On stability and application of extremum seeking control without steady-state oscillation☆
Introduction
Extremum seeking is a kind of adaptive control which can drive and maintain the input and output of the controlled object to their respective extrema. Extremum seeking control will work without any explicit knowledge about the input–output characteristics as long as the extrema exist, which is its greatest advantage. Therefore, extremum seeking is a model independent control scheme. Though extremum seeking scheme has been developed for several decades, the first rigorous stability analysis of the classic extremum seeking scheme was published by Wang and Krstić, 2000a, Wang and Krstić, 2000b. It appears that this paper renewed research interest in the theory of extremum seeking, and consequently the last decade has witnessed the significant development and numerous applications (Moase, Manzie, Nešić, & Mareels, 2010). Among them, Tan, Nešić, and Mareels (2008) studied how the different excitation signal in extremum seeking scheme affects the system performance. Lavretsky, Hovakimyan, and Calise (2003) pointed out that the amplitude of the excitation signal plays an important role in the system performance.
Extremum seeking scheme is essentially a method that achieves and maintains the function extremum by obtaining gradient information of the unknown function. Therefore, the extremum seeking scheme could easily converge to one of the local extrema if they exist. Committed to finding solutions to the problem, Tan, Nešić, Mareels, and Astolfi (2009) proposed a global extremum seeking scheme. A monotonically decreasing time function was designed to adjust the amplitude of the excitation signal, so that the searching arguments are expanded to get a certain ability to overcome the possible convergence to a local extremum. However, the performance of the scheme proposed by Tan et al. (2009) is not prominent because the excitation signal amplitude only varies with time. The same as the ordinary extremum seeking scheme, the improved scheme still has a large steady-state oscillation, which is undesirable, even not allowed for most practical systems.
In order to achieve the purpose of real-time optimality, the scheme is required to have a fast convergence rate. Krstić (1999) proposed a fast adaptive extremum seeking scheme to solve this problem. By introducing a dynamic compensator, the gain and phase margins of the feedback loop were improved to enhance the stability of the system and increase the response speed of the system. Compared with ordinary extremum seeking scheme it has better dynamic performance (Zuo, Hu, & Shi, 2006). However, the dynamic compensator design is dependent on the prior knowledge of the Wiener–Hammerstein model of the plant (Ariyur & Krstić, 2003), which brings inconvenience to its application.
As an application of extremum seeking control, the antilock braking systems with extremum seeking control have been investigated extensively (Dinçmen et al., 2014, Drakunov et al., 1995, Tunay, 2001, Yu and Özgüner, 2002, Zhang and Ordóñez, 2007). They all have reached the control purpose that the braking time is short and the tires are not locked during braking. However, most of them do bring additional oscillations in steady-state due to the perturbation signal and the sliding mode, respectively. Zhang and Ordóñez (2007) proposed an extremum seeking scheme based on numerical optimization through combining the numerical optimization algorithm with state regulation. The steady-state oscillation is successfully avoided when using the asymptotic state regulator numerical optimization based extremum seeking control. However, its real-time implementation will be affected when using a sophisticated optimization algorithm. Besides, the complex parameter adjustment of the scheme is not convenient for its application, too.
Motivated by the above study, we proposed a novel extremum seeking scheme in which the excitation signal amplitude can change adaptively with the extremum estimation error (Wang, Chen, & Zhao, 2014). The same as the scheme proposed by Tan et al. (2009), our scheme can avoid falling into local extrema effectively. The difference from the scheme proposed by Tan et al. (2009) is that our scheme can eliminate the steady-state oscillation due to the fact that the excitation signal amplitude will fast converge to zero with the decrease of the extremum estimation error. Furthermore it also has a strong adaptability to the extremum perturbation. In this paper, we improve the scheme further and give the rigorous stability proof of this improved scheme using Singular Perturbation Theory, Averaging Method, The Center Manifold Theorem and Lyapunov Method. We also give a simulation example to validate the characteristics of non-oscillating steady state of the improved scheme. Besides, the application of the improved scheme to antilock braking systems sufficiently illustrates the practicability of the scheme.
This paper is organized as follows. The problem formulation is given in Section 2. The main results are stated in Section 3. In Section 4 we discuss the improved scheme in detail. Section 5 consists of simulation example and application to ABS. Finally, a brief summary of the full text is given in Section 6. Some lemmas which are used in proof and auxiliary results are presented in the Appendix.
Section snippets
Problem formulation
Consider a general single input and single output (SISO) nonlinear model as follows: where and are continuously differentiable, is the state, is the input and is the measurable output. Suppose there exists a family control laws of the following form: where is a scalar parameter. The closed-loop system then has an equilibrium point parameterized by . For simplicity, is assumed to be scalar and (1), (2) is assumed to be
Main results
In this section we will investigate stability of a novel ESCWSSO. This novel scheme is an improvement of the perturbation-based extremum seeking scheme. The analysis of the proposed scheme lends itself to an understanding of this improvement. The proposed ESCWSSO is shown in Fig. 1.
