Full Order Observer Design Using Simulink
David Pyne EE 692
Goal of the Project
To design a Simulink library block that automatically generates a Full Order Observer for a given linear dynamical system Desired Characteristics:
Scaleable Practical Easy to Use
Agenda
What is a Full Order Observer? Overview of Linear Dynamical Systems Library Block Concept Design of the Observer Simulink Implementation A Simple Example A More Complex Example Questions
What is a Full Order Observer?
A mathematical model of the entire dynamical system An estimator of the unmeasurable states
Flexible Scaleable Highly accurate
Overview of Linear Dynamical Systems
Systems characterized by the following model:
x1 . . . xn
MIMO
y1 . . . ym
Overview of Linear Dynamical Systems
The concept of State
Key
system attributes The minimum number of measurements needed Not always consistent
Library Block Concept
Systems in Simulink
A series of function blocks Connected like a circuit Used to model dynamical systems
Library Block Concept
Example system
Library Block Concept
Inputs
System A and B matrices Roots of observer Preconfigured Simulink block
Outputs
Design of the Observer
The full order observer equation
K0 is to be designed by the user
Eigenvalues of the characteristic error dynamics polynomial
Design of the Observer
Define the error dynamics equation
Simulink Implementation
Design Tasks
Implement the observer equation using existing Simulink blocks
Create a subsystem by grouping smaller units
Create the user interface
Validate user input Load user matrices into observer equation
Verify correct output
Simulink Implementation
Simulink Implementation
Observer user interface Requires limited knowledge of observer design by the user
Simulink Implementation
Validation of user input A must be square B must have same number of rows as A C is assumed to be of the form [1 0 0 0] K must be the same length as C
The system must be completely observable
Simulink Implementation
Complete observability check
Complete controllability check
Simulink Implementation
Check for repeated roots in K vector
If root multiplicity is greater than the rank of C use ACKER()
Not terribly reliable Breaks down for higher order systems
Otherwise use PLACE()
Much more robust Based on algorithm designed by Kautsky and Nichols
Simulink Implementation
Block inputs
y Scalar system output u System control command
Any errors are warnings for the user Command line output of the calculated K0 matrix Vector estimate of system state
Block outputs
A Simple Example
A Simple Example
Output from new block tracks system exactly after locking on
A Simple Example
Error between actual and estimate converges to zero
A More Complex Example
A More Complex Example
A More Complex Example
A More Complex Example
A More Complex Example
A More Complex Example
Output from new block tracks system exactly after locking on
A More Complex Example
Error between actual and estimate converges to zero
Summary
A Simulink full order observer library block was created
Accurate Easy to use Scaleable
Saves the modern control designer (or student) time Reduces the pain and suffering inherent in the design for higher order systems
References
Dorf, R.C. and Bishop, R.H.,Modern Control Systems, Tenth Edition, New Jersey:Pearson, Prentice Hall Publishing, 2004 Johnson, C.D., Stabilization of Linear Dynamical Systems with Respect to Arbitrary Linear Subspaces, Journal of Mathematical Analysis and Applications, Vol. 44 (1973), No. 1, pp. 175-185 Johnson, C.D., A Unified Canonical Form For Controllable and Uncontrollable Linear Dynamical Systems, International Journal of Control, Vol. 13 (1971), No. 3, pp. 497-517 Kalman, R.E., Mathematical Description of Linear Dynamical Systems, SIAM Journal of Controls, Vol. 1 (1963), pp. 152-192 Kautsky, J. and N.K. Nichols, "Robust Pole Assignment in Linear State Feedback," International Journal of Control, Vol. 41 (1985), pp. 1129-1155 Kolman, B and Hill, D.R., Linear Algebra With Applications, Seventh Edition, New Jersey:PearsonPrentice Hall Publishing, 2001
Questions?
