This paper presents adaptive fuzzy finite time sliding mode control of microelectromechanical system gyroscope with uncertainty and external disturbance. Firstly, fuzzy system is employed to approximate the uncertainty nonlinear dynamics. Secondly, nonlinear sliding mode hypersurface and double exponential reaching law are selected to design the finite time convergent sliding mode controller. Thirdly, based on Lyapunov methods, adaptive laws are presented to adjust the fuzzy weights and the system can be guaranteed to be stable. Finally, the effectiveness of the proposed method is verified with simulation.
National Natural Science Foundation of China613040986020400560974109Aeronautical Science Foundation of China2015ZA53003Natural Science Basic Research Plan in Shaanxi Province2014JQ83262015JM62722016KJXX-86Fundamental Research Funds for the Central Universities3102015AX0013102015BJ(II)CG017Fundamental Research Funds of Shenzhen Science and Technology ProjectJCYJ20160229172341417International Science and Technology Cooperation Program of China2014DFA115801. Introduction
MEMS gyroscopes have become the most growing microsensors in recent years due to the characteristics of compact size, low cost, and high sensitivity. Most MEMS gyroscopes sales in the market are vibrating silicon micromechanical gyroscopes, whose basic principle is to generate and detect Coriolis Effect. As depicted in Figure 1, under assumption that the proof mass m of gyroscope rotates around z-axis at a speed of Ω→ and makes uniform motion along the x-axis at a speed of ν→, a Coriolis force of F→=-2mΩ→×ν→ is produced along y-axis.
Coriolis Effect.
In the last few years, numerous advanced control approaches with intelligent design have been studied to realize the trajectory tracking in [1–4] and to handle the system parametric uncertainties and disturbances, and the adaptive control can be found in [5–7]. For control of MEMS gyroscope, Park and Horowitz firstly applied adaptive state feedback control method [8]. Both drive shaft and sensitive axis were subjected to feedback control force in this control method, which administered two axial modal vibration track specified reference trajectories, weakening the boundary between drive mode and test mode as well.
Sliding mode control changes its structure to force the system in accordance with a predetermined trajectory. Batur et al. developed a sliding mode control for MEMS gyroscope system in [9]. Since then, adaptive sliding mode control approach with the advantages of variable structure methods and adaptive control strategies are presented to control MEMS gyroscopes in [10, 11].
Due to the necessity of ideal sliding mode, good dynamic quality, and high robustness, several methods are extended to improve the performance. Yu and Man investigated a nonlinear sliding mode hypersurface to ensure that systems from any point of the sliding mode surface were able to reach the balance point in a limited time in [12–16]. Bartoszewicz [17] examined the reaching laws introduced by Gao and Hung in [18] and proposed an enhanced version of those reaching laws, which was more appropriate for systems subject to constraints. Recently, Fallaha et al. studied a novel approach, which allowed chattering reduction on control input while keeping tracking performance in steady-state regime [19]. This approach consisted of designing a nonlinear reaching law by using an exponential function that dynamically adapted to the variations of the controlled system. Mei and Wang in [20] proposed a nonlinear sliding mode surface which converged to the equilibrium point with a higher speed than both linear sliding mode surface and terminal sliding mode surface. In addition, a new two-power reaching law was proposed to make the system move toward the sliding mode faster.
As a matter of fact, the methods mentioned above are highly dependent on the structure of the nonlinearity, while, currently, accurate model is unavailable. Thus, fuzzy model has been widely used to approximate nonlinear objects in [21, 22]. Robust adaptive sliding mode control with on-line identification for the upper bounds of external disturbance and estimator for the nonlinear dynamics of MEMS gyroscope uncertainty parameters was proposed in [23].
In this paper, an adaptive fuzzy sliding mode control strategy with nonlinear sliding mode hypersurface and double exponential reaching law is developed to track MEMS gyroscope. Furthermore, it converges faster compared with strategies using conventional sliding mode surface in [23] and terminal sliding mode surface in [12–16].
The rest of this paper is organized as follows. The dynamics of MEMS gyroscope with parametric uncertainties and disturbances are given in Section 2. Controller design and stability analysis are discussed in Section 3. Numerical simulations are conducted to verify the superiority of the proposed approach in Section 4, compared with conventional adaptive fuzzy sliding mode control. Conclusions are drawn in Section 5.
