Region of Attraction Comparison for Gradient Projection Anti-windup Compensated Systems Justin Teo1 1 DSO Jonathan P. How2 National Laboratories, Singapore 2 Aerospace Controls Laboratory Department of Aeronautics & Astronautics Massachusetts Institute of Technology 50th IEEE Conf. Decision & Control and European Control Conf. December 14, 2011 J. Teo, J. P. How (DSO, ACL–MIT) ROA Comparison for GPAW Compensated Systems Dec. 14, 2011 1 / 12 Outline Outline 1 Motivation 2 GPAW Overview 3 General ROA Comparison 4 ROA Comparison for GPAW Compensated Systems 5 Conclusions and Future Work J. Teo, J. P. How (DSO, ACL–MIT) ROA Comparison for GPAW Compensated Systems Dec. 14, 2011 2 / 12 Motivation “Conventional” Anti-windup Problem r controller u sat(·) plant anti-windup compensator Given nominal plant and controller J. Teo, J. P. How (DSO, ACL–MIT) ROA Comparison for GPAW Compensated Systems Dec. 14, 2011 3 / 12 Motivation “Conventional” Anti-windup Problem r controller u sat(·) plant − anti-windup compensator Given nominal plant and controller, design anti-windup compensator such that [Tarbouriech and Turner 2009]: unconstrained response recovered when no controls saturate performance improved when some controls saturate region of attraction (ROA) maximized when r ≡ 0 J. Teo, J. P. How (DSO, ACL–MIT) ROA Comparison for GPAW Compensated Systems Dec. 14, 2011 3 / 12 Motivation “Conventional” Anti-windup Problem r controller u sat(·) plant − anti-windup compensator Given nominal plant and controller, design anti-windup compensator such that [Tarbouriech and Turner 2009]: unconstrained response recovered when no controls saturate performance improved when some controls saturate region of attraction (ROA) maximized when r ≡ 0 Question Can some solutions (corresponding to some initial conditions) that were stable (uncompensated) become unstable with anti-windup? J. Teo, J. P. How (DSO, ACL–MIT) ROA Comparison for GPAW Compensated Systems Dec. 14, 2011 3 / 12 Motivation Maximizing ROA Not Adequate r controller u sat(·) ROA: Rn anti-windup compensator r controller plant u uncompensated sat(·) plant compensated Rn − anti-windup compensator J. Teo, J. P. How (DSO, ACL–MIT) ROA: Raw ROA Comparison for GPAW Compensated Systems Dec. 14, 2011 4 / 12 Motivation Maximizing ROA Not Adequate r controller u sat(·) ROA: Rn anti-windup compensator r controller plant u sat(·) − anti-windup compensator J. Teo, J. P. How (DSO, ACL–MIT) Raw plant ROA: Raw Rn Maximizing ROA not adequate Want Raw ⊃ Rn ROA Comparison for GPAW Compensated Systems Dec. 14, 2011 4 / 12 Motivation Maximizing ROA Not Adequate r controller u sat(·) ROA: Rn anti-windup compensator r controller plant u sat(·) − anti-windup compensator Raw plant ROA: Raw Rn Maximizing ROA not adequate Want Raw ⊃ Rn Or Raw ⊃ Ω, where Ω ⊂ Rn is an ROA estimate J. Teo, J. P. How (DSO, ACL–MIT) ROA Comparison for GPAW Compensated Systems Dec. 14, 2011 4 / 12 Motivation Maximizing ROA Not Adequate r controller u sat(·) ROA: Rn anti-windup compensator r controller plant u