Double Inverted Pendulum.pptxVIP

Double Inverted PendulumDigital Control System Wenjun Li Outline Motivation Mathematic Model Optimal Controller Full-Order Observer States Feedback Conclusion Future Work Motivation Apply digital control theory to a real dynamic system Build control-simulation models with Simulink and simulate the results Analyse the results to get some conclusion Mathematic Model Non-linear Multiple variables Absolutely unstable Mathematic Model X1 X2 X3 X4 X5 X6 S(position) ?1(angle) ?2-?1(angel) ?(velocity) ?1(angular velocity) ?2-?1(angular velocity) ? =AX+Bu X(k+1)=GX(k)+Hu(k) Y=Cx Y(K)=CX(k) Linearization at (0,0,0,0,0,0) Continuous to Discrete with sampling time T=5ms Controllable Observerable Unstable Optimal Controller When step k comes to infinite, the performance of system J J=∑[X*(k)QX(k)+u*(k)Ru(k)] The optimal feedback matrix K=[R+H*PH]^-1H*PG Where P is a positive definite matrix and is also the solution of Riccati difference equation P=Q+G*[P^-1+HR^-1H*]^-1G At last the input u(k) would be u(k)=-KX(k) R=1 Q= K=[0.522,15.284,46.185,1.789,6.065,6.579] Optimal Controller Initial state [0 -0.05 0.05 0 0] Full-Order Observer General observer model X(k+1)=FX(k)+GY(k)+Hu(k) Full-Order Observer Original state Estimated states Putting all six eigenvalues of observer at zero , the observer can track states correctlly. States Feedback Combining optimal controller and full–order observer ,we can obtain a regular states-feedback controller States Feedback Output without observer Output with observer Conclusion Optimal controller replaces the original eigenvalues with new eigenvalues to satisfy with some conditions Full-order observer can track states correctly Full-order observer can provide the estimated states to controller to meet requirements Future Work Design controller and observer blocks Implement blocks to some more complex cases Thank you