2. Dynamics of MEMS Gyroscope
The basic principle of z-axis vibratory MEMS gyroscope is shown in Figure 2, which can be described as a quality-stiffness-damping system. Owing to mechanical coupling caused by fabrication imperfections, the dynamics can be derived as (1)mx¨+dxxx˙+dxy-2mΩz∗y˙+kxx-mΩz∗2x+kxyy=ux∗,my¨+dxxy˙+dxy+2mΩz∗x˙+kyy-mΩz∗2y+kxyx=uy∗,where m is the mass of proof mass; Ωz∗ is the input angular velocity; x, y represent the system generalized coordinates; dxx, dyy represent damping terms; dxy represents asymmetric damping term; kxx, kyy represent spring terms; kxy represents asymmetric spring terms; and ux∗, uy∗ represent the control forces.
The basic principle diagram of z-axis vibratory MEMS gyroscope.
On issues related to the study of mechanism, the law described by model is required to be independent of dimensions. So, it is necessary to establish nondimensional vector dynamics. Because of the nondimensional time t∗=ωot, both sides of (1) should be divided by reference frequency ωo2, reference length qo, and reference mass m. Then the dynamics can be rewritten in vector forms:(2)q∗qo+D∗mωoq∗qo+2S∗ωoq∗qo-Ωz∗2ωo2q∗qo+K1∗mωo2q∗qo=u∗mωo2qo,where q∗=xy, u∗=ux∗uy∗, D∗=dxxdxydxydyy, S∗=0-Ωz∗Ωz∗0, K1∗=kxxkxykxykyy.
New parameters are defined as follows: (3)q=q∗qo,u=u∗mωo2qo,Ωz=Ωz∗ωo,D=D∗mωo,K1=K1∗mωo2,S=-S∗ωo.Thus, the final form of the nondimensional vector dynamics is(4)q¨=2S-Dq˙+Ωz2-K1q+u.
In presence of parametric uncertainties and external disturbance, based on (4), state equation of dynamics is established as(5)q¨=A+ΔAq˙+B+ΔBq+Cu+dt,where A∈R2×2, B∈R2×2, C∈R2×2 are system known matrices; ΔA, ΔB are parametric uncertainties; and d(t) is an external disturbance. Besides, A=2S-D, B=Ωz2-K1.
If the system total interference (consisting of parametric uncertainties and external disturbance) is represented by P(t), we know (6)q¨=Aq˙+Bq+Cu+Pq˙,q,t,where Pq˙,q,t=ΔAq˙+ΔBq+dt.
It is vital that (6) must meet the following assumptions.
Assumption 1.
The total interference Pq˙,q,t≤Pc, where Pc is an unknown positive vector.
Assumption 2.
The total interference Pq˙,q,t meets sliding mode matching conditions; namely, ΔA=CH1, ΔB=CH2, dt=CH3, where H1, H2, H3 are unknown matrices with appropriate dimensions.
Assumption 3.
A, B are observability matrices.
Based on the above assumptions, the controller can be designed to compensate the total interference.
3. Adaptive Fuzzy Finite Time Sliding Mode Control
The fuzzy model of P(t) could be composed of M IF-THEN rules, and the ith rule has the form
IF x˙i is A1i and y˙i is A2i and xi is A3i and yi is A4i
THEN P^q˙,q∣θPi is Bi, i=1,2,…,M.
Based on singleton fuzzifier, product inference, and center-average defuzzifier, its output can be expressed as (7)P^q˙,q∣θP=θPTμq˙,q,where μq˙,q=(ηA1i×ηA2i×ηA3i×ηA4i)/(∑j=1MηA1j×ηA2j×ηA3j×ηA4j), ηA1i, ηA2i, ηA3i, ηA4i are membership function values of the fuzzy variables x˙, y˙, x, y with respect to fuzzy sets A1, A2, A3, A4, respectively.
The fuzzy sets of input variables are defined as {N,Z,P}, where N is negative, Z is zero, and P is positive. Then membership functions of x˙ are selected as the following triangular functions: (8)ηNx˙i=1x≤-3-13x-3≤x≤0ηZx˙i=13x+1-3≤x≤0-13x+10≤x≤3ηPx˙i=1x≥313x0≤x≤3.
