33 9 2016 9 DOI: 10.7641/CTA.2016.50970 Control Theory & Applications Vol. 33 No. 9 Sep. 2016,, (, 214122) :,.,,.,,,.. : ; ; ; ; : TP273 : A Optimal algorithm and application for point to point iterative learning control via updating reference trajectory TAO Hong-feng, DONG Xiao-qi, YANG Hui-zhong (Key laboratory of Advanced Process Control for Light Industry (Ministry of Education), Jiangnan University, Wuxi Jiangsu 214122, China) Abstract: For the output tracking control problem of discrete linear system with non-repetitive disturbance, a point to point iterative learning control algorithm based on updating reference trajectory is proposed. Firstly, the iterative learning controller is optimized by constructing performance index with norm function, and the corresponding convergence conditions are given, then the system output can track with the desired points in updating reference trajectory. Furthermore, when the system output is affected by non-repetitive disturbance in some trials, a new multi-objective performance index function is constructed by Lagrange multiplier algorithm, and the robust iterative learning controller is optimized to improve the convergence speed and tracking accuracy. Finally, the simulation results of the motor driven single mechanical arm control system show the effectiveness and feasibility of the proposed algorithm. Key words: updating reference trajectory; iterative learning control; performance index; Lagrange multiplier; point to point 1 (Introduction) (iterative learning control), [1 3].,,,,.,,,,,. [4],,,. [5],,., [6],. [7] P : 2015 12 08; : 2016 05 26.. E-mail: taohongfeng@hotmail.com; Tel.: +86 13621514714. :. (61273070, 61203092). Supported by National Natural Science Foundation of China (61273070, 61203092).
1208 33 CB 0 0 0., CAB CB 0 0, G = CA 2 B CAB CB 0,......, CA T 1 B CA T 2 B CA T 3 B CB. [8] d k =[CA CA 2 CA 3 CA T ] T x k (0),,,. u k y k u k =[u k (0) u k (1) u k (T 1)] T, y k =[y k (1) y k (2) y k (T )] T. [9], (1) k +1, ω, k+1,,., y k+1 = Gu k+1 + d k+1 + ω k+1, (3), [10], sup ω k+1 ξ, t [0,T ]. :,,. [11],. [12 13],. b x0 = sup t [0,T ] d k+1 d k.,,,.,,.,. 2 (Problem formulation) : { xk (t +1)=Ax k (t)+bu k (t), (1) y k (t) =Cx k (t), : x k (t) R m, u k (t) R r, y k (t) R n k, x k (0) k, t [0,T] T ; A, B, C, CB 0. (1), : : y k = Gu k + d k, (2),,, y k (t) y d (t), t {0, 1,,T}.,,, : r k (t i )=y d (t i ), i =1,,M, 0 t 1 <t 2 < <t M T., r k+1 (t) =r 0 (t)+h(t)f(t), (4) : h(t) =(t t 1 )(t t 2 ) (t t M ), f(t), r 0 (t). (4),,, r k (t i )=r 0 (t i )=y d (t i ). r k+1 =[r k+1 (0) r k+1 (1) r k+1 (T )] T, H =diag{h(0) h(1) h(t )}, f k =[f(0) f(1) f(t )] T. (4) : r k+1 = r 0 + Hf k. (5), r k (t i )=r 0 (t i )=y d (t i ), (5) : r k+1 = r k + Hf k. (6) (6) f k = F (r k y k ), F
