Canonical references and the offline source mirror

E.1 How to use the offline source mirror

All canonical arXiv PDFs cited in the book are mirrored under sources/pdfs/ in the project’s working directory, with filenames of the form arxiv_<id>_<short-title>.pdf — for example, arxiv_2406.09246_OpenVLA.pdf. Companion blog posts, vendor technical reports, and product pages that lack arXiv equivalents are saved as HTML snapshots under sources/html/ with descriptive filenames matching their domain — for example, huggingface_blog_pi0_and_pi0_fast.html or figure_ai_news_helix.html. The machine-readable index sources/SOURCES_INDEX.txt lists every URL → local-file mapping, plus the curl-based download_sources.sh script used to (re)generate the mirror.

Four citation-hygiene rules apply throughout the book.

Prefer the arXiv ID (for example, arXiv:2406.09246) over a hosted URL. arXiv IDs are stable; URLs are not.

When a model is best described by a vendor blog or technical report rather than a peer-reviewed paper — Helix, Helix 02, RoboBrain2.0, ρα (Rho-alpha) — the canonical URL is given and the snapshot date is recorded in SOURCES_INDEX.txt.

Survey citations are kept up to date in each chapter’s reference list and in §E.2 of this appendix. A reader who wants a single point of entry to the literature starts with the surveys listed under Chapter 1.

Numerical claims — parameter counts, training-hours figures, success rates — are always cited to the model’s primary reference. The Model Zoo in Appendix F reuses those numbers verbatim and points back to the primary-reference column rather than re-citing the originals.

E.2 Reading list, by chapter

Chapter 1. The robot learning problem

• Kober, J., Bagnell, J. A., & Peters, J. (2013). Reinforcement Learning in Robotics: A Survey. International Journal of Robotics Research, 32(11), 1238–1274.
• Sapkota, R. et al. (2025). Vision-Language-Action Models: Concepts, Progress, Applications and Challenges. arXiv:2505.04769.
• Zhang et al. (2025). Pure Vision Language Action (VLA) Models — A Comprehensive Survey. arXiv:2509.19012.
• Ni et al. (2025). Embodied Arena — A Comprehensive, Unified, and Evolving Evaluation Platform. arXiv:2509.15273.
• Xu et al. (2025). An Anatomy of Vision-Language-Action Models — From Modules to Milestones and Challenges. arXiv:2512.11362.

Chapter 2. Your first VLA, end-to-end

• Kim, M. J. et al. (2024). OpenVLA — An Open-Source Vision-Language- Action Model. arXiv:2406.09246.
• Liu, B. et al. (2023). LIBERO — Benchmarking Knowledge Transfer for Lifelong Robot Learning. NeurIPS 2023 Datasets and Benchmarks track.
• Garcia et al. (2025). LIBERO-Para — A Diagnostic Benchmark and Metrics for Paraphrase Robustness in VLA Models. arXiv:2603.28301.

Chapter 3. Math and ML prerequisites in 30 minutes

• Goodfellow, I., Bengio, Y., & Courville, A. (2016). Deep Learning. MIT Press.
• Bishop, C. M. (2006). Pattern Recognition and Machine Learning. Springer.
• Paszke, A. et al. (2019). PyTorch — An Imperative Style, High- Performance Deep Learning Library. NeurIPS 2019.

Chapter 4. Classical action models: planning and inverse dynamics

• Fikes, R. E., & Nilsson, N. J. (1971). STRIPS — A New Approach to the Application of Theorem Proving to Problem Solving. Artificial Intelligence, 2(3–4), 189–208.
• McDermott, D. et al. (1998). PDDL — The Planning Domain Definition Language. Yale Center for Computational Vision and Control TR-98-003.
• Helmert, M. (2006). The Fast Downward Planning System. Journal of Artificial Intelligence Research, 26, 191–246.
• Garrett, C. R., Lozano-Pérez, T., & Kaelbling, L. P. (2020). PDDLStream — Integrating Symbolic Planners and Blackbox Samplers via Optimistic Adaptive Planning. ICAPS 2020.
• Ahn, M. et al. (2022). Do As I Can, Not As I Say — Grounding Language in Robotic Affordances (SayCan). arXiv:2204.01691.
• LaValle, S. M. (2006). Planning Algorithms. Cambridge University Press.
• Featherstone, R. (2008). Rigid Body Dynamics Algorithms. Springer.
• Khatib, O. (1987). A Unified Approach for Motion and Force Control of Robot Manipulators — The Operational Space Formulation. IEEE Journal of Robotics and Automation, 3(1).
• Hogan, N. (1985). Impedance Control — An Approach to Manipulation. ASME Journal of Dynamic Systems, Measurement, and Control.

