Laplace-domain input/output description of an LTI system ̇𝐱=𝐴𝐱+𝐵𝐮, 𝐲=𝐶𝐱+𝐷𝐮 (extends the resolvent matrix to the forced problem).

Laplace Transform of the Complete System

  • For the complete system, Laplace gives:
    • ℒ︀{̇𝐱=𝐴𝐱+𝐵𝐮}ℒ︀{𝐲=𝐶𝐱+𝐷𝐮}
    • (𝑠𝐼−𝐴)𝐗(𝑠)=𝐱0+𝐵𝐔(𝑠),𝐔(𝑠)=ℒ︀{𝐮(𝑡)}𝐘(𝑠)=𝐶𝐗(𝑠)+𝐷𝐔(𝑠)
  • 𝐗(𝑠)=(𝑠𝐼−𝐴)−1𝐱0⏟𝐗zi(𝑠)+(𝑠𝐼−𝐴)−1𝐵𝐔(𝑠)⏟𝐗zs(𝑠)
  • 𝐱(𝑡)=ℒ︀−1{𝐗(𝑠)}=𝐱zi(𝑡)+𝐱zs(𝑡)
  • with
    • 𝐱zi(𝑡)=ℒ︀−1{(𝑠𝐼−𝐴)−1𝐱0}=𝑒𝐴𝑡𝐱0𝐱zs(𝑡)=ℒ︀−1{(𝑠𝐼−𝐴)−1𝐵𝐔(𝑠)}=ℒ︀−1{(𝑄(𝑠)𝐵𝑟(𝑠))𝐔(𝑠)}
    • (using (𝑠𝐼−𝐴)−1=𝑄(𝑠)𝑟(𝑠))

Transfer Function Matrix

  • Output:
    • 𝐘(𝑠)=𝐶𝐗(𝑠)+𝐷𝐔(𝑠)=𝐶(𝑠𝐼−𝐴)−1𝐱0⏟𝐲zi(𝑡)+[𝐶(𝑠𝐼−𝐴)−1𝐵+𝐷]⏟𝐺(𝑠)𝐔(𝑠)⏟𝐲zs(𝑡)

Transfer Function Matrix (TFM)

  • 𝐺(𝑠)=𝐶(𝑠𝐼−𝐴)−1𝐵+𝐷
  • 𝑝×𝑚 transfer function matrix for the system (TFM)

Impulse Response Matrix

  • We know
    • 𝐲zs(𝑡)=ℒ︀−1{𝐺(𝑠)𝐔(𝑠)}=∫𝑡0𝐺(𝑡−𝜏)𝐮(𝜏)d𝜏
    • 𝐺(𝑡−𝜏): impulse response matrix

Impulse Response Matrix (LTI)

  • 𝐺(𝑡)=ℒ︀−1{𝐺(𝑠)}=𝐶𝑒𝐴𝑡𝐵+𝐷𝛿(𝑡)
  • The 𝑝×𝑚 impulse response matrix is the inverse transform of the TFM

Remember

  • The above is for LTI only! Laplace breaks down for LTV systems, since
    • ℒ︀{𝐴(𝑡)𝐱(𝑡)}≠𝐴(𝑡)𝐗(𝑠)
    • similarly ≠ℒ︀{𝐴(𝑡)}𝐗(𝑠)
  • and the same holds for ℒ︀{𝐵(𝑡)𝐮(𝑡)}, etc.
  • From the modal decomposition of the response (with 𝐱0=𝟎), 𝐺(𝑠) satisfies
    • 𝐘(𝑠)=𝐺(𝑠)𝐔(𝑠)
    • ⇒[𝐺(𝑠)]𝑖𝑗=𝑌𝑖(𝑠)𝑈𝑗(𝑠)
  • Substitute the modal expressions for 𝑌𝑖(𝑠) (𝐜𝖳𝑖 rows of 𝐶, 𝐛𝑗 columns of 𝐵, 𝑑𝑖𝑗 entries of 𝐷, 𝐯𝑘 / 𝐰𝑘 right / left eigenvectors):

TFM in Partial-Fraction Form

  • [𝐺(𝑠)]𝑖𝑗=∑𝑛𝑘=1(𝐜𝖳𝑖𝐯𝑘)(𝐰𝖳𝑘𝐛𝑗)𝑠−𝜆𝑘+𝑑𝑖𝑗
  • Instantly available in partial fraction form!
    • 𝐜𝖳𝑖𝐯𝑘: how much of mode 𝑘 is visible in output 𝑖
    • 𝐰𝖳𝑘𝐛𝑗: strength of the impact of input 𝑗 on mode 𝑘

Invariance Under Coordinate Changes

  • For LTI systems, a change of state coordinates 𝐱=𝑃𝐳 does not change the TFM:
    • 𝐺𝑥(𝑠)=𝐺𝑧(𝑠)
    • i.e. 𝐶(𝑠𝐼−𝐴)−1𝐵+𝐷=𝐶𝑝(𝑠𝐼−𝐴𝑝)−1𝐵𝑝+𝐷𝑝; proof in State Coordinate Transformations