9 : 1209, (6) r k+1 = r k + HF(r k y k ). (7) (7) H F, λ k = HF, λ k,, (7), (8) r k+1 = r k + λ k (r k y k ). (8) r k+1 y k = (I + λ k )(r k y k ) I + λ k (r k y k ). I + λ k =1, r k+1 y k (r k y k ). f k.,,,. 3 (Point to point robust iterative learning control) : { uk+1 = u k + L(r k+1 y k ), (9) r k+1 = r k + λ k (r k y k ), : λ k, r k k 1, L. 1 (1), (9), : ρ(i GL) < 1, k, : 0 lim e k+1 1 k 1 ρ (ξ + b x 0 )., ξ =0, b x0 =0, lim e k+1 =0. k k +1, (1) e k+1 = r k+1 y k+1 = r k + λ k (r k y k ) Gu k d k GL(r k + λ k (r k y k ) y k )+d k d k+1 ω k+1 = (I GL)(I + λ k )(r k y k )+ d k d k+1 ω k+1. (10) (10) (I GL) = ρ, I + λ k =1, (10) e k+1 ρ e k + ξ + b x0, (11) k e k+1 = ρ k+1 e 0 + 1 ρk+1 1 ρ (ξ + b x 0 ). (12) ρ <1, 0 lim k e k+1 1 1 ρ (ξ + b x 0 )., ξ =0, b x0 =0, k.,,, (11) e k+1 ρ e k. 0 <ρ<1, e k+1 e k,. 3,, 4, 5,. 4 (Norm performance index optimal iterative learning without disturbance), u k+1,., : min J k+1 = e k+1 2 + u Q k+1 u k 2 + u R k+1 2, S (13) Q, R S. e k+1 = r k+1 Gu k+1 d k+1, e k+1 e k+1 = k (I + λ i )r 0 k (I + λ i )λ 0 Gu 0 i=0 i=1 k (I + λ i )λ 1 Gu 1 i=2 (I + λ k )λ k 1 Gu k 1 λ k Gu k Gu k+1 λ k d k d k+1 k (I + λ i )λ 1 d 1 i=2 (I + λ k )λ k 1 d k 1. (14) (14) (13), u k+1 G T Q(r k+1 Gu k+1 d k+1 )+ R(u k+1 u k )+Su k+1 =0. (15) (G T QG + R + S)u k+1 = Ru k + G T Qr k+1 G T Qd k+1. (16), I + λ k =1,
1210 33 λ k 0, r k (t i )=r 0 (t i ), i =1, 2,,M. (16) : (G T QG + R + S)u k+1 = Ru k + G T Qr 0 G T Qd k+1. (17) (17) R S ri si, (r + s)i >0, G T QG > 0, (17) u k+1 =(G T QG + R + S) 1 Ru k + (G T QG + R + S) 1 G T Qr 0 (G T QG + R + S) 1 G T Qd k+1. (18), (17) (G T QG + R + S)u k+1 = (R + G T QG)u k + G T Q(r 0 y k ) G T Q(d k+1 d k ), (19) u k+1 =(G T QG + R + S) 1 (R + G T QG)u k (G T QG + R + S) 1 G T Q(d k+1 d k )+ (G T QG + R + S) 1 G T Qe k. (20) Z u =(G T QG + R + S) 1 (R + G T QG), Z e = (G T QG + R + S) 1 G T Q, (20) u k+1 = Z u u k + Z e e k Z e (d k+1 d k ). (21) 2 (21), (1) : 1) : ρ((g T QG + R + S) 1 R) < 1; 2) S ; 3) x k (0), k,. (1) u d, (21) u d u k+1 = u d Z u u k Z e e k +Z e (d k+1 d k ). (22) (22) G, e k+1 Z u Z e G e k + U, U = G Z u G u d + Z e d k+1 d k + Z u d k d k+1., ρ(z u Z e G)= ρ((g T QG + R + S) 1 R) < 1, (23)., S, x k (0) : Z u, G Z u G =0, d k+1 d k =0, Z u d k d k+1 =0, e k+1 ρ e k e k, (24) e k+1. 2 Amann [14 15], Owens [16 17],, (9), [15],.,, [16]. 3, 4,, (1). 