Showing papers by "Shun-Feng Su published in 2009"

Dynamic Slip-Ratio Estimation and Control of Antilock Braking Systems Using an Observer-Based Direct Adaptive Fuzzy–Neural Controller

Abstract: This paper proposes an antilock braking system (ABS), in which unknown road characteristics are resolved by a road estimator. This estimator is based on the LuGre friction model with a road condition parameter and can transmit a reference slip ratio to a slip-ratio controller through a mapping function. The slip-ratio controller is used to maintain the slip ratio of the wheel at the reference values for various road surfaces. In the controller design, an observer-based direct adaptive fuzzy-neural controller (DAFC) for an ABS is developed to online-tune the weighting factors of the controller under the assumption that only the wheel slip ratio is available. Finally, this paper gives simulation results of an ABS with the road estimator and the DAFC, which are shown to provide good effectiveness under varying road conditions.

Fuzzy Sliding Controller Design with Adaptive Approximate Error Feedback

Abstract: In this paper, a novel state error feedback sliding controller is proposed. In the controller, an optimal feedback gain is required and in this study it is assumed to be unknown. Usually, a rudimentary feedback gain is used. Besides, in order to approximate the state error feedback sliding controller with the optimal feedback gain, an adaptive fuzzy system is employed. Thus, the proposed control scheme consists of an adaptive fuzzy system and a state error feedback sliding controller with a rudimentary feedback gain. In the system framework, the rudimentary state error feedback sliding controller can be viewed as the approximate error estimator of the adaptive fuzzy system. Therefore, such an estimated error can be fed back to the learning of the fuzzy system through a modified adaptive law. With such an approximate error feedback, it is clearly evident from our simulation that the learning speed of the proposed learning scheme is faster than that of the original scheme. Also, with the proposed controller, the system stability not only is guaranteed, but also becomes more stable. time. Generally, the conventional sliding controller design has a large and inelastic control component to guarantee the system stability. It is often obtained based on the upper bounds of the system uncertainties. Since the robust controller can only use such bounds to ensure the stability if no other information is used, chattering phenomena seems unavoidable [18-20]. Please note that the integral sliding surface [29] or other sliding surfaces [30-31] are not considered in this study. An idea of resolving the above problem is to incorporate certain learning capability into the system to provide information for the robust controller. In the sliding control design for uncertain systems, adaptive approximation techniques are usually employed as learning mechanisms. In this kind of approaches, numerical [27-28, 32-34] or intelligent [12-17] approximation systems are used to estimate unknown parameters in the control law [12-17, 26-28] or to adjust the boundary layer to eliminate chattering phenomena [21-26]. In this study, an adaptive fuzzy approximation system is considered. Adaptive fuzzy systems are adaptive systems with the incorporation of linguistic fuzzy information in a form of fuzzy IF-THEN rules [5-6][8-9] and are usually employed to estimate some elements in the so-call equivalent controller. Basically, these design approaches are similar to that of indirect adaptive fuzzy control systems [6][10-11]. In this paper, a novel idea for adaptive fuzzy sliding control systems is proposed. By referring to [14], it can be found that the stable (sliding) condition, which is the negative Lyapunov energy derivative, is the key idea of the system stability. Thus, it is possible to consider the traditional stable process of approaching the sliding surface only in the design process. Of course, this design direction may violate the basis of the variable structure system design [7]. But, it may exist interesting