the exact original 520-episode UR7e simulated training log, with selectable fields, trailing windows, per-episode inspection and unmodified CSV/provenance downloads. This is one sequential training run, not a held-out test set or physical-robot evaluation. See the [original source](https://github.com/devkumar-projects/UR7e-Reinforcement-Learning-Suite/blob/fbba2e4ef768219b7f925c091ff027d025883500/line-follower-simulation-complete/results/monitor.csv) and [Kaggle dataset](https://www.kaggle.com/datasets/devkumarkaggle/ur7e-simulated-robot-line-following-training-log). | |
| - The separate [3D ICP Lab](https://huggingface.co/spaces/devkumarhf/ICP-Lab) offers interactive rigid alignment of synthetic point clouds. | |
| ## Provenance and attribution | |
| `projects.json` records the audited Git commits and source paths. Reports remain linked to their original repositories; this Space does not republish third-party images, proprietary simulation software or the CMM code whose licence requires clarification. Each project retains its documented authors and upstream licence. Supervisors and software-support contributors are distinguished from authors. A repository or demonstration is not presented as a peer-reviewed publication. | |
| The source UR7e data hash is `ac08bd9bae19f6883adfe51203ce7a759d8cea39fd3fcbf0ec6f3ad290b99814`. `UPSTREAM-UR7E-LICENSE` retains its MIT notice. The Control Lab translation follows the MIT-licensed RST source by Dev Kumar. Workbench presentation and independent implementation: Dev Kumar, MIT. | |
| ## Reproduce and validate | |
| Serve this directory with `python3 -m http.server 8135 --bind 127.0.0.1`, then open `http://127.0.0.1:8135`. Run `node test.mjs` for domain-wide algebraic checks, internal-model response checks, source hashes and schema validation. Browser tests separately cover navigation, filtering, controls and mobile layout. No package installation is required by the application. | |
| Explore [GitHub](https://github.com/devkumar-projects), [Kaggle](https://www.kaggle.com/devkumarkaggle) and [ORCID](https://orcid.org/0009-0008-2279-9163). HAL records will be linked only after real deposits are identifiable; none are invented here. | |
| ## Reusable trained policy | |
| The [UR7e SAC model](https://huggingface.co/devkumarhf/ur7e-sac-line-following) provides actual trained weights, portable CPU inference and a paired SAC/DLS evaluation. [Run it live in the Policy Lab](https://huggingface.co/spaces/devkumarhf/UR7e-Policy-Lab) and inspect the [Kaggle experiment](https://www.kaggle.com/code/devkumarkaggle/ur7e-sac-vs-dls-reproducible-policy-benchmark). This Workbench remains the source-linked engineering overview; the Policy Lab specializes in the trained robot model. | |