sat(·) − anti-windup compensator Raw plant ROA: Raw Rn Maximizing ROA not adequate Want Raw ⊃ Rn Or Raw ⊃ Ω, where Ω ⊂ Rn is an ROA estimate Results for GPAW compensated planar LTI systems available [Teo and How 2010a,c]. Develop similar results for MIMO nonlinear systems J. Teo, J. P. How (DSO, ACL–MIT) ROA Comparison for GPAW Compensated Systems Dec. 14, 2011 4 / 12 GPAW Overview Overview of GPAW Compensation References [Teo and How 2009, 2010a,b,c,d, Teo 2011] Gradient Projection Anti-windup (GPAW) scheme constructed for saturated nonlinear MIMO plants driven by MIMO nonlinear controllers J. Teo, J. P. How (DSO, ACL–MIT) ROA Comparison for GPAW Compensated Systems Dec. 14, 2011 5 / 12 GPAW Overview Overview of GPAW Compensation References [Teo and How 2009, 2010a,b,c,d, Teo 2011] Gradient Projection Anti-windup (GPAW) scheme constructed for saturated nonlinear MIMO plants driven by MIMO nonlinear controllers For “strictly proper” nonlinear controllers ẋc = fc (xc , y, r) uc = gc (xc ) GPAW,Γ=ΓT >0 −−−−−−−−−−→ ẋg = RI ∗ (xg , y, r)fc (xg , y, r) ug = gc (xg ) Projection operator RI ∗ projects controller state onto unsaturated region K := {x | sat(gc (x)) = gc (x)} J. Teo, J. P. How (DSO, ACL–MIT) ROA Comparison for GPAW Compensated Systems Dec. 14, 2011 5 / 12 GPAW Overview Overview of GPAW Compensation References [Teo and How 2009, 2010a,b,c,d, Teo 2011] Gradient Projection Anti-windup (GPAW) scheme constructed for saturated nonlinear MIMO plants driven by MIMO nonlinear controllers For “strictly proper” nonlinear controllers ẋc = fc (xc , y, r) uc = gc (xc ) GPAW,Γ=ΓT >0 −−−−−−−−−−→ ẋg = RI ∗ (xg , y, r)fc (xg , y, r) ug = gc (xg ) Projection operator RI ∗ projects controller state onto unsaturated region K := {x | sat(gc (x)) = gc (x)} Achieves “controller state-output consistency”, a unique property: sat(ug ) ≡ ug or sat(gc (xg )) ≡ gc (xg ) Extension of “conditional integration” method [Fertik and Ross 1967] Can be realized in 3 equivalent ways (closed-form expressions also) J. Teo, J. P. How (DSO, ACL–MIT) ROA Comparison for GPAW Compensated Systems Dec. 14, 2011 5 / 12 GPAW Overview GPAW Scheme Visualization Nominal controller: ẋc = fc (xc , y, r) fc2 fc1 uc = gc (xc ) H2 ∇h2 xg3 fc0 fc3 H 3( K f g3 Boundaries: H1 , H2 , G3 xg1 f g1 fg2 ) H1 xg2 (x g3 ug = gc (xg ) ∇h 1 ∇h 3 GPAW controller: ẋg = RI ∗ fc (xg , y, r) xg0 ) J. Teo, J. P. How (DSO, ACL–MIT) x g3 Unsaturated region: Nominal update: Projections: G3 Gradients: ∇hi (xg ) = ±∇gci (xg ) K := {x̄ | sat(gc (x̄)) = gc (x̄)} fci := fc (xgi , y(ti ), r(ti )) for xgi := xg (ti ) fgi := RI ∗ fci ROA Comparison for GPAW Compensated Systems Dec. 14, 2011 6 / 12 General ROA Comparison General ROA Comparison Result System: ẋ = f (x), equilibrium: xeq , solution from x0 : φ(t, x0 ) ROA: RA (xeq ) := {x | limt→∞ φ(t, x) = xeq } ROA estimate: Ω ⊂ RA (xeq ) ROA estimate ΩV associated with Lyapunov function, V : V (xeq ) = 0, V (x) > 0, ∀x ∈ ΩV \ {xeq } ∂V (x) f (x) ≤ −α(kx − xeq k), ∀x ∈ ΩV V̇ (x) = ∂x M is positively invariant if forward solution