Mathematical Modeling, Analysis, and Advanced Control of Complex Dynamical SystemsView this Special Issue
Composite Antidisturbance Control for a Class of Nonlinear Stochastic Systems via Disturbance Observer
The stabilization problem is investigated in this paper for a class of nonlinear systems with disturbances. The disturbances are supposed to be classified into two types. One type in the input channel is generated by an exogenous system, which can represent the constant or harmonic signals with unknown phase and magnitude. The other type is stochastic disturbance. Two kinds of nonlinear dynamics in the plants are considered, respectively, which correspond to the known and unknown functions. By integrating the disturbance observers with conventional control method, the first type of disturbances can be estimated and rejected. Simultaneously, the desired dynamic performances can be guaranteed. An example is given to show the effectiveness of the proposed scheme.
Though the stochastic stabilization theory emerged in the 1960s , the progress has been slow. This was mainly due to a fundamental theory obstacle in the Lyapunov analysis; the Itô differentiation introduces not only the gradient but also the Hessian term of the Lyapunov function. Along with the advances in differential geometric theory  and the discovery of a simple constructive formula for Lyapunov stabilization , the stochastic stabilization problem was reexamined and some constructive results have been achieved. The stabilization of nonlinear stochastic systems was considered in the work of Florchinger [4–7], who, among other things, extended the concept of control Lyapunov functions and Sontag’s stabilization formula to stochastic setting. Pan and Başer  solved the stabilization problem for a class of strict-feedback systems representative of stabilization results for deterministic systems. Deng and Krstić  developed a simpler control strategy for strict-feedback systems and then extended the results on inverse optimal stabilization for general systems to the stochastic case. The authors of  considered the dissipativity analysis and dissipativity-based sliding mode control for a class of continuous-time switched stochastic systems and  designed multistep predictive controller for a class of Markov jump convex polyhedron linear parameter time-varying systems with both constraints on input and output. The adaptive neural tracking control problem was the concern in  for a class of strict-feedback stochastic nonlinear systems with unknown dead zone. However, most of these results were focused on systems that only have one kind of disturbance-stochastic disturbance. In , Hinrichsen proposed the stochastic robust control method for systems with deterministic and stochastic disturbances, which enabled us to deal with a broader class of systems. However, only stability of the nominal system in the absence of deterministic disturbances was the concern in this approach, which means that the stability cannot be guaranteed in the presence of both deterministic and stochastic disturbances.
Disturbance-observer-based control (DOBC) approach, which is based on the idea of feed-forward compensation, appeared in the late 1980s and has attracted considerable attention in control theory literatures [14–17]. The controller design of this method can be accomplished in two steps. First, a disturbance observer is designed to estimate the deterministic disturbance and then compensate it. Second, feedback controller is designed to stabilize the nominal system without disturbance. DOBC approach has its roots in many mechanical applications in the last two decades, in particular for linear systems [18–20]. Recently, some attempts have been made to establish theoretic justification of these DOBC applications and extend DOBC from linear systems to nonlinear systems [21, 22]. Besides,  designed output feedback controller for a class of Markovian jump repeated scalar nonlinear systems and  investigated the problem of composite DOBC and control for Markovian jump systems with nonlinearity and multiple disturbances.
This paper considers the application of DOBC approach to a class of nonlinear stochastic systems. The nonlinear dynamics are described by known and unknown nonlinear functions, respectively. And apart from the stochastic noises, the deterministic disturbance is supposed to be generated by an exogenous system as investigated in [18, 19], which is not confined to be bounded in norm . By using the disturbance estimation, the DOBC strategy can be integrated with the conventional stabilization controllers to reject the deterministic disturbance and globally stabilize the closed-loop systems in probability. Finally, simulations on an A4D aircraft model show the effectiveness of the proposed approaches.
2. Problem Statement
The following MIMO stochastic system with nonlinearity is described as where , and , are the state, control input, and disturbance, respectively. is a nonlinear vector function satisfying bounded condition as described in Assumption 1. is an -dimensional standard Brownian motion defined on a complete probability space with being a sample space, being a -field, and being the probability measure. , , , and are given system matrices with corresponding dimensions.
