an introduction to
play

An INTRODUCTION to SWITCHING ADAPTIVE CONTROL Daniel Liberzon - PowerPoint PPT Presentation

An INTRODUCTION to SWITCHING ADAPTIVE CONTROL Daniel Liberzon Coordinated Science Laboratory and Dept. of Electrical & Computer Eng., Univ. of Illinois at Urbana-Champaign Based on joint work with J.P. Hespanha (UCSB) and A.S. Morse


  1. An INTRODUCTION to SWITCHING ADAPTIVE CONTROL Daniel Liberzon Coordinated Science Laboratory and Dept. of Electrical & Computer Eng., Univ. of Illinois at Urbana-Champaign Based on joint work with J.P. Hespanha (UCSB) and A.S. Morse (Yale) UTC-IASE online seminar, July 21-22, 2014 1 of 34

  2. SWITCHING CONTROL y u Plant: P y u P Classical continuous feedback paradigm: C y u P But logical decisions are often necessary: C 1 C 2 l o g i c 2 of 34

  3. REASONS for SWITCHING • Nature of the control problem • Sensor or actuator limitations • Large modeling uncertainty • Combinations of the above 3 of 34

  4. REASONS for SWITCHING • Nature of the control problem • Sensor or actuator limitations • Large modeling uncertainty • Combinations of the above 4 of 34

  5. MODELING UNCERTAINTY unmodeled dynamics parametric   0 uncertainty Also, noise and disturbance Adaptive control (continuous tuning) vs. supervisory control (switching) 5 of 34

  6. EXAMPLE Scalar system: , otherwise unknown (purely parametric uncertainty) stable  not implementable Controller family: Could also take controller index set 6 of 34

  7. SUPERVISORY CONTROL ARCHITECTURE Supervisor candidate controllers  u 1 Controller y u u 2 Plant Controller . . . u m Controller . . .  – switching signal, takes values in – switching controller 7 of 34

  8. TYPES of SUPERVISION • Prescheduled (prerouted) • Performance-based (direct) • Estimator-based (indirect) 8 of 34

  9. TYPES of SUPERVISION • Prescheduled (prerouted) • Performance-based (direct) • Estimator-based (indirect) 9 of 34

  10. OUTLINE • Basic components of supervisor • Design objectives and general analysis • Achieving the design objectives (highlights) 10 of 34

  11. OUTLINE • Basic components of supervisor • Design objectives and general analysis • Achieving the design objectives (highlights) 11 of 34

  12. SUPERVISOR  y 1 e 1 y   Multi- y 2 e 2 estimation errors: .  . . Estimator  e p y p u .  . . . . . Want to be small Then small indicates likely 12 of 34

  13. EXAMPLE Multi-estimator: exp fast => 13 of 34

  14. EXAMPLE disturbance Multi-estimator: exp fast => 14 of 34

  15. STATE SHARING Bad! Not implementable if is infinite The system produces the same signals 15 of 34

  16. SUPERVISOR   y 1 e 1 y 1   Monitoring  Multi- y 2 e 2 2 .  . Signals . Estimator   e p y p u Generator p .  . . . . . . . . Examples: 16 of 34

  17. EXAMPLE Multi-estimator: – can use state sharing 17 of 34

  18. SUPERVISOR   y 1 e 1 y 1   Monitoring  Multi- y 2 e 2  Switching 2 .  . Signals . Estimator Logic   e p y p u Generator p .  . . . . . . . . Basic idea: , controllers: Justification? Plant gives stable small => small => plant likely in => closed-loop system (“certainty equivalence”) 18 of 34

  19. SUPERVISOR   y 1 e 1 y 1   Monitoring  Multi- y 2 e 2  Switching 2 .  . Signals . Estimator Logic   e p y p u Generator p .  . . . . . . . . Basic idea: , controllers: Justification? Plant gives stable small => small => plant likely in => closed-loop system only know converse! small => Need: gives stable closed-loop system This is detectability w.r.t. 19 of 34

  20. DETECTABILITY Linear case: plant in closed loop with view as output Want this system to be detectable “output injection” matrix 9 L q : A q ¡ L q C q is Hurwitz x = (A q ¡ L q C q )x + L q e q _ asympt. stable 20 of 34