5 (Norm index optimal iterative learning with non-repetitive disturbance) 1, k +1 e k+1 = r 0 Gu k+1 d k+1 ω k+1. (25) min J k+1 = r 0 Gu k+1 d k+1 ω k+1 2 + Q u k+1 u k 2 + u R k+1 2. S (26), min max J(u k+1,ω k+1 )= u k+1 ω k+1 r 0 Gu k+1 d k+1 ω k+1 2 + Q u k+1 u k 2 + u R k+1 2. S (27) ω k+1,, (27) ω k+1 r 0 Gu k+1 d k+1 ω k+1 2 Q. : max r 0 Gu k+1 d k+1 ω k+1 2. ω Q d(k+1) ω k+1 ξ, γ, max J 1 (ω k+1, γ), J 1 (ω k+1,γ)= r 0 Gu k+1 d k+1 ω k+1 2 Q γ( ω k+1 2 ξ 2 ). (28), γ 0, -, (28) ω k+1 γ
9 : 1211 Q(r 0 Gu k+1 d k+1 )=(Q γi)ω k+1, (29) ξ 2 = ω k+1 2. (30) Q γi 0, (26). (29) ω k+1 =(Q γi) + Q(r 0 Gu k+1 d k+1 ), (31) (Q γi) + Q γi. (31) (30) ξ 2 =(r 0 Gu k+1 d k+1 ) T Q(Q γi) + (Q γi) + Q(r 0 Gu k+1 d k+1 ). (32) (31) (32), ω k+1, γ u k+1,,, Θ(γ): Θ(γ) =(r 0 Gu k+1 d k+1 ) T Q(I+ (Q γi) + Q) (r 0 Gu k+1 d k+1 ) γξ 2. (33) (33) γ (32). (28) (33) : min Θ(γ) = 0 γ Q min (r 0 Gu k+1 d k+1 ) T Q(I+ 0 γ Q (Q γi) + Q)(r 0 Gu k+1 d k+1 )+γξ 2. (34),, max J 1 (ω k+1,γ)= min Θ(γ). (35) 0 γ Q (27) min min J(u k+1,ω k+1 )= 0 γ Q u k+1 Θ(γ)+ u k+1 2 + u S k+1 u k 2. R (36), (36) u k+1 (G T ΣG + R + S)u k+1 = Ru k + G T Σr 0 G T Σd k+1. (37) = Q[I +(Q γi) + Q], (37) u k+1 =(G T ΣG + R + S) 1 Ru k + (G T ΣG + R + S) 1 G T Σr 0 (G T ΣG + R + S) 1 G T Σd k+1. (38) ρ((g T ΣG + R + S) 1 R) < 1. (39)., (33) γ ξ 2 =(r 0 Gu k+1 d k+1 ) T Q(Q γi) + (Q γi) + Q(r 0 Gu k+1 d k+1 ). (40), (38) (40) (36), k, u k+1 =(G T ΣG + S) 1 G T Σr 0 (G T ΣG + S) 1 G T Σd k+1. (41) e k+1 = r 0 G(G T ΣG + S) 1 G T Σr 0 + G(G T ΣG + S) 1 G T Σd k+1 d k + ω k+1. (42) 4,, S,,. 6 (Simulation study), [18], ẋ 1 = x 2, ẋ 2 = N Ψ t x 1 B c Ψ t x 2 + K t Ψ t x 3, ẋ 3 = K b Γ x 2 R r Γ x 3 + 1 Γ u, y = x 1, (43) : N = m 2 gl + m 1 gl, Ψ t = Ξ +(1/3)m 2 l 2 + (1/10)m 1 l 2 D c, x 1 = θ, x 2 = θ, x 3 = i, g, θ, i, K t, K b, B c, D c, l, m 1, m 2, R r, u, Ξ, Γ. K t =1N m, K b =0.085 V s/rad, R r =0.075 Ω, B c =0.015 kg m 2 /s, D c =0.05, l =0.6m,m 1 =0.05 kg,m 2 =0.01 kg, Ξ =0.05 kg m 2,Γ=0.0008 Ω, g=9.8m/s 2. t =2s, t s =0.1s, 1.021 0.043 0.009 0.666 A= 0.298 0.076 0.023,B= 10.961, 0.329 0.127 0.038 1.421 C =[100], D=0.
1212 33 T {0, 1, 2,, 20},, M =7, t i {0, 2, 6, 10, 14, 18, 20}. x k (0) = k 1 [00.1 0.01] T. λ k, 0. Q, R, S Q = I, R =0.005I, S =0., P : u k+1 = u k + k p e k, (9), (21), 1 3. P.,,, 2,., 3,,, P. 3 Fig. 3 The curves of system performance index without disturbance 1 Fig. 1 The curves of system actual output and reference point tracking (43) 10 0.01k(sin t +cost), 5, γ 0.5, 4 5. 2 Fig. 2 The curves of system output error without disturbance 1 (43) (9), π 2 π, 2. 2,,, 4 Fig. 4 The curves of system output error with disturbance 4, 10,,,,, P.
