EE611 Deterministic Systems Realizations Kevin D. Donohue Electrical and Computer Engineering University of Kentucky Minimal Realizations Realizations are not unique, they may have different bases and different orders while still realizing the same transfer matrix. A minimal realization is one with the smallest possible dimension. An important condition for transfer functions if that of being coprime. g s= y s N s = u s D s For a SISO TF N(s) and D(s) are coprime iff they have no common factors. Controllable Canonical Form Consider SISO system with strictly proper TF: 3 2 1 s 2 s 3 s4 y s N s g s= = 4 = u s s 1 s3 2 s 2 3 s 4 D s this corresponds to: 4 3 2 3 2 d y d y d y dy d u d u du y= 4 u 1 2 3 4 1 2 3 4 3 2 3 2 dt dt dt dt dt dt dt State variables can defined as derivatives of y such that derivatives of the input are accounted for. Define auxiliary variable v such that: −1 y s= N s D s u s= N s v s where: −1 D s u s= v s u s= D s v s State Variable Definition Define state variables from v as: and ẋ 1 is determined from u s= D s v s ẋ 1=u−1 x1 −2 x 2−3 x3 −4 x 4 and y is obtained from x 4=v x 3 = ẋ 4 =v̇ x 2= ẋ 3 =v̈ x 1 = ẋ 2 =v y s= N s v s y=1 x 12 x 2 3 x3 4 x 4 Controllable Canonical Form Can be realized with [][ ẋ 1 −1 − 2 −3 −4 ẋ 2 0 0 0 = 1 ẋ 3 0 1 0 0 0 0 1 0 ẋ 4 y t =[ 1 2 3 4 ] ][ ] [ ] x1 1 x2 0 ut 0 x3 0 x4 [] x1 x2 x3 x4 Controllable Canonical Form Can be realized with β 1 β2 β3 u(t) + ẋ1 1 s −α1 1 ẋ 2 s 1 ẋ 3 s 1 ẋ 4 s −α2 β4 + y(t) −α3 −α4 Minimal Realizations y s N s = Given a SISO TF g s= : u s D s The controllable canonical form is observable iff N(s) and D(s) are coprime. State equation {A, b, c, d} is a minimal realization of proper rational TF g s iff {A, b} is controllable and {A, c} is observable or iff dim [A] = deg[g s]. All minimal realization of g s are equivalent. Controllable Canonical Form Consider SISO system with strictly proper TF: 3 2 1 z 2 z 3 z4 y z N z g z= = 4 = u z z 1 z 3 2 z 2 3 z 4 D z this corresponds to: y [n4 ] 1 y [n3 ] 2 y [n2] 3 y [n1] 4 y [n] =1 u [n3]2 u [n2]3 u [n1 ]4 u [n ] Can define state variables as time delays on y such that time delays of the input are accounted for. Define auxiliary variable v[n] such that: −1 y z = N z D z u z = N z v z where: −1 D z u z= v z u z= D z v z State Variable Definition x 4 [n]=v [n ] x 3 [n ]= x 4 [n1]=v [n1] x 2 [n ]= x 3 [n1]=v [n2] x 1 [n ]= x 2 [n1 ]=v [n3] Define state variables from v as: and x 1 [n1] is determined from u z= D z v z x 1 [n1]=u[n]−1 x 1 [n]−2 x2 [n]−3 x3 [n]− 4 x 4 [n] and y is obtained from y z= N z v z y [n]=1 x1 [n]2 x 2 [n]3 x3 [n]4 x 4 [n] Controllable Canonical Form Can be realized with [ ][ x 1 [n1] −1 −2 −3 − 4 x 2 [n1] 0 0 0 = 1 x 3 [n1] 0 1 0 0 0 0 1 0 x 4 [n1] y [n]=[ 1 2 3 4 ] ][ ] [ ] [] x1 [n] x 2 [n] x3 [n] x 4 [n] x1 [n] 1 x 2 [n] 0 u[n] 0 x3 [n] 0 x 4 [n] Controllable Canonical Form Can be realized with β 1 β2 β3 u[n] x1 [n1] + x 2 [n1] -1 x 4 [n1] -1 -1 Z −α1 x 3 [n1] -1 Z Z Z −α2 β4 + y[n] −α3 −α4 Observable Canonical Form Given an observable system ẋ=A xb ut y t =c x with characteristic polynomial s=∣s I−A∣=sn1 s n−1 2 sn−2... n−1 s n system can be transformed into observable canonical form by x =P x [ n−1 n−2 n−2 n−3 −1 P=Q = ⋮ ⋮ 1 1 1 0 ... ... ⋰ ... ... 1 1 ⋮ 0 0 ][ ] c 1 cA 0 c A2 ⋮ c A3 0 ⋮ 0 n−1 cA Observable Canonical Form [ ][ 0 0 ... ̇x 1 1 0 ... ̇x 2 ⋮ =⋮ ⋮ ⋱ 0 0 ... ̇x n−1 0 0 ... ̇x n ][ ] [ ] [] 0 − n n x 1 0 − n−1 x 2 n−1 ⋮ ⋮ ⋮ ⋮ ut 0 − 2 x n−1 2 1 − 1 1 x n x 1 x 2 y t = [ 0 0 ... 0 1 ] ⋮ x n−1 x n n−1 n−2 1 s 2 s ⋯n−1 sn y s g s= = u s sn 1 sn−1 2 sn−2 ⋯ n−1 s n Computing Coprime Fractions Consider equivalent proper rational polynomials: N s N s g s= = D s Ds N(s) and D(s) are not coprime iff ∃ a D s ∋ deg [ D s] < deg[D(s)]. Let the coefficients of D s and N s be unknowns and their polynomial degrees one less than deg[D(s)]. The above equation can be multiplied out and like terms equated to form a set of linear equations (equal to zero). The coefficient matrix in terms of coefficients N(s) and D(s) with unknowns coefficients of D s and N s , form the Sylvester resultant. Coprime and Sylvesters Resultant Consider equivalent proper rational polynomials: N s N s g s= = D s Ds N(s) and D(s) are coprime iff Sylvesters resultant is nonsingular. s D N s Lecture Note Homework U10.1 Implement the system below in Simulink using delays, summers, and gains. [ ] [] 0.5 −0.25 1 x [k 1]= x[ k ] u[ k ] 0 0.1 2 y [ k ]= [ 1 0 ] x [k ] Assume the system is relaxed at k = 0. A) Plot the output when the input is the unit impulse. B) Plot the output when the input is the unit step. Lecture Note Homework U10.2 Find a minimal realization of the system below. Implement the system below in Simulink using integrators, summers, and gains. 2 y s N s s 3s2 g s= = 6 = u s s 7.5 s5 31.5 s4 84 s3123 s 284 s20 D s Assume the system is relaxed at t = 0. A) Plot the output when the input is the unit impulse. B) Plot the output when the input is the unit step.