The corresponding membership functions of these fuzzy sets labels are depicted in Figure 3. In addition, the membership functions of y˙i, xi, and yi are the same with x˙i.
The membership functions for x˙i.
Based on the aforementioned fuzzy sets and membership functions, the fuzzy rules are described in Table 1. Therefore, 81 fuzzy rules are chosen.
Fuzzy control rules.
x˙i
N
Z
P
y˙i
N
Z
P
xi
N
Z
P
yi
N
Z
P
N: negative; Z: zero; P: positive.
The control target for MEMS gyroscope is to maintain the proof mass oscillation at given frequency and amplitude, such as xd=Axsinωxt, yd=Aysinωyt in the x and y directions, respectively. So, reference model can be designed as (9)q¨d=Adqd,where qd=xdyd, Ad=-ωx200-ωy2.
And the tracking error is defined as (10)e=q-qd.Nonlinear sliding mode hypersurface is chosen as (11)s=e˙+αen1/m1+βem2/n2,where α>0, β>0; m1>n1>0, m2>n2>0; what is more, m1, n1, m2, n2 are odd.
Then the reaching law is designed as the following double exponential function: (12)s˙=-k1sasgns-k2sbsgns,where k1>0, k2>0, 0<a<1, b>1.
It should be noted that the converge speed depends on parameters such as α, β, m1, n1, m2, n2 and k1, k2, a, b.
According to (6), equivalent control law is obtained as(13)ueq=C-1q¨-Aq˙-Bq-P^q˙,q∣θP=C-1q¨d+e˙-Aq˙-Bq-P^q˙,q∣θP.And the derivative of sliding surface (11) is (14)s˙=e¨+αn1m1en1/m1-1+βm2n2em2/n2-1.Then substituting (12) into (14), (15)e¨=-k1sasgns-k2sbsgns-αn1m1en1/m1-1-βm2n2em2/n2-1.And substituting (15) into (13),(16)ueq=C-1q¨d-k1sasgns-k2sbsgns-αn1m1en1/m1-1-βm2n2em2/n2-1-Aq˙-Bq-P^q˙,q∣θP.Besides, a robust item is designed to guarantee that the system is asymptotically stable: (17)us=-C-1Ks.Thus, the adaptive fuzzy finite time sliding mode controller is obtained as(18)u=ueq+us.According to (10), we have (19)e¨=q¨-q¨d=Aq˙+Bq+Cu+Pq˙,q,t-q¨d.Substituting (18) into (19), (20)e¨=Aq˙+Bq+q¨d-k1sasgns-k2sbsgns-αn1m1en1/m1-1-βm2n2em2/n2-1-Aq˙-Bq-P^q˙,q∣θP-Ks+Pq˙,q,t-q¨d=Pq˙,q,t-P^q˙,q∣θP-Ks-k1sasgns-k2sbsgns-αn1m1en1/m1-1-βm2n2em2/n2-1.Substituting (20) into (14), (21)s˙=Pq˙,q,t-P^q˙,q∣θP-Ks-k1sasgns-k2sbsgns.The optimal parameters are set as(22)θP∗=argminsupP^q˙,q∣θP-Pq˙,q,tθP∈ΩP,q˙,q∈R2×2,where ΩP is a set of θP.
And the minimum approximation errors are defined as (23)w=Pq˙,q,t-P^q˙,q∣θP.Substituting (23) into (21), we derive(24)s˙=P^q˙,q∣θP∗-P^q˙,q∣θP+w-Ks-k1sasgns-k2sbsgns.Considering (7), (24) can be expressed as(25)s˙=φPTμq˙,q+w-Ks-k1sasgns-k2sbsgns,where φP=θP∗-θP.
So adaptive law can be selected as(26)φ˙P=-rsTμq˙,q.Namely, (27)θ˙Px=rs1μxq˙,q,θ˙Py=rs2μyq˙,q,where φ˙P=-θ˙P.
Lyapunov function is defined as(28)V=12sTs+1rφPTφP.Differentiate V with respect to time yields, and substitute (26) as(29)V˙=sTw-sTKs-k1sTsasgns-k2sTsbsgns.