starting in M stays in M J. Teo, J. P. How (DSO, ACL–MIT) ROA Comparison for GPAW Compensated Systems Dec. 14, 2011 7 / 12 General ROA Comparison General ROA Comparison Result System: ẋ = f (x), equilibrium: xeq , solution from x0 : φ(t, x0 ) ROA: RA (xeq ) := {x | limt→∞ φ(t, x) = xeq } ROA estimate: Ω ⊂ RA (xeq ) ROA estimate ΩV associated with Lyapunov function, V : V (xeq ) = 0, V (x) > 0, ∀x ∈ ΩV \ {xeq } ∂V (x) f (x) ≤ −α(kx − xeq k), ∀x ∈ ΩV V̇ (x) = ∂x M is positively invariant if forward solution starting in M stays in M Lemma (ROA Comparison for General Systems) Assume xeq is asymptotically stable equilibrium for two autonomous systems ẋ = f1 (x) and ẋ = f2 (x). Let RA1 (xeq ) and RA2 (xeq ) be associated ROAs. Let ΩV ⊂ RA1 (xeq ) be ROA estimate associated with Lyapunov function V . If Ω2 ⊂ ΩV is positively invariant for ẋ = f2 (x) and ∂V (x) ∂V (x) ∂x f2 (x) ≤ ∂x f1 (x) for all x ∈ Ω2 , then Ω2 ⊂ RA2 (xeq ) J. Teo, J. P. How (DSO, ACL–MIT) ROA Comparison for GPAW Compensated Systems Dec. 14, 2011 7 / 12 ROA Comparison for GPAW Compensated Systems An ROA Comparison Result Plant, Σp ẋ = f (x, sat(u)) Nominal Controller, Σc ẋc = fc (xc , y) y = g(x, sat(u)) u = gc (xc ) J. Teo, J. P. How (DSO, ACL–MIT) GPAW Controller, Σgpaw ẋg = RI ∗ fc (xg , y) u = gc (xg ) ROA Comparison for GPAW Compensated Systems Dec. 14, 2011 8 / 12 ROA Comparison for GPAW Compensated Systems An ROA Comparison Result Plant, Σp ẋ = f (x, sat(u)) Nominal Controller, Σc ẋc = fc (xc , y) GPAW Controller, Σgpaw ẋg = RI ∗ fc (xg , y) y = g(x, sat(u)) u = gc (xc ) u = gc (xg ) n Assume zeq ∈ R × (K \ ∂K) is asymptotically stable for Σn and Σg Nominal System: Σn : Σ p + Σ c ROA: Rn (zeq ) GPAW System: Σg : Σp + Σgpaw ROA: Rg (zeq ) J. Teo, J. P. How (DSO, ACL–MIT) ROA Comparison for GPAW Compensated Systems Dec. 14, 2011 8 / 12 ROA Comparison for GPAW Compensated Systems An ROA Comparison Result Plant, Σp ẋ = f (x, sat(u)) Nominal Controller, Σc ẋc = fc (xc , y) GPAW Controller, Σgpaw ẋg = RI ∗ fc (xg , y) y = g(x, sat(u)) u = gc (xc ) u = gc (xg ) n Assume zeq ∈ R × (K \ ∂K) is asymptotically stable for Σn and Σg Nominal System: Σn : Σ p + Σ c ROA: Rn (zeq ) GPAW System: Σg : Σp + Σgpaw ROA: Rg (zeq ) ROA estimate: ΩV = {z̄ | V (z̄) ≤ c} ⊂ Rn (zeq ) for some Lyapunov function V (z) = V (x, xc ) of Σn J. Teo, J. P. How (DSO, ACL–MIT) ROA Comparison for GPAW Compensated Systems Dec. 14, 2011 8 / 12 ROA Comparison for GPAW Compensated Systems An ROA Comparison Result Plant, Σp ẋ = f (x, sat(u)) Nominal Controller, Σc ẋc = fc (xc , y) GPAW Controller, Σgpaw ẋg = RI ∗ fc (xg , y) y = g(x, sat(u)) u = gc (xc ) u = gc (xg ) n Assume zeq ∈ R × (K \ ∂K) is asymptotically stable for Σn and Σg Nominal System: Σn : Σ p + Σ c ROA: Rn (zeq ) GPAW System: Σg : Σp + Σgpaw ROA: Rg (zeq ) ROA estimate: ΩV = {z̄ | V (z̄) ≤ c} ⊂ Rn (zeq ) for some Lyapunov function V (z) = V (x, xc ) of Σn Theorem (ROA Bounds for GPAW Compensated System) If there exists a Γ = ΓT > 0 such that ∂V (x̄, x̄c ) ∂V (x̄, x̄c ) RI ∗ fc ≤ fc , ∂xc ∂xc ∀(x̄, x̄c ) ∈ ΩV ∩ (Rn × K) then Σg with Γ has ROA satisfying (ΩV ∩ (Rn × K)) ⊂ Rg (zeq ) J. Teo, J. P. How (DSO, ACL–MIT) ROA Comparison for GPAW Compensated Systems Dec. 14, 2011 8 / 12 ROA Comparison for GPAW Compensated Systems Observations on ROA Comparison ROA comparison is relative to uncompensated system If ΩV = Rn (zeq ), conclusion is Rn (zeq ) ∩ (Rn × K) ⊂ Rg (zeq ) J. Teo, J. P. How (DSO, ACL–MIT) ROA Comparison for GPAW Compensated Systems Dec. 14, 2011 9 / 12 ROA Comparison for GPAW Compensated Systems Observations on ROA Comparison ROA comparison is relative to uncompensated system If ΩV = Rn (zeq ), conclusion is Rn (zeq ) ∩ (Rn × K) ⊂ Rg (zeq ) States loosely that ROA of Σg is “not less than” ROA estimate ΩV (or ROA Rn (zeq )) Specialized with additional assumptions (e.g. LTI) – yields LMI conditions [Teo 2011] J. Teo, J. P. How (DSO, ACL–MIT) ROA Comparison for GPAW Compensated Systems Dec. 14, 2011 9 / 12 ROA Comparison for GPAW Compensated Systems Observations on ROA Comparison ROA comparison is relative to uncompensated system If ΩV = Rn (zeq ), conclusion is Rn (zeq ) ∩ (Rn × K) ⊂ Rg (zeq ) States loosely that ROA of Σg is “not less than” ROA estimate ΩV (or ROA Rn (zeq )) Specialized with additional assumptions (e.g. LTI) – yields LMI conditions [Teo 2011] Main condition: ∂V (x̄,x̄c ) ∗ ∂xc RI fc ≤ ∂V (x̄,x̄c ) ∂xc fc independent of sat(·) Can be used in two ways: comparison against ROA estimate of unconstrained system or nominal system Result is applicable to MIMO nonlinear systems, but likely conservative J. Teo, J. P. How (DSO, ACL–MIT) ROA Comparison for GPAW Compensated Systems Dec. 14, 2011 9 / 12 ROA Comparison for GPAW Compensated Systems Application of ROA Comparison Result Example nonlinear planar system [Khalil 2002] ( ẋ = − sat(u) ( ẋ = − sat(u) Σn : Σ : 0, if A1 gs u̇ = x + (x2 − 1)u u̇ = 2 x + (x − 1)u, otherwise J. Teo, J. P. How (DSO, ACL–MIT) ROA Comparison for GPAW Compensated Systems Dec. 14, 2011 10 / 12 ROA Comparison for GPAW Compensated Systems Application of ROA Comparison Result Example nonlinear planar system [Khalil 2002] ( ẋ = −u ( ẋ = −u Σu : Σ : 0, if A1 g u̇ = x + (x2 − 1)u u̇ = 2 x + (x − 1)u, otherwise Compare with ROA estimate ΩV of unconstrained system: 4 3 φu (t, z0 ) 3 2 x 2 φg u(t, z0 ) 1 0 1 −1 u −2 0 0 1 2 3 4 0 1 2 3 4 5 6 7 8 9 10 5 6 7 8 9 10 2 ΩV 1 u −1 −2 Rg u(zeq) φu (t, z0 ) Ru (zeq ) φg u(t, z0 ) −3 −3 −2 −1 0 x J. Teo, J. P. How (DSO, ACL–MIT) 1 2 0 −1 −2 3 time (s) ROA Comparison for GPAW Compensated Systems Dec. 14, 2011 10 / 12 ROA Comparison for GPAW Compensated Systems Application of ROA Comparison Result Example nonlinear planar system [Khalil 2002] ( ẋ = −u ( ẋ = −u Σu : Σ : 0, if A1 g u̇ = x + (x2 − 1)u u̇ = 2 x + (x − 1)u, otherwise Compare with ROA estimate ΩV of unconstrained system: 4 3 φu (t, z0 ) 3 2 x 2 φg u(t, z0 ) 1 0 1 −1 u −2 0 0 1 2 3 4 0 1 2 3 4 5 6 7 8 9 10 5 6 7 8 9 10 2 ΩV 1 u −1 −2 Rg u(zeq) φu (t, z0 ) Ru (zeq ) φg u(t, z0 ) −3 −3 −2 −1 0 x 1 2 0 −1 −2 3 time (s) Toy example defeats methods for LTI