Assumption 1. For any , , nonlinear function satisfy where is a given constant weighting matrix.
Assumption 2. The disturbance in the control input path is supposed to be generated by the following exogenous systems: where , are known constant weighting matrices.
Remark 3. In fact, many kinds of disturbances in engineering can be described by Assumption 2, for example, unknown constant and harmonics with unknown phase and magnitude.
The following assumption is the necessary condition for the DOBC problem.
Assumption 4. is controllable and is observable.
In this paper, we suppose system states are available, which means that only the estimation of disturbance needs to be focused on. In this situation, the objective of DOBC is to design an observer for system (1) to estimate the unknown disturbance , and then construct a composite controller with the disturbance estimation and a conventional controller so that the disturbance can be rejected and the stability in probability of the resulting composite system can be guaranteed.
3. DOBC for the Case with Known Nonlinearity
In this section, the nonlinearity function is supposed to be given and Assumptions 1, 2, and 4 hold. Due to the fact that the states of system are available, the disturbance observer is designed as where is the estimation of and is the auxiliary variable generated by
The objective of disturbance rejection can be achieved by designing the observer gain such that (6) satisfies the desired stability in probability.
In the following, the objective is to design gain and such that system (9) is asymptotical stabilization in probability. For the convenience of research, the following lemma is presented.
Lemma 5 (see [29, 30]). For a given stochastic system if there exists function , , constants , and a nonnegative , such that then, one has the following.(i)The equilibrium is globally stable in probability and the solution satisfied , when , , and is continuous.(ii)The equilibrium is globally asymptotically stable in probability, when , , and is positive definite.Here, the differential operator for differentiable function is defined as
Theorem 6. Consider system (1) with disturbance (3) under Assumptions 1, 2, and 4. For some , if there exist , and constant satisfying and , satisfying where , then the closed loop system (8) under DOBC law (7) with gain and observer (4) with gain are global asymptotical stabilization in probability.
It is noted that for all and , , . In addition, along with (6), (8), the Itó differential of satisfies Via Young inequality, we get Then we have
A Lyapunov function candidate for (9) is chosen as ; hence, it is easy to get where
Using Lemma 5, it can be verified that system (9) is global asymptotical stabilization in probability if and hold.
Based on Schur complement, is equivalent to , where and is equivalent to , where and . In addition, represents the corresponding elements in the symmetric matrix.
is premultiplied and postmultiplied simultaneously by ; then it is equivalent to , where and .
Thus, (14), (15) can be obtained.
On the other hand, (14), (15) hold, meaning that there exist , such that , ; that is, , . Hence, we have Therefore, the closed-loop system (9) is global asymptotical stabilization in probability when the control gain is selected as and observer gain is selected as . The proof is completed.
4. DOBC for the Case with Unknown Nonlinearity
In this section, the nonlinear function is supposed to be unknown, which means disturbance observer should be designed different from Section 3. In this case, the disturbance observer can be constructed as
Compared with (6), the estimation error satisfies
Theorem 7. Consider system (1) with disturbance (3) under Assumptions 1, 2, and 4. For some , , if there exist , and constant satisfying and , satisfying where , then the closed loop system (28) under DOBC law (7) with gain and observer (26) with gain is global asymptotical stabilization in probability.
5. Simulation Example
In , DOBC strategy was employed to a deterministic system of A4D aircraft with disturbance, and better system performance was obtained than some previous results . However, stochastic noise should be considered when higher precision of system performance was required. In this section, the stochastic system that represents the longitudinal dynamics of A4D aircraft is considered, which is described as follows: with the following coefficient:
The parameter matrices for disturbance that are described by (3) are given by
Case 1 (with known nonlinearity). In this case, the nonlinear dynamic is supposed to be denoted by in the simulation. In order to satisfy Assumption 1, is selected as . The initial value of state is taken to be and is selected as . Based on Theorem 6, it can be solved that
When the stochastic robust control strategy is applied to (33), which was first given in , it can be shown from Figure 1 that asymptotical stabilization in probability cannot be guaranteed in the presence of disturbance . Figures 2 and 3 show the system response and estimation error of system disturbance for the case with known nonlinearity, respectively. The simulation results show that asymptotical stabilization can be achieved using the method proposed in this paper and that the proposed disturbance observer is fine and effective.
