"An agent becomes interesting at the exact moment the world refuses to be a dataset."
A Patient Embodied AI Agent
Control for AI Practitioners matters because embodied intelligence is a closed loop. The agent must turn partial observations into useful state, choose actions under uncertainty, and learn from the consequences in a physical or simulated world.
The core move is to connect Control for AI Practitioners to action. A static model can be accurate and still be useless if it cannot support timely, safe, and recoverable behavior.
Chapter Overview
Chapter 7 develops Control for AI Practitioners as a working piece of the embodied AI stack. The chapter starts with the role this topic plays in the sense, represent, predict, decide, act, observe, and learn loop, then turns that role into a concrete implementation pattern.
The practical thread uses python-control, CasADi, Drake, do-mpc, ROS 2 control where appropriate, while the theory thread keeps the mechanism visible. The reader should leave with both a mental model and a build path.
Prerequisites
Readers should be comfortable with Python, tensors, and the perception-action loop. When the chapter uses geometry, control, or probability, the relevant appendices provide a compact refresher.
Chapter Roadmap
- 7.1 Open-loop vs. closed-loop controlBuild the concept, inspect the assumptions, and connect it to tools and evaluation.
- 7.2 Feedback, error, stability, overshoot, oscillationBuild the concept, inspect the assumptions, and connect it to tools and evaluation.
- 7.3 PID control, intuition and tuningBuild the concept, inspect the assumptions, and connect it to tools and evaluation.
- 7.4 State-space control, LQRBuild the concept, inspect the assumptions, and connect it to tools and evaluation.
- 7.5 Model predictive control (MPC) as receding-horizon optimizationBuild the concept, inspect the assumptions, and connect it to tools and evaluation.
- 7.6 Operational-space and whole-body control (preview for humanoids)Build the concept, inspect the assumptions, and connect it to tools and evaluation.
- 7.7 Controllers vs. policies; when learning helps and when it makes control unsafeBuild the concept, inspect the assumptions, and connect it to tools and evaluation.
This chapter uses the right-tool principle. Build the mechanism once, then reach for maintained tools such as python-control, CasADi, Drake, do-mpc, ROS 2 control when the task moves from learning exercise to working system.
Hands-On Lab: Build the Chapter System
Objective
Turn the chapter concept into a small working artifact: define the interface, run a baseline, inspect failure modes, then replace the hand-built part with a library shortcut.
Steps
- Define observations, actions, state, and evaluation metrics.
- Implement the smallest useful version from scratch.
- Run the maintained library version and compare behavior.
- Log success, failure, latency, and robustness.
- Write a short postmortem explaining what changed between the simple version and the practical version.
What's Next?
Continue with Section 7.1: Open-loop vs. closed-loop control, where the chapter moves from motivation to the first concrete idea.
This chapter was strengthened as a full production pass across curriculum, explanation, flow, examples, code, visuals, exercises, cross-references, style, bibliography, controller review, and publication QA. The through-line is feedback, stability, PID, LQR, MPC, operational-space control, policies, and safety filters, always tied to a runnable artifact.
The chapter turns control into a contract for timely correction, so learned policies can be compared with classical feedback on the same plant and safety envelope. This chapter is the control-vs-policy spine for reinforcement learning, MPC, safe deployment, locomotion, humanoids, and safety chapters.
| Tool or Library | What It Handles | Verification Check |
|---|---|---|
| python-control | analyzes linear systems, transfer functions, state-space models, and feedback loops | Verify units, sample time, poles, stability margin, and reference scaling. |
| CasADi | formulates optimization-based controllers with constraints and horizons | Verify constraints, warm start, solver status, and deadline behavior. |
| Drake | models dynamical systems, multibody plants, optimization, and controllers | Verify scalar type, plant finalization, frame convention, and solver status. |
| do-mpc | formulates optimization-based controllers with constraints and horizons | Verify constraints, warm start, solver status, and deadline behavior. |
| ROS 2 control | supports practical work on feedback, stability, PID, LQR, MPC, operational-space control, policies, and safety filters | Verify the library output against the hand-built baseline on one small case. |
Extend the lab by implementing one hand-built baseline, one maintained-library version using python-control, CasADi, Drake, do-mpc, ROS 2 control, and one perturbation test. Save configuration, logs, summary metrics, latency, and two representative failure cases in a single folder.
The recommended teaching rhythm is concept, minimal implementation, library shortcut, diagnostic exercise, then failure analysis. That sequence keeps Control for AI Practitioners attached to an inspectable system artifact rather than treating it as notation alone.
For this chapter, the practical stack is a set of choices, not a shopping list. The hand-built fragment keeps feedback semantics visible. python-control supports linear analysis, CasADi supports nonlinear optimization, Drake supports model-based design and verification, do-mpc packages receding-horizon control, and ROS 2 control connects algorithms to real controller interfaces.
Before leaving the chapter, the reader should be able to state one theory claim, one implementation claim, one evaluation claim, and one realistic failure mode. If any of those four are missing, revisit the lab and the EvidenceRecord exercise.
A strong chapter session ends with an artifact: a script, a trace, a simulator run, a data card, or a reproducible evaluation panel. The artifact is what turns reading into embodied-system-building practice.
Bibliography & Further Reading
Foundational Papers, Tools, and References
Sutton, R. S., and Barto, A. G.. "Reinforcement Learning: An Introduction." (2018). http://incompleteideas.net/book/the-book-2nd.html
A foundation for value functions, policy gradients, exploration, and the RL framing used throughout the book.
Todorov, E., Erez, T., and Tassa, Y.. "MuJoCo: A physics engine for model-based control." (2012). https://mujoco.org/
The simulator lineage behind much modern robot learning, now extended through MJX and Warp workflows.
Brohan, A. et al.. "RT-1: Robotics Transformer for real-world control at scale." (2022). https://arxiv.org/abs/2212.06817
A landmark in large-scale robot policy learning with transformer policies.
Brohan, A. et al.. "RT-2: Vision-Language-Action Models Transfer Web Knowledge to Robotic Control." (2023). https://arxiv.org/abs/2307.15818
A central reference for connecting web-scale VLM knowledge to robot actions.
Open X-Embodiment Collaboration. "Open X-Embodiment: Robotic Learning Datasets and RT-X Models." (2023). https://arxiv.org/abs/2310.08864
The cross-embodiment data and transfer reference used by the data chapters.
Chi, C. et al.. "Diffusion Policy: Visuomotor Policy Learning via Action Diffusion." (2023). https://arxiv.org/abs/2303.04137
The practical diffusion policy reference for imitation learning and continuous action generation.
Hafner, D. et al.. "Mastering Diverse Domains through World Models." (2023). https://arxiv.org/abs/2301.04104
DreamerV3, a modern reference for latent world models and imagination-based control.
Hugging Face. "LeRobot." (2024). https://github.com/huggingface/lerobot
The open robot-learning stack used for datasets, policies, demos, and low-cost embodied AI workflows.
Official documentation and source repositories for Control for AI Practitioners.
Use official docs to check install commands, current APIs, and version caveats before applying Control for AI Practitioners in a lab or project.