Linear Quadratic Regulator (LQR) Control Design Tutorial

Added:

Robust H-infinity (H∞) Control and the Challenge of Model Uncertainty

While the Linear Quadratic Regulator (LQR) offers an optimal control solution under perfect model conditions, its primary limitation is its vulnerability to model uncertainty and external disturbances. LQR optimizes for average performance based on user-defined Q and R weighting matrices but does not explicitly account for system variations or unmodeled dynamics. In real-world applications, a perfect mathematical model is rarely available. Robust Control, particularly H-infinity (H∞) control, arises as a major alternative. Instead of optimizing average-case performance, H∞ control designs for the worst-case scenario, mathematically ensuring stability and performance despite bounded model perturbations and noise. Additionally, while LQR assumes full-state feedback, real systems often require observers; when LQR is paired with a Kalman Filter (LQG), it famously loses its guaranteed stability margins. H∞ and other robust control methods systematically address these safety-critical vulnerabilities, making them essential for high-reliability engineering.

State-space representation of dynamical systems, including state equations, input vectors, and output equations.
Fundamentals of state feedback control design and the concept of pole placement.
System controllability and observability to ensure the physical viability of feedback control.
Basic linear algebra, particularly quadratic forms, eigenvalues, and positive (semi-)definite matrices.
Linear Quadratic Gaussian (LQG) control, which combines LQR with a Kalman Filter for systems with noise and unmeasurable states.
Solving the Continuous and Discrete Algebraic Riccati Equations (ARE) both analytically and numerically.
Model Predictive Control (MPC) to extend LQR concepts to systems with strict state and control input constraints.
Robust control theories, such as H-infinity control, to handle model parameter uncertainties and external disturbances.
148.5K views3.2Klikes1:36:06@ChristopherLumOriginal Release: 2018-12-03

The Linear Quadratic Regulator (LQR) is a powerful control design technique that computes the optimal full-state feedback gain matrix K by solving an optimization problem, where the controller minimizes the cost function J = ∫₀^∞ (xᵀQx + uᵀRu) dt subject to the system dynamics ẋ = Ax + Bu; the matrices Q (symmetric positive semi-definite) and R (symmetric positive definite) serve as tuning parameters that allow engineers to trade-off between state regulation and control effort, with the optimal controller given by u = -Kx where K = R⁻¹BᵀS and S satisfies the Algebraic Riccati Equation AᵀS + SA - SBR⁻¹BᵀS + Q = 0.