Case 2 (with unknown nonlinearity). When nonlinear term is unknown, we assume in simulation, where is supposed to be stochastic input that obeys uniform distribution. Based on Theorem 7, it can be solved that
It is clear from Figure 4 that all states of system converge to zero and estimation errors of disturbance also converge to zero as shown in Figure 5. As has been shown above, we can see that asymptotical stabilization in probability is guaranteed and satisfied performance of the closed-loop control systems is achieved.
In this paper, the DOBC approach is investigated for a class of nonlinear systems with deterministic and stochastic disturbances. Feasible design procedures are proposed under different conditions to estimate and reject deterministic disturbance for the plants with known and unknown nonlinearity. Based on the estimation of disturbances, the composite control laws can guarantee the composite closed-loop systems to be global asymptotical stabilization in probability in the presence of disturbances. Simulation for an aircraft model shows the efficiency of the proposed algorithms.
This work was supported partly by the National Natural Science Foundation of China under Grants 61374131, 60925012, and 61121003 and the National Basic Research Program of China (973 Program) under Grant 2012CB720003.
H. J. Kushner, Stochastic Stability and Control, Academic Press, New York, NY, USA, 1967.View at: MathSciNet
A. Isidori, Nonlinear Control Systems, Springer, New York, NY, USA, 3rd edition, 1995.View at: MathSciNet
L. Wu, W. X. Zheng, and H. Gao, “Dissipativity based sliding mode control of switched stochastic systems,” IEEE Transactions on Automatic Control, vol. 58, no. 3, pp. 785–793, 2013.View at: Google Scholar
Y. Y. Yin, Y. Shi, and F. Liu, “Constrained model predictive control on conven polyhedron stochastic linear parameter varying systems,” International Journal of Innovative Computing, Information and Control, vol. 9, no. 10, pp. 4193–4204, 2013.View at: Google Scholar
H. Q. Wang, B. Chen, and C. Lin, “Adaptive neural tracking control for a class of stochastic nonlinear systems with unknown dead-zone,” International Journal of Innovative Computing, Information and Control, vol. 9, no. 8, pp. 3257–3269, 2013.View at: Google Scholar
R. Bickel and M. Tomizuka, “Passivity-based versus disturbance observer based robot control: equivalence and stability,” Journal of Dynamic Systems, Measurement and Control, vol. 121, no. 1, pp. 41–47, 1999.View at: Google Scholar
L. Xie, X. He, W. Zhang, and X. Xu, “Robust control for nonlinear uncertain stochastic systems with multiple time delays,” Control and Decision, vol. 16, pp. 779–782, 2001.View at: Google Scholar
L. Guo, X. Y. Wen, and X. Xin, “Hierarchical composite anti-disturbance control for robotic systems using robust disturbance observer,” in Proceedings of the Robot Intelligence, Advanced Information and Knowledge, pp. 229–243, 2010.View at: Google Scholar
P. E. Caines and K. Nassiri-Toussi, “On the adaptive stabilization and ergodic behaviour of stochastic systems with jump-Markov parameters via nonlinear filtering,” in Topics in Stochastic Systems: Modelling, Estimation and Adaptive Control, vol. 161 of Lecture Notes in Control and Information Science, pp. 185–215, Springer, Berlin, Germany, 1991.View at: Publisher Site | Google Scholar | Zentralblatt MATH | MathSciNet
K. Nassiri-Toussi and P. E. Caines, “On the adaptive stabilization and ergodic behaviour of stochastic jump parameter systems via non-linear filtering,” in Proceedings of the 30th IEEE/CDC Conference on Decision and Control, December 1991.View at: Google Scholar