Owing to the fuzzy approximation theory, adaptive fuzzy system can approximate nonlinear system closely. Therefore, V˙≤0; namely, the system is asymptotically stable.
4. Simulation Study
In this section, numerical simulations are investigated to track the position and speed trajectories of MEMS gyroscope, compensate parametric uncertainties and external disturbances, and verify the superiority of the proposed approach compared with conventional adaptive fuzzy sliding mode control strategy using linear sliding mode surface. Those two methods are defined as follows.
Method 1.
Define the adaptive fuzzy sliding mode control proposed in this paper as Method 1, whose sliding mode surface is shown in (11), and the reaching law is expressed in (12).
Method 2.
Define the conventional adaptive fuzzy sliding mode control as Method 2, whose sliding mode surface is =e˙+βe, and the reaching law is s˙=0.
Parameters of the MEMS gyroscope are as follows:(30)m=0.57×10-8kg,dxx=0.429×10-6Ns/m,dyy=0.0429×10-6Ns/m,dxy=0.0429×10-6Ns/m,kxx=80.98N/m,kyy=71.62N/m,kxy=5N/m,Ωz=5.0rad/s.
Since the position of proof mass ranges within the scope of submillimeter and the natural frequency is generally in the range of kilohertz, assume that reference length is qo=10×10-6 m, reference frequency is ωo=1 kHz, and the reference trajectories are xd=sin6.71t, yd=1.2sin5.11t, respectively.
Then set other simulation parameters as(31)A=-0.0750.0025-0.0175-0.0075,B=-14207-877-877-12564,C=1001,K=1000001000,α=100010,β=100010,m1=3,n1=2,m2=3,n2=1,Pt=3.2×10-65×10-6+5×10-6sin5.11t+0.3,r=0.01,a=0.5,b=10,k1=1000,k2=1000.
And select the initial state values of the system as 0.8010T.
Then the position and speed trajectories of Method 1 are shown in Figures 4 and 5 and those of Method 2 are depicted in Figures 6 and 7.
Position tracking of Method 1.
Speed tracking of Method 1.
Position tracking of Method 2.
Speed tracking of Method 2.
The position tracking error and speed tracking error of Methods 1 and 2 are shown in Figures 8–11, respectively.
Position tracking error of gyroscope x.
Position tracking error of gyroscope y.
Speed tracking error of gyroscope x.
Speed tracking error of gyroscope y.
Through the tracking simulation of MEMS gyroscope, the proposed approach is with satisfying performance; in addition, in comparison to Method 2, the convergence time is shortened to 0.3′′ from 0.6′′.
5. Conclusion and Future Work
An adaptive fuzzy finite time sliding mode control strategy using nonlinear sliding mode hypersurface and double exponential reaching law is proposed to compensate parametric uncertainties and external disturbance of MEMS gyroscope in this paper. Based on Lyapunov methods, the stability of system can be guaranteed. Simulations verify that, compared with conventional adaptive fuzzy linear sliding mode control strategy, the convergence time of finite time convergent control strategy proposed in this paper is shortened to 0.3′′ from 0.6′′; namely, convergence has been significantly improved. For future work, the novel adaptive online constructing fuzzy algorithm [24] can be employed for more efficient learning while disturbance observer based design [25, 26] can be considered to improve system performance.
Competing Interests
The authors declare that they have no competing interests.
Acknowledgments
This work was supported by National Natural Science Foundation of China (61304098, 60204005, and 60974109), Aeronautical Science Foundation of China (2015ZA53003), Natural Science Basic Research Plan in Shaanxi Province (2014JQ8326, 2015JM6272, and 2016KJXX-86), Fundamental Research Funds for the Central Universities (3102015AX001, 3102015BJ(II)CG017), Fundamental Research Funds of Shenzhen Science and Technology Project (JCYJ20160229172341417), and International Science and Technology Cooperation Program of China under Grant no. 2014DFA11580.
