Tracking of LTI Systems with Unstable Zero Dynamics using Sliding Mode Control S. Janardhanan
Problem Statement Consider the nth order system with relative degree of r, What is relative degree?
Two-Part Control – Nominal Control Assuming the system output is following the reference trajectory Thus the nominal control is
Two-Part Control – Incremental Control The incremental input Du(k) is designed in such a way as to stabilize the error dynamics, and bring the states to the reference. A error based incremental DSMC control Du(k) = FDx(k) + gsign(s(k)) s(k)=cT Dx(k) is used here
Two Part Control Thus, the total control input (state feedback based) would be
Determination of Nominal Zero Dynamics Trajectory (x2,0) When the reference is being tracked perfectly, the system dynamics is of the form without knowledge of initial state of x2,0, the future value cannot be calculated
Diagonal Transformation Define a transformation T2 such that the following are satisfied.
Reference Computation Thus we calculate the zero dynamics reference as
Assumptions, Restrictions … It is assumed that z(k)=0, k<0 It is assumed that z(k) is such that it keeps the marginally stable dynamics bounded It is assumed that the ‘flow’ of z(k) is pre-determined and known. The zero dynamics reference are null in both infinite past and infinite future.
Bounded Preview The calculation of the xu dynamics requires z(k) up to infinity : impractical Algorithm is modified for a ‘bounded preview’. This keeps error in zero dynamics reference.
Output Feedback Version The multirate output feedback version of the aforementioned control algorithm is of the form
Numerical Example to track
Simulation Results : Control Input
Simulation Results : Reference and Response
Simulation Results : Tracking Error
Simulation Results : Bounded Zero Dynamics