State Variable Description of LTI systems Eytan Modiano Eytan Modiano Slide 1 Learning Objectives • Understand concept of a state • Develop state-space model for simple LTI systems – – – RLC circuits Simple 1st or 2nd order mechanical systems Input output relationship • Develop block diagram representation of LTI systems • Understand the concept of state transformation – Eytan Modiano Slide 2 Given a state transformation matrix, develop model for the transformed system The State of a System • The “state” of a system is the minimum information needed about the system in order to determine its future behavior – Given the state at time t0, and input up to time t > t0; can determine the output for time t. • State Variables – Set of variables of smallest possible size that together with any input to the system is sufficient to determine the future behavior (I.e., output) of the system. – Each state variable has “memory” E.g., voltage in capacitor – Each state variable has an “initial condition” E.g., its state at time t0 – State variables are typically associated with energy storage • State vector: vector of state variables Eytan Modiano Slide 3 Example: RLC circuit e1 R C ic + V(t) - dv " !!ic = C V (t) $ dv ! =! dt # & dt RC !ic = !V / R $% V (t) = V (0)e!t / RC • V(t) is the state of the system at time t – • Initial condition - V(0) is the voltage across the capacitor at time 0 V(t) V(0) If we know v(t) at any time t, we know it for all future time – No input in this case t Eytan Modiano Slide 4 State Variables • In electric circuits, the energy storage devices are the capacitors and inductors – – They contain all of the state information or “memory” in the system State variables: Voltage across capacitors Current through inductors • In mechanical systems, energy is stored in springs and masses – State variables Spring displacement Mass position and velocity • Example” Single mass M, moving in one dimension (x), under force F – State variables (x1, x2) x1 = mass position x2 = velocity F Eytan Modiano Slide 5 M x1 = x ! x!1 = x2 "$ x2 = x! # x!2 = !! x=F/M F = M!! x Position (x) State Equations • State equations in matrix form: ! ! x1 $ x = # & ,!!input :!u = F / M " x2 % State!equations :! x" = Ax + Bu ! x"1 $ ! 0 1 $ ! x1 $ ! 0 $ # x" & = # 0 0 & # x & + # 1 & F / M " 2% " %" 2% " % • Output equations: – suppose output is y=x1 (position) ! x1 $ y = [1 0 ] # & " x2 % x! = Ax + Bu y = Cx + Du • General form of state equations: • We will focus (to start) mainly on homogeneous case: Eytan Modiano Slide 6 x! = Ax General form of state equations • In general a system can have – n states, state vector X(t), m inputs, u(t), and l outputs, y(t) ! x1 (t) $ ! u1 (t) $ ! y1 (t) $ ! ! ! # & # & X(t) = # " & ,!!!U(t) = # " & ,!!Y (t) = ## " && #" xn (t) &% #"un (t) &% #" yn (t) &% • The state and output equations can be written as: • A,B,C,D are constant matrices " " " X! = AX + BU " " " Y = CX + DU – – – – Eytan Modiano Slide 7 A is an nxn “system dynamics” matrix B is an nxm “input matrix” C is an lxn matrix relating states to outputs D is an lxm matrix relating inputs to outputs Why the state-space approach? • Very general approach to describe Linear time-invariant (LTI) systems – – – • Rich theory describing the solutions Simplifies analysis of complex systems with multiple inputs and outputs Approach is central to “modern” control History of state-space approach – State-space approach to control system design was