  21. SUPERVISOR   y 1 e 1 y 1   Monitoring  Multi- y 2 e 2  Switching 2 .  . Signals . Estimator Logic   e p y p u Generator p .  . . . . . . . . We know: is small Switching logic (roughly): This (hopefully) guarantees that is small small => stable closed-loop switched system Need: This is switched detectability 21 of 34

  22. DETECTABILITY under SWITCHING plant in closed Switched system: loop with view as output Want this system to be detectable: Assumed detectable for each frozen value of Output injection: need this to be asympt. stable Thus needs to be “ non-destabilizing ” : • switching stops in finite time • slow switching (on the average) 22 of 34

  23. SUMMARY of BASIC PROPERTIES Multi-estimator: 1. At least one estimation error ( ) is small • when • is bounded for bounded & Candidate controllers: 2. For each , closed-loop system is detectable w.r.t. Switching logic: 3. is bounded in terms of the smallest 4. Switched closed-loop system is detectable w.r.t. provided this is true for every frozen value of : for 3, want to switch to conflicting for 4, want to switch slowly or stop 23 of 34

  24. SUMMARY of BASIC PROPERTIES Multi-estimator: 1. At least one estimation error ( ) is small • when • is bounded for bounded & Candidate controllers: 2. For each , closed-loop system is detectable w.r.t. Switching logic: 3. is bounded in terms of the smallest 4. Switched closed-loop system is detectable w.r.t. provided this is true for every frozen value of 1 + 3 => is small Analysis: => state is small  2 + 4 => detectability w.r.t. 24 of 34

  25. OUTLINE • Basic components of supervisor • Design objectives and general analysis • Achieving the design objectives (highlights) 25 of 34

  26. CANDIDATE CONTROLLERS y fixed Plant u  Controller y e Multi- q q  estimator y 26 of 34

  27. CANDIDATE CONTROLLERS y fixed Plant P u  Controller y e Multi- q q  estimator C  y E  Linear: overall system is detectable w.r.t. if i. system inside the box is stable ii. plant is detectable => , , Need to show: C E P => => => => C , P E ii i 27 of 34

  28. CANDIDATE CONTROLLERS y fixed Plant P u  Controller y e Multi- q q  estimator C  y E  Linear: overall system is detectable w.r.t. if i. system inside the box is stable ii. plant is detectable Nonlinear: same result holds if stability and detectability are interpreted in the ISS / OSS sense: external signal 28 of 34

  29. CANDIDATE CONTROLLERS y fixed Plant P u  Controller y e Multi- q q  estimator C  y E  Linear: overall system is detectable w.r.t. if i. system inside the box is stable ii. plant is detectable Nonlinear: same result holds if stability and detectability are interpreted in the integral-ISS/OSS sense: 29 of 34

  30. CANDIDATE CONTROLLERS y fixed Plant P u  Controller y e Multi- q q  estimator C  y E  Linear: overall system is detectable w.r.t. if i. system inside the box is stable ii. plant is detectable For minimum-phase plants, it is enough to ask that the system inside the box be output-stabilized 30 of 34

  31. SWITCHING LOGIC: DWELL-TIME – dwell time – monitoring signals Initialize Wait time units Find ? no yes Detectability is preserved if is large enough  Obtaining a bound on in terms of is harder Not suitable for nonlinear systems (finite escape) 31 of 34

  32. SWITCHING LOGIC: HYSTERESIS – hysteresis constant – monitoring signals Initialize Find ? yes no or (scale-independent) 32 of 34

  33. SWITCHING LOGIC: HYSTERESIS Initialize Find ? yes no finite, bounded switching stops in finite time => This applies to exp fast, Linear, bounded average dwell time => 33 of 34

  34. TOY EXAMPLE: PARKING PROBLEM x 2 x    cos p 1 w 1 1 x    w sin p 1  2 1    w p 2 2 x 1 Unknown parameters correspond to the 1 , p p 2 radius of rear wheels and distance between them 34 of 34

Recommend


More recommend