Chapter 5. Learning from rewards: MDPs and reinforcement learning

• Bellman, R. (1957). Dynamic Programming. Princeton University Press.
• Howard, R. A. (1960). Dynamic Programming and Markov Processes. MIT Press.
• Puterman, M. L. (1994). Markov Decision Processes — Discrete Stochastic Dynamic Programming. Wiley.
• Sutton, R. S., & Barto, A. G. (2018). Reinforcement Learning — An Introduction (2nd ed.). MIT Press.
• Kober, J., Bagnell, J. A., & Peters, J. (2013). Reinforcement Learning in Robotics — A Survey. International Journal of Robotics Research, 32(11).
• Kaelbling, L. P., Littman, M. L., & Cassandra, A. R. (1998). Planning and Acting in Partially Observable Stochastic Domains. Artificial Intelligence, 101(1–2).

Chapter 6. Learning from demonstrations: behavior cloning and imitation learning

• Pomerleau, D. A. (1988). ALVINN — An Autonomous Land Vehicle in a Neural Network. NeurIPS 1988.
• Ross, S., Gordon, G. J., & Bagnell, J. A. (2011). A Reduction of Imitation Learning and Structured Prediction to No-Regret Online Learning (DAgger). AISTATS 2011.
• Ng, A. Y., & Russell, S. J. (2000). Algorithms for Inverse Reinforcement Learning. ICML 2000.
• Argall, B. D., Chernova, S., Veloso, M., & Browning, B. (2009). A Survey of Robot Learning from Demonstration. Robotics and Autonomous Systems, 57(5).

Chapter 7. Deep RL for control: DQN to SAC and PPO

• Mnih, V. et al. (2015). Human-level Control through Deep Reinforcement Learning (DQN). Nature, 518(7540).
• Lillicrap, T. P. et al. (2016). Continuous Control with Deep Reinforcement Learning (DDPG). ICLR 2016.
• Schulman, J. et al. (2017). Proximal Policy Optimization Algorithms (PPO). arXiv:1707.06347.
• Fujimoto, S., van Hoof, H., & Meger, D. (2018). Addressing Function Approximation Error in Actor-Critic Methods (TD3). ICML 2018.
• Haarnoja, T. et al. (2018). Soft Actor-Critic. ICML 2018.
• Tobin, J. et al. (2017). Domain Randomization for Transferring Deep Neural Networks from Simulation to the Real World. IROS 2017.

Chapter 8. Sequence models meet control

• Vaswani, A. et al. (2017). Attention Is All You Need. NeurIPS 2017.
• Chen, L. et al. (2021). Decision Transformer — Reinforcement Learning via Sequence Modeling. NeurIPS 2021.
• Janner, M., Li, Q., & Levine, S. (2021). Offline Reinforcement Learning as One Big Sequence Modeling Problem (Trajectory Transformer). NeurIPS 2021.
• Fu, J. et al. (2020). D4RL — Datasets for Deep Data-Driven Reinforcement Learning. arXiv:2004.07219.
• Liu et al. (2024). RoboMamba — Multimodal State Space Model for Efficient Robot Reasoning and Manipulation. arXiv:2406.04339.
• Wu et al. (2025). A Survey on Efficient Vision-Language-Action Models. arXiv:2510.24795.

Chapter 9. World models and model-based learning

• Ha, D., & Schmidhuber, J. (2018). World Models. NeurIPS 2018.
• Hafner, D. et al. (2019). Learning Latent Dynamics for Planning from Pixels (PlaNet). ICML 2019.
• Hafner, D., Lillicrap, T., Ba, J., & Norouzi, M. (2020). Dream to Control — Learning Behaviors by Latent Imagination (Dreamer). ICLR 2020.
• Hafner, D. et al. (2023). Mastering Diverse Domains through World Models (DreamerV3). arXiv:2301.04104.
• Bruce, J. et al. (2024). Genie — Generative Interactive Environments. ICML 2024.
• LeCun, Y. (2022). A Path Towards Autonomous Machine Intelligence. OpenReview position paper.