9 : 1213 5 Fig. 5 The curves of system performance index with disturbance 5,.,,,,. 7 (Conclusions),,.,.,.. (References): [1] ZHANG Hongwei, YU Fashan, BU Xuhui, et al. Robust iterative learning control for permanent magnet linear motor [J]. Electric Machines and Control, 2012, 6(16): 81 86. (,,,. [J]., 2012, 6(16): 81 86.) [2] BU X H, YU F S, HOU Z S, et al. Iterative learning control for a class of linear discrete-time switched systems [J]. Automatica, 2013, 39(9): 1564 1569. [3] WANG H B, WANG Y. Iterative learning control for nonlinear systems with uncertain state delay and arbitrary initial error [J]. Control Theory & Application, 2011, 9(4): 541 547. [4] XU J X, HUANG D Q. Initial state iterative learning for final state control in motion systems [J]. Automatica, 2008, 44(12): 3162 3169. [5] DINH T V, FREEMAN C T, LEWIN P L, et al. Convergence and robustness of a point-to-point iterative learning control algorithm [C] //The 51st IEEE Conference on Decision and Control. Mauli, Hawaii, USA: IEEE, 2012: 4678 4683. [6] SON T D, AHN H S, KEVIN L M. Iterative learning control in optimal tracking problems with specified data points [J]. Automatica, 2013, 49(1): 1465 1471. [7] AN Tongjian, LIU Xiangguan. Point to point robust iterative learning control via reference trajectory regulating [J]. Journal of Zhejiang U- niversity, 2015, 49(1): 87 93. (,. [J]., 2015, 49(1): 87 92.) [8] SON T D, AHN H S. Terminal iterative learning control for optimal multiple point tracking [C] //2011 American Control Conference. San Francisco, USA: IEEE, 2011: 3651 3656. [9] CHI R H, HOU Z S, WANG D W, et al. An optimal terminal iterative learning control approach for nonlinear discrete-time systems [J]. Control Theory & Applications, 2012, 29(8): 1025 1030. [10] FREEMAN C T, CAI Z L, ROGERS E, et al. Iterative learning control for multiple point-to-point tracking application [J]. IEEE Transactions Control Systems Technology, 2011, 19(3): 590 600. [11] FREEMAN C T, TAN Y. Point-to-point iterative learning control with mixed constraints [C] //2011 American Control Conference. San Francisco, USA: AACC, 2011: 3657 3662. [12] SHEN D, WANG Y Q. Iterative learning control for stochastic pointto-point tracking system [C] //The 12th International Conference on Control. Guangzhou: IEEE, 2012: 480 485. [13] SON T D, NGUYEN D H, AHN H S. Iterative learning control for optimal multiple point tracking [C] //The 50th IEEE Conference on Decision and Control and European Control Conference. Oriando, USA: IEEE, 2011: 6025 6030. [14] AMANN N, OWENS D H, ROGERS E. Iterative learning control using optimal feedback and feedforward actions [J]. International Journal of Control, 1996, 62(2): 277 293. [15] AMANN N, OWENS D H, ROGERS E. Robustness of norm optimal iterative learning control [C] //1996 International Conference on Control. Exeter, UK: IEEE, 1996: 1119 1124. [16] OWENS D H, CHU B, SONG J M. Parameter optimal iterative learning control using polynominal representations of the inverse plant [J]. International Journal of Control, 2012, 85(5): 533 544. [17] OWENS D H. Multivariable norm optimal and parameter optimal iterative learning control: a unified formulation [J]. International Journal of Control, 2012, 85(8): 1010 1025. [18] LIU Jinkun. Design of Robot Control System and MATLAB Simulation [M]. Beijing: Tsinghua University Press, 2008. (. MATLAB [M]. :, 2008.) : (1979 ),,,,,, E-mail: taohongfeng@hotmail.com; (1989 ),,,, E-mail: d qsimon@163.com; (1955 ),,,,, E-mail: yhzjn@163.com.