systems, feedback linearizable systems, and nonlinear anti-windup [Morabito et al. 2004] J. Teo, J. P. How (DSO, ACL–MIT) ROA Comparison for GPAW Compensated Systems Dec. 14, 2011 10 / 12 Conclusions and Future Work References I H. A. Fertik and C. W. Ross. Direct digital control algorithm with anti-windup feature. ISA Trans., 6(4):317 – 328, 1967. H. K. Khalil. Nonlinear Systems. Prentice Hall, Upper Saddle River, NJ, 3 edition, 2002. F. Morabito, A. R. Teel, and L. Zaccarian. Nonlinear antiwindup applied to Euler-Lagrange systems. IEEE Trans. Robot. Autom., 20(3):526 – 537, June 2004. doi: 10.1109/TRA.2004.824933. S. Tarbouriech and M. Turner. Anti-windup design: an overview of some recent advances and open problems. IET Control Theory Appl., 3(1):1 – 19, Jan. 2009. doi: 10.1049/iet-cta:20070435. C. S. J. Teo. Gradient Projection Anti-windup Scheme. PhD thesis, Massachusetts Institute of Technology, Cambridge, MA, Feb. 2011. URL http://hdl.handle.net/1721.1/63043. J. Teo and J. P. How. Anti-windup compensation for nonlinear systems via gradient projection: Application to adaptive control. In Proc. 48th IEEE Conf. Decision and Control & 28th Chinese Control Conf., pages 6910 – 6916, Shanghai, China, Dec. 2009. doi: 10.1109/CDC.2009.5400075. J. Teo and J. P. How. Analysis of gradient projection anti-windup scheme. In Proc. American Control Conf., pages 5966 – 5972, Baltimore, MD, June/July 2010a. J. Teo and J. P. How. Geometric properties of gradient projection anti-windup compensated systems. In Proc. American Control Conf., pages 5973 – 5978, Baltimore, MD, June/July 2010b. J. Teo and J. P. How. Gradient projection anti-windup scheme on constrained planar LTI systems. Technical Report ACL10-01, MIT, Cambridge, MA, Mar. 2010c. URL http://hdl.handle.net/1721.1/52600. Aerosp. Controls Lab. J. Teo and J. P. How. Corrections to ‘geometric properties of gradient projection anti-windup compensated systems’. Technical Report ACL10-02, MIT, Cambridge, MA, June 2010d. URL http://hdl.handle.net/1721.1/56001. Aerosp. Controls Lab. J. Teo, J. P. How (DSO, ACL–MIT) ROA Comparison for GPAW Compensated Systems Dec. 14, 2011 11 / 12 Conclusions and Future Work Conclusions and Future Work Maximizing ROA not adequate - want Rn ⊂ Raw Developed ROA comparison result for general systems Specialized to GPAW compensated saturated MIMO nonlinear systems Toy example defeats 3 classes of state-of-the-art anti-windup methods Current result still likely conservative Plenty of work to find qualitatively similar but less conservative results J. Teo, J. P. How (DSO, ACL–MIT) ROA Comparison for GPAW Compensated Systems Dec. 14, 2011 12 / 12 Conclusions and Future Work Conclusions and Future Work Maximizing ROA not adequate - want Rn ⊂ Raw Developed ROA comparison result for general systems Specialized to GPAW compensated saturated MIMO nonlinear systems Toy example defeats 3 classes of state-of-the-art anti-windup methods Current result still likely conservative Plenty of work to find qualitatively similar but less conservative results