ChenM.GeS. S.Adaptive neural output feedback control of uncertain nonlinear systems with unknown hysteresis using disturbance observer201562127706771610.1109/tie.2015.24550532-s2.0-84947769466XuB.YangC.PanY.Global neural dynamic surface tracking control of strict-feedback systems with application to hypersonic flight vehicle201526102563257510.1109/tnnls.2015.2456972MR34531962-s2.0-84944036101XuB.FanY.ZhangS.Minimal-learning-parameter technique based adaptive neural control of hypersonic flight dynamics without back-stepping201516420120910.1016/j.neucom.2015.02.0692-s2.0-84935535286YangQ.JagannathanS.SunY.Robust integral of neural network and error sign control of MIMO nonlinear systems201526123278328610.1109/tnnls.2015.2470175MR34532742-s2.0-84958113022XuB.ZhangQ.PanY.Neural network based dynamic surface control of hypersonic flight dynamics using small-gain theorem201617369069910.1016/j.neucom.2015.08.017LiuY.-J.TongS.Adaptive NN tracking control of uncertain nonlinear discrete-time systems with nonaffine dead-zone input201545349750510.1109/tcyb.2014.23294952-s2.0-84923343195HeW.ChenY.YinZ.Adaptive neural network control of an uncertain robot with full-state constraints201646362062910.1109/tcyb.2015.24112852-s2.0-84926319520ParkS.HorowitzR.New adaptive mode of operation for MEMS gyroscopes2004126480081010.1115/1.18492522-s2.0-17044391497BaturC.SreeramreddyT.KhasawnehQ.Sliding mode control of a simulated MEMS gyroscope20064519910810.1016/s0019-0578(07)60069-x2-s2.0-31644444586EbrahimiA.Regulated model-based and non-model-based sliding mode control of a MEMS vibratory gyroscope20142862343234910.1007/s12206-014-0525-42-s2.0-84902329548FeiJ.BaturC.A novel adaptive sliding mode control with application to MEMS gyroscope2009481737810.1016/j.isatra.2008.10.0082-s2.0-57549089047MichailZ. A. K.Terminal attractors in neural networks19892225927410.1016/0893-6080(89)90036-1YuX.ZhihongM.Fast terminal sliding-mode control design for nonlinear dynamical systems200249226126410.1109/81.983876MR18903862-s2.0-0036475055YuX.ManZ.Terminal sliding mode control of MIMO systems199744111065107010.1109/81.641769GhanbariA.Moghanni-Bavil-OlyaeiM. R.Adaptive fuzzy terminal sliding-mode control of MEMS z-axis gyroscope with extended Kalman filter observer20142183191YanW. F.HouS. X.FangY. M.FeiJ.Robust adaptive nonsingular terminal sliding mode control of MEMS gyroscope using fuzzy-neural-network compensator201610.1007/s13042-016-0501-7BartoszewiczA.A new reaching law for sliding mode control of continuous time systems with constraints201537451552110.1177/01423312145432982-s2.0-84924956656GaoW.HungJ. C.Variable structure control of nonlinear systems: a new approach1993401455510.1109/41.1848202-s2.0-0027542351FallahaC. J.SaadM.KanaanH. Y.Al-HaddadK.Sliding-mode robot control with exponential reaching law201158260061010.1109/TIE.2010.20459952-s2.0-78651368332MeiH.WangY.Fast convergent sliding mode variable structure control of robot2009385552557WangL.-X.MendelJ. M.Fuzzy basis functions, universal approximation, and orthogonal least-squares learning19923580781410.1109/72.1590702-s2.0-0026928374ZengX.SinghM. G.Singh approximation theory of fuzzy systems-MIMO case19953221923510.1109/91.388175FeiJ.WangS.Robust adaptive sliding mode control of MEMS gyroscope using T-S fuzzy model2014771-236137110.1007/s11071-014-1300-xMR32151552-s2.0-84902298295WangN.ErM. J.Direct adaptive fuzzy tracking control of marine vehicles with fully unknown parametric dynamics and uncertainties20162451845185210.1109/tcst.2015.25105872-s2.0-84954045108XuB.SunF.PanY.ChenB.Disturbance observer based composite learning fuzzy control of nonlinear systems with unknown dead zone201610.1109/tsmc.2016.2562502XuB.Disturbance observer-based dynamic surface control of transport aircraft with continuous heavy cargo airdrop201610.1109/tsmc.2016.2558098