Model Reference Adaptive Control Survey of Control Systems (MEM 800) Presented by Keith Sevcik.
Download ReportTranscript Model Reference Adaptive Control Survey of Control Systems (MEM 800) Presented by Keith Sevcik.
Model Reference Adaptive Control Survey of Control Systems (MEM 800) Presented by Keith Sevcik Concept Model ymodel Adjustment Mechanism Controller Parameters uc Controller u Plant yplant Design controller to drive plant response to mimic ideal response (error = yplant-ymodel => 0) Designer chooses: reference model, controller structure, and tuning gains for adjustment mechanism MIT Rule Tracking error: e y plant ymodel Form cost function: Update rule: 1 2 J ( ) e ( ) 2 d J e e dt sensitivity derivative – Change in is proportional to negative gradient of J MIT Rule Can chose different cost functions EX: J ( ) e( ) d e sign (e) dt 1, e 0 where sign (e) 0, e 0 1, e 0 From cost function and MIT rule, control law can be formed MIT Rule EX: Adaptation of feedforward gain Reference Model Gm ( s) koG( s) ymodel Adjustment Mechanism - s θ Π + Plant uc Π u G p ( s) k G( s) yplant MIT Rule Y ( s) kG( s) where k is unknown For system U ( s) Y (s) koG ( s) Goal: Make it look like U c ( s) using plant Gm (s) koG(s) (note, plant model is scalar multiplied by plant) MIT Rule Choose cost function: Write equation for error: 1 2 d e J ( ) e ( ) e 2 dt e y ym kGU GmU c kGU c koG U c Calculate sensitivity derivative: Apply MIT rule: e k kGUc ym ko d k ' ym e ym e dt ko MIT Rule Gives block diagram: Reference Model Gm ( s) koG( s) ymodel Adjustment Mechanism - s θ Π + Plant uc Π u G p ( s) k G( s) yplant considered tuning parameter MIT Rule NOTE: MIT rule does not guarantee error convergence or stability usually kept small Tuning crucial to adaptation rate and stability. MRAC of Pendulum System J c mgdc sin d1 T (s) d1 2 T ( s) Js cs mgdc d2 dc d1 T ( s) 1.89 2 T ( s) s 0.0389s 10.77 MRAC of Pendulum Controller will take form: Model ymodel Adjustment Mechanism Controller Parameters uc Controller u 1.89 2 s 0.0389s 10.77 yplant MRAC of Pendulum Following process as before, write equation for error, cost function, and update rule: e y plant ymodel 1 2 J ( ) e ( ) 2 d J e e dt sensitivity derivative MRAC of Pendulum Assuming controller takes the form: u 1uc 2 y plant e y plant ymodel G p u Gmuc 1.89 y plant G p u 2 1uc 2 y plant s 0.0389s 10.77 1.891 y plant 2 uc s 0.0389s 10.77 1.89 2 MRAC of Pendulum 1.891 e 2 uc Gmuc s 0.0389 s 10.77 1.89 2 e 1.89 2 uc 1 s 0.0389 s 10.77 1.89 2 e 1.89 2 1 uc 2 2 2 s 0.0389 s 10.77 1.89 2 1.891 2 y plant s 0.0389 s 10.77 1.89 2 MRAC of Pendulum If reference model is close to plant, can approximate: s 0.0389s 10.77 1.89 2 s a1m s a0 m 2 2 a1m s a0 m e 2 uc 1 s a1m s a0 m a1m s a0 m e 2 y plant 2 s a1m s a0 m MRAC of Pendulum From MIT rule, update rules are then: a1m s a0 m d1 e e 2 uc e dt 1 s a1m s a0 m a1m s a0 m d 2 e e 2 y plant e dt 2 s a1m s a0 m MRAC of Pendulum Block Diagram Reference Model bm s a1m s a0 m ymodel 2 - uc Π θ1 + 1.89 s 2 0.0389s 10.77 - a1m s a0 m s a1m s a0 m 2 Π θ2 s Π + e Plant Π s yplant a1m s a0 m s a1m s a0 m 2 MRAC of Pendulum Simulation block diagram (NOTE: Modeled to reflect control of DC motor) omega^2 s+am ym Reference Model Error Saturation 35 4.41 2/26 Step s2 +.039s+10.77 Degrees Degrees to Volts Plant 180/pi y Radians to Degrees -gamma s Theta1 gamma Theta2 s am s+am am s+am MRAC of Pendulum Simulation with small gamma = UNSTABLE! 150 ym g=.0001 100 50 0 -50 -100 0 200 400 600 800 1000 1200 MRAC of Pendulum Solution: Add PD feedback omega^2 s+am ym Reference Model 1.5 du/dt Error D Saturation 35 1 4.41 2/26 Step Degrees s2 +.039s+10.77 P Degrees to Volts Plant 180/pi y Radians to Degrees -gamma s Theta1 gamma Theta2 s am s+am am s+am MRAC of Pendulum Simulation results with varying gammas 45 ym g=.01 g=.001 g=.0001 40 3.56 ym 2 s 2.67 s 3.56 35 30 25 Designed such that : Ts 3 sec .707 20 15 10 5 0 0 500 1000 1500 2000 2500 LabVIEW VI Front Panel LabVIEW VI Back Panel Experimental Results Experimental Results PD feedback necessary to stabilize system Deadzone necessary to prevent updating when plant approached model Often went unstable (attributed to inherent instability in system i.e. little damping) Much tuning to get acceptable response Conclusions Given controller does not perform well enough for practical use More advanced controllers could be formed from other methods – Modified (normalized) MIT – Lyapunov direct and indirect – Discrete modeling using Euler operator Modified MRAC methods – Fuzzy-MRAC – Variable Structure MRAC (VS-MRAC)