For more information on GPAW compensation http://acl.mit.edu/projects/gpaw.htm http://dspace.mit.edu/handle/1721.1/63043 J. Teo, J. P. How (DSO, ACL–MIT) ROA Comparison for GPAW Compensated Systems Dec. 14, 2011 12 / 12 Conclusions and Future Work Conclusions and Future Work Maximizing ROA not adequate - want Rn ⊂ Raw Developed ROA comparison result for general systems Specialized to GPAW compensated saturated MIMO nonlinear systems Toy example defeats 3 classes of state-of-the-art anti-windup methods Current result still likely conservative Plenty of work to find qualitatively similar but less conservative results For more information on GPAW compensation http://acl.mit.edu/projects/gpaw.htm http://dspace.mit.edu/handle/1721.1/63043 Questions? J. Teo, J. P. How (DSO, ACL–MIT) ROA Comparison for GPAW Compensated Systems Dec. 14, 2011 12 / 12 Backup Slides ROAs of Saturated Planar LTI Systems, Rn ⊂ Rg Unstable plant, stable controller Symmetric constraints Asymmetric constraints 5 5 φn (t, z0 ) 4 φn (t, z0 ) 4 φg (t, z0 ) φg (t, z0 ) 3 3 2 2 1 Rn = Rg 0 u u 1 Rg 0 −1 −1 −2 −2 −3 −3 −4 −4 −5 −2.5 −5 −2.5 −2 −1.5 −1 −0.5 0 0.5 x 1 1.5 2 2.5 Rn −2 −1.5 −1 −0.5 0 x 0.5 1 1.5 2 2.5 Stable plant, unstable controller 3 Rn 2 Rg u 1 0 −1 φn (t, z0 ) −2 φg (t, z0 ) −3 −5 J. Teo, J. P. How (DSO, ACL–MIT) −4 −3 −2 −1 0 x 1 2 3 4 5 ROA Comparison for GPAW Compensated Systems Dec. 14, 2011 13 / 12 Backup Slides ROA Comparison Indicates True Advantage Existing anti-windup results are in “absolute” sense may not indicate any advantages of anti-windup scheme ROA comparison result is in “relative” sense directly shows advantage of GPAW scheme first in new anti-windup paradigm J. Teo, J. P. How (DSO, ACL–MIT) ROA Comparison for GPAW Compensated Systems Dec. 14, 2011 14 / 12 Backup Slides ROA Comparison Indicates True Advantage Existing anti-windup results are in “absolute” sense may not indicate any advantages of anti-windup scheme ROA comparison result is in “relative” sense directly shows advantage of GPAW scheme first in new anti-windup paradigm States loosely that ROA of Σg is not less than ROA estimate ΩV Applies for asymmetric saturation constraints Specialized with additional assumptions (e.g. LTI) J. Teo, J. P. How (DSO, ACL–MIT) ROA Comparison for GPAW Compensated Systems Dec. 14, 2011 14 / 12 Backup Slides ROA Comparison Indicates True Advantage Existing anti-windup results are in “absolute” sense may not indicate any advantages of anti-windup scheme ROA comparison result is in “relative” sense directly shows advantage of GPAW scheme first in new anti-windup paradigm States loosely that ROA of Σg is not less than ROA estimate ΩV Applies for asymmetric saturation constraints Specialized with additional assumptions (e.g. LTI) Main condition: ∂V (x̄,x̄c ) ∗ ∂xc RI fc ≤ ∂V (x̄,x̄c ) ∂xc fc independent of sat(·) Can be used in two ways: comparison against ROA estimate of unconstrained system or nominal system J. Teo, J. P. How (DSO, ACL–MIT) ROA Comparison for GPAW Compensated Systems Dec. 14, 2011 14 / 12