Watch the videos and use the interactive companions
Five original six-minute narrated videos, each with geometric animation, eight chapters, visible captions, downloadable transcripts, primary paper sources, and an interactive HTML laboratory.
| Paper | Interactive companion |
|---|---|
| Integrated Task and Motion Planning — Garrett et al. | Grid A*, collision checking, action skeletons, and movable obstacles |
| PDDLStream | Seeded stream evaluation, dependencies, reachability, and geometric paths |
| OWL-TAMP | Continuous predicates and intersections with feasible geometry |
| TiPToP | Candidate optimization, perception bias, and scene disturbance |
| GaP | Skill-graph trials, retry edges, reliability, recovery, and throughput |
Open index.html after downloading or cloning the repository. Keep the media, previews, captions, and transcripts directories alongside it. Viewing and experiments work offline without installation or API keys. Individual HTML entry points are included for each paper.
For an optional local server with video seeking:
python3 serve.py --port 8768Then open http://127.0.0.1:8768. Videos are H.264/AAC MP4, 1280×720, 24 fps, exactly 6:00 each, with embedded subtitles and chapter markers. README.txt contains rebuild instructions. The original narration, renderer, and companion code are in src; verification results are in qa.
Toy experiments are original illustrations, not paper benchmark replications. Primary sources and scope notes appear in each companion and transcript. The narration is synthetic and does not impersonate an author or video creator. Research PDFs are linked rather than redistributed.
GitHub Pages serves the main branch from / with .nojekyll.
Related collection: Learning Without a Reset — six detailed reset-free RL lessons and labs.