The scheme shown in Fig. 1 introduces two new design parameters instead of the excitation signal amplitude in the classic extremum seeking scheme. is the cutoff frequency of the low-pass filter and is a
Discussions
In this section we will discuss main characteristics of the proposed scheme including non-oscillating steady state, the ability of passing local extremum and the selection of the design parameters.
From (23), (25) we can get the following formula in the vicinity of extreme point. (35) shows that the excitation signal amplitude locally exponentially converges to zero. We can always make by choosing the initial value . It can be seen
Examples
This section consists of two parts. In the first part, we give the simulation results of the ESCWSSO. The application of the scheme to ABS is given in the second part to illustrate the advantage of it through comparing with the ABS with classic perturbation-based extremum seeking scheme and sliding-mode-based extremum seeking scheme.
Conclusion and future work
This paper proposed a novel extremum seeking scheme without steady-state oscillation. The scheme greatly improved the dynamic and static performance of the extremum seeking scheme. This scheme not only can converge fast, but also can effectively avoid falling into local extreme points. The most important is that the scheme eliminates the steady-state oscillation which hinders the online applications of the classical schemes. The rigorous stability analysis of the scheme was given and the
Libin Wang received his B.S. degree in Control Science and Engineering from Hebei University of Technology in 2012, and the M.S. degree in Control Science and Engineering from Harbin Institute of Technology (HIT) in 2014. He is now a Ph.D. student at the Control and Simulation Center, HIT, China. His current research interests include extremum seeking, decoupling control, adaptive control, and sliding mode control.
References (20)
- et al.
On the choice of dither in extremum seeking systems: A case study
Automatica
(2008) - et al.
On non-local stability properties of extremum seeking control
Automatica
(2006) - et al.
On global extremum seeking in the presence of local extrema
Automatica
(2009) - et al.
Real-time optimization by extremum-seeking control
(2003) Applications of centre manifold theory
(1981)- Deschênes, J. S. (2012). Demodulation considerations in extremum seeking control loops. In Proceedings of the American...
- et al.
Extremum-seeking control of ABS braking in road vehicles with lateral force improvement
IEEE Transactions on Control Systems Technology
(2014) - et al.
ABS control using optimum search via sliding modes
IEEE Transactions on Control Systems Technology
(1995) Nonlinear systems
(2002)- Krstić, M. (1999). Toward faster adaptation in extremum seeking control. In Proceeding of the 38th IEEE conference on...
Cited by (72)
100 years of extremum seeking: A survey
2024, AutomaticaExtended Kalman filter based extremum seeking algorithm for model-free control of building thermal environment under uncertainty
2023, Journal of Building EngineeringReal-Time optimization as a feedback control problem – A review
2022, Computers and Chemical EngineeringFuture perspective and current situation of maximum power point tracking methods in thermoelectric generators
2022, Sustainable Energy Technologies and AssessmentsCitation Excerpt :As long as the extrema exist, the ESC algorithm works without any input–output information and this is the biggest advantage of the method. Therefore ESC is a model-independent control scheme [127]. ESC is suitable for nonlinearity situations in systems with a local minimum or local maximum.
Extremum Seeking for Stefan PDE with Moving Boundary and Delays
2022, IFAC-PapersOnLine
Libin Wang received his B.S. degree in Control Science and Engineering from Hebei University of Technology in 2012, and the M.S. degree in Control Science and Engineering from Harbin Institute of Technology (HIT) in 2014. He is now a Ph.D. student at the Control and Simulation Center, HIT, China. His current research interests include extremum seeking, decoupling control, adaptive control, and sliding mode control.
Songlin Chen received the M.S. degree in Control Science and Control Engineering from Harbin University of Science and Technology in 2002, and the Ph.D. degree in Guidance, Navigation and Control from Harbin Institute of Technology (HIT) in 2006. He is currently an associate professor at Control and Simulation Center, HIT. His current research interests include high-performance servo systems, nonlinear and adaptive control, active disturbance rejection control, and robot control.
Kemao Ma received his Bachelor degree in engineering from the Department of Information and Control Engineering, Xi’an Jiaotong University, China, in 1992, and his Ph.D. degree from Harbin Institute of Technology, China, in 1998. Since 2006, he has been a professor at the Control and Simulation Center, School of Astronautics, Harbin Institute of Technology, China. His current research interests are nonlinear control, non-smooth analysis and control, robust control and their applications to guidance and control of flight vehicles.
- ☆
This work was partially supported by the Natural Science Foundation of China under grant numbers 61174001, 61321062. The material in this paper was not presented at any conference. This paper was recommended for publication in revised form by Associate Editor Raul Ordóñez under the direction of Editor Miroslav Krstic.
- 1
Tel.: +86 451 86403507; fax: +86 451 86414580.