introduced in the 1950’s Up to that point “classical” control used root-locus or frequency response methods (more in 16.060) Eytan Modiano Slide 8 – “New” approach was named “modern” control and still have that name – Related state space approach to describing differential equations is over 100 years old Block Diagram Representation Integrator block diagram $ ∫ x(t) "# dx(t) = ax(t) + bu(t) dt x(t) = $ t "# t ax(! ) + bu(! )d! • • x(! )d! x(t) - state variables u(t) - inputs System block diagram u(t) b + ! x(t) ∫ x(t) a Eytan Modiano Slide 9 State is “fed-back” into the system x(t) x(t) ! integrator!output ! ! integrator!input x(t) ! = ax(t) + bu(t) x(t) General system block diagram " " " X! = AX + BU " " " Y = CX + DU u(t) b + " ! x(t) d ∫ ! x(t) c + a Force-mass example: F Eytan Modiano Slide 10 1/M x1 = x ! x!1 = x2 "$ x2 = x! # x!2 = !! x = x!1 = F / M x!2 (t) ∫ x2 (t) = x!1 ∫ x1 (t) ! y(t) State Transformation • The state variable description of a system is not unique • Different state variable descriptions are obtained by “state transformation” – – • New state variables are weighted sum of original state variables Changes the form of the system equations, but not the behavior of the system Some examples: original system ~ x1(t), x2(t) – Transformed systems ~ z1(t), z2(t) (1) z1(t) = x2(t), z2(t)=x1(t) (2) z1(t) = x1(t)+x2(t), z2(t)=x2(t) (3) z1(t) = 2x1(t) - x2(t), z2(t) = x1(t)+2x2(t) • Eytan Modiano Slide 11 State transformation can simplify system description and analysis State Transformation, continued ! • In general, we can transform x " " x! = Tx • " to a new state vector x! by, Where T is the state transformation matrix ! • Relationship between x " and x! must be one-to-one (i.e., the mapping must be invertible) ⇒ T must be non-singular ⇒ T-1 must exist original!system :!!!!! x! = Ax + Bu Transformed!system :! x" = Tx ! x"! = Tx! = TAx + TBu ! x"! = TAT "1 x + TBu!!(recall :!!x = T "1 x" ) !! A" = TAT "1 ,!!! B" = TB " + Bu " !! x"! = Ax Eytan Modiano Slide 12 State Transformation (example) • Try to write down the state equations explicitly • Notice that the state transformation is a linear combination of the original system states • Notice that the new transformed system has a much simpler (to understand) structure – – Eytan Modiano Slide 13 Two Decoupled first order differential equations The system is still the same - just the description is simpler Original!system:!!! x! = Ax + Bu " !1 4 % "2 % A=$ ,!!!B = $ ' ' # 4 1& #4 & State!transformation: !x1 + x2 x1 + x2 x"1 = ,! x" 2 = 2 2 1 " !1 1% " !1 1% !1 T= $ ,!!T = $ ' ' 1 1 2 # 1 1& # & " !5 0 % " "1 % !1 " A = TAT = $ ,!! B = TB = $ ' ' # 0 3& # 3& x"!1 = !5 x"1 + u x"!2 = 3x" 2 + 3u State variable description of RLC circuits y(t) u(t) • • • • Eytan Modiano Slide 14 + - R1 e1 i1 + v1 - e2 i2 R2 C1 C2 + v2 - d i1 (t) = C1 v1 (t) dt d i2 (t) = C2 v2 (t) dt Input: voltage source ~ u(t) System state: capacitor voltages ~ v1(t), v2(t) Output: current through resistor R1 ~ y(t) Node equations: v1 ! u(t) v1 ! v2 dv1 v2 ! v1 u(t) ! v1 e1 :!! + i1 + =0" = ! R1 R2 dt C1 R2 C1 R1 v2 ! v1 dv2 v2 ! v1 v1 ! v2 e2 :!! + i1 = 0 " =! = R2 dt C2 R2 C2 R2 Example, continued " 1 1 % 1 1 v!1 = !v1 $ + + v2 + u(t) ' C1 R2 C1 R1 # C1 R1 C1 R2 & 1 1 v!2 = !v2 + v1 C2 R2 C2 R2 1 ( 1 *! C R ! C R 1 1 1 2 V! = * 1 * * C2 R2 ) Eytan Modiano Slide 15 1 + ( 1 + C1 R2 - V + * C1 R1 - u(t) * 1 ! *) 0 -, C2 R2 , ( !1 + (1+ u(t) ! v1 y(t) = .Y = * 0 - V + * - u(t) R1 ) R1 , ) R1 ,