d953030747
* feat: add v1/v2 versioning and language selector for mdbook - Copy current content to v1/ directory (1st Edition) - Create v2/ directory with new TOC structure (2nd Edition) and placeholder chapters - Add version selector (V1/V2) and language toggle (EN/ZH) in top-right nav bar - Add build scripts: build_mdbook_v1.sh, build_mdbook_v2.sh - Update assemble_docs_publish_tree.py to support v1/v2 deployment layout - Fix mdbook preprocessor to use 'sections' key (v0.4.43 compatibility) - Update .gitignore for new build artifact directories - Deployment layout: / = v2 EN, /cn/ = v2 ZH, /v1/ = v1 EN, /v1/cn/ = v1 ZH Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> * build: update CI to build and verify all four books (v1/v2 x EN/ZH) - Clarify step names: "Build v2 (EN + ZH)" and "Build v1 (EN + ZH)" - Add verification step to check all four index.html outputs exist - Deploy workflow assembles: / = v2 EN, /cn/ = v2 ZH, /v1/ = v1 EN, /v1/cn/ = v1 ZH Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> * fix: gracefully skip missing TOC entries instead of crashing resolve_toc_target() now returns None for missing files instead of raising FileNotFoundError. This fixes v1 EN build where chapter index files reference TOC entry names that don't match actual filenames. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> --------- Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com>
2.9 KiB
Control System
Although control theory has been firmly rooted in model-based design philosophy, abundant data and machine learning methods have brought new opportunities to control theory. The intersection of control theory and machine learning encompasses a wide range of research directions and applications in various real-world systems.
Linear Quadratic Control
On the theoretical side, Linear-Quadratic Control is a classical control method. If a dynamical system can be represented by a set of linear differential equations and its constraints are quadratic functionals, such problems are called linear-quadratic problems. The solution to such problems is the Linear-Quadratic Regulator (LQR). Recently, there has been research on graph neural networks in distributed linear-quadratic control, which transforms linear-quadratic problems into self-supervised learning problems, enabling the discovery of optimal distributed controllers based on graph neural networks. The researchers also derived sufficient conditions for the stability of the resulting closed-loop systems.
Model Predictive Control
Model Predictive Control (MPC) is an advanced process control method used to control a process while satisfying a set of constraints. The main advantage of MPC is that it allows optimization of the current time step while considering future time steps. Therefore, it differs from the Linear-Quadratic Regulator. MPC also has the ability to predict future events and can take control actions accordingly. Recent research has combined optimal control and machine learning and applied them to visual navigation tasks in unfamiliar environments: for example, a learning-based perception module generates a series of waypoints to guide the robot to the target through collision-free paths, a model-based planner uses these waypoints to generate smooth and dynamically feasible trajectories, and then feedback control executes them on the physical system. Experiments show that compared to purely geometry-based mapping or end-to-end learning-based approaches, this new system can reach the target position more reliably and efficiently.
Stability Analysis of Control Systems
Because safety is critical for robotic applications, some reinforcement learning methods improve safety by learning the uncertainties of dynamics, encouraging methods that are safe, robust, and can formally certify learned control policies. :numref:safe\_learning\_control shows the framework diagram of a Safe Learning Control system. Lyapunov functions are effective tools for evaluating the stability of nonlinear dynamical systems, and recently Neural Lyapunov has been proposed to incorporate safety considerations.
:width:800px
🏷️safe\_learning\_control