Chapter 10. Diffusion and flow models for action generation

• Ho, J., Jain, A., & Abbeel, P. (2020). Denoising Diffusion Probabilistic Models. NeurIPS 2020.
• Song, Y. et al. (2021). Score-Based Generative Modeling through Stochastic Differential Equations. ICLR 2021.
• Chi, C. et al. (2023). Diffusion Policy — Visuomotor Policy Learning via Action Diffusion. RSS 2023.
• Zhao, T. Z., Kumar, V., Levine, S., & Finn, C. (2023). Learning Fine-Grained Bimanual Manipulation with Low-Cost Hardware (ACT). RSS 2023.
• Lipman, Y. et al. (2023). Flow Matching for Generative Modeling. ICLR 2023.
• Liu, X., Gong, C., & Liu, Q. (2023). Flow Straight and Fast — Learning to Generate and Transfer Data with Rectified Flow. ICLR 2023.

Chapter 11. The VLA recipe: from CLIP to RT-1

• Radford, A. et al. (2021). Learning Transferable Visual Models from Natural Language Supervision (CLIP). ICML 2021.
• Jang, E. et al. (2022). BC-Z — Zero-Shot Task Generalization with Robotic Imitation Learning. CoRL 2021.
• Brohan, A. et al. (2022). RT-1 — Robotics Transformer for Real-World Control at Scale. arXiv:2212.06817.
• Reed, S. et al. (2022). A Generalist Agent (Gato). TMLR.

Chapter 12. Scaling up: PaLM-E, RT-2, OpenVLA, Octo (and friends)

• Driess, D. et al. (2023). PaLM-E — An Embodied Multimodal Language Model. arXiv:2303.03378.
• Brohan, A. et al. (2023). RT-2 — Vision-Language-Action Models Transfer Web Knowledge to Robotic Control. arXiv:2307.15818.
• Kim, M. J. et al. (2024). OpenVLA. arXiv:2406.09246.
• Octo Model Team / Ghosh, D. et al. (2024). Octo — An Open-Source Generalist Robot Policy. arXiv:2405.12213.
• O’Neill, A. et al. (2024). Open X-Embodiment — Robotic Learning Datasets and RT-X Models. arXiv:2310.08864.
• Liu et al. (2024). RDT-1B — A Diffusion Foundation Model for Bimanual Manipulation. arXiv:2410.07864.

Chapter 13. Smooth control: π0 and flow-matching action heads

• Black, K. et al. (2024). π0 — A Vision-Language-Action Flow Model for General Robot Control. arXiv:2410.24164 (companion technical report at pi.website/download/pi0.pdf).
• Pertsch, K. et al. (2025). FAST — Efficient Action Tokenization for Vision-Language-Action Models. arXiv:2501.09747.
• Frontier Robotics (2025). SimVLA — A Simple VLA Baseline for Robotic Manipulation. arXiv:2602.18224.
• Xiaomi Robotics (2025). Xiaomi-Robotics-0. arXiv:2602.12684.
• Hugging Face (2025). SmolVLA. arXiv:2506.01844.

Chapter 14. Dual-system architectures: Helix and GR00T N1

• Figure AI (2024). Helix — A Vision-Language-Action Model for Generalist Humanoid Control. figure.ai/news/helix.
• Figure AI (2026). Helix 02 — A General-Purpose Humanoid System. figure.ai/news/helix-02.
• NVIDIA (2025). GR00T N1 — An Open Foundation Model for Generalist Humanoid Robots. arXiv:2503.14734.
• Driess, D. et al. (2023). PaLM-E (antecedent of VLM + controller decoupling). arXiv:2303.03378.
• Ahn, M. et al. (2022). Do As I Can, Not As I Say (SayCan). CoRL 2022.

