计算机控制理论第十一章.pptVIP

Chapter 11Optimal Design Methods:A State-Space Approach 11.1 Problem Formulation The Process The Criterion Admissible Control Laws Full state information is assumed. 11.2 Linear Quadratic Control(LQ Control) The Deterministic Case Theorem 11.1 LQ-control of a deterministic system Dynamic programming, Richard E. Bellman 1957 An optimal trajectory on the time interval [T1, T3] must be optimal also on the subintervals [T1, T2] and [T2, T3]. Dynamic programming in linear quadratic control Let S(k) be defined by Dynamic programming with T1= k ? 1, T2 =k, T3=N gives The completion of squares calculation then gives Jocopo Francesco Riccati, 1676–1754 Solution via the Riccati equation Define S(k) by the Riccati equation Then the control law u(k)=?L(k)x(k) minimizes the cost.The minimum is · Time-varying controller · Often or special LQ – First order system: x(k +1)= x(k)+u(k) Loss function defined by q1, q2, q0, and N Riccati equation Controller The stochastic LQ control problem Minimize subject to where v is white noise with Time-varying stochastic LQ control Minimize subject to where Dynamic programming again Define S(k) by the Riccati equation and Mean value of quadratic form Assume E x = 0 and cov x = R Solution to stationary problem Minimize subject to where v is white noise with How to solve the LQ problem in Matlab Use Matlab! [L,S,E] = DLQR(A,B,Q,R,N) Linear quadratic regulator design for discrete-time systems. u[n] = -Lx[n] minimizes the cost function J = Sum {xQx + uRu + 2*xNu} subject to the constraint equation: x[n+1] = Ax[n] + Bu[n] E = EIG(A-B*L) lqrd solves LQ starting with continuous-time loss function. Theorem: Stability of closed-loop system Assume that is positive definite and that there exists a positive-definite steady-state solution S to the algebraic Riccati equation. Then u(k)=?Lx(k) gives an asymptotically stable closed- loop system Sampling a loss function If u(t) is constant over the