Chapter 15. Datasets, benchmarks, and evaluation; specialized and adjacent VLAs

• Wei et al. (2025). LiLo-VLA — Compositional Long-Horizon Manipulation via Linked Object-Centric Policies. arXiv:2602.21531.
• Long-VLA Team (2025). Unleashing Long-Horizon Capability of Vision- Language Models for Manipulation (Long-VLA). arXiv:2508.19958.
• Tianjin University / Huawei (2025). Embodied-R1 — Reinforced Embodied Reasoning for Manipulation. arXiv:2508.13998.
• Huang et al. (2024). LEO — An Embodied Generalist Agent in 3D World. arXiv:2311.12871.
• OpenDriveVLA Team (2025). OpenDriveVLA — Towards End-to-End Autonomous Driving with VLAs. arXiv:2503.23463.
• Zheng et al. (2025). UniAct — A Universal Action Model for Cross- Embodiment Robot Manipulation. arXiv:2501.10105.
• TinyVLA Team (2024). TinyVLA — Toward Fast, Data-Efficient Vision- Language-Action Models for Robotic Manipulation. arXiv:2409.12514.

Chapter 16. Fine-tuning a VLA for your robot

• Hu, E. J. et al. (2022). LoRA — Low-Rank Adaptation of Large Language Models. ICLR 2022.
• Kim, M. J. et al. (2024). OpenVLA. arXiv:2406.09246.
• Octo Model Team / Ghosh, D. et al. (2024). Octo. arXiv:2405.12213.
• Wu et al. (2025). A Survey on Efficient Vision-Language-Action Models. arXiv:2510.24795.
• TinyVLA Team (2024). TinyVLA. arXiv:2409.12514.
• Hugging Face (2025). SmolVLA. arXiv:2506.01844.
• Liu et al. (2024). RoboMamba. arXiv:2406.04339.

Chapter 17. Evaluation, safety, and deployment

• García, J., & Fernández, F. (2015). A Comprehensive Survey on Safe Reinforcement Learning. Journal of Machine Learning Research, 16.
• Alshiekh, M. et al. (2018). Safe Reinforcement Learning via Shielding. AAAI 2018.
• Brunke, L. et al. (2022). Safe Learning in Robotics — From Learning- Based Control to Safe Reinforcement Learning. Annual Review of Control, Robotics, and Autonomous Systems, 5.
• Amodei, D. et al. (2016). Concrete Problems in AI Safety. arXiv:1606.06565.
• Garcia et al. (2025). LIBERO-Para. arXiv:2603.28301.
• Ni et al. (2025). Embodied Arena. arXiv:2509.15273.

Chapter 18. Open problems and what comes next

• Zhang et al. (2025). Pure Vision Language Action (VLA) Models — A Comprehensive Survey. arXiv:2509.19012.
• Sapkota, R. et al. (2025). Vision-Language-Action Models — Concepts, Progress, Applications and Challenges. arXiv:2505.04769.
• Xu et al. (2025). An Anatomy of Vision-Language-Action Models — From Modules to Milestones and Challenges. arXiv:2512.11362.
• Wu et al. (2025). A Survey on Efficient Vision-Language-Action Models. arXiv:2510.24795.
• Stanford HAI (2025). AI Index Report 2025 — Robotics chapter.
• Firoozi, R. et al. (2023). Foundation Models in Robotics — Applications, Challenges, and the Future. arXiv:2312.07843.

E.3 Where the mirror is incomplete

Three categories of citation in this book are not held in the offline mirror. Textbooks — Sutton and Barto, Bishop, Strang, LaValle, Featherstone, Murray-Li-Sastry — are commercial works and not mirrored; the reader is expected to either own a copy or to access them through institutional channels. Older pre-arXiv papers — Fikes and Nilsson (1971), Pomerleau (1988), Khatib (1987), Hogan (1985), McDermott et al. (1998) — predate arXiv and are cited by venue; canonical scans exist on author homepages and university repositories and are linked to in SOURCES_INDEX.txt where stable URLs were available at snapshot time. Vendor technical reports — the Figure AI Helix and Helix 02 posts, the BAAI RoboBrain2.0 report, the Microsoft ρα report — are HTML pages that have been snapshotted to sources/html/ with the date of capture recorded; if the live version has since changed, the snapshot is the canonical reference for what is claimed in the book.

A reader who finds a missing reference is asked to open an issue on the book’s companion repository so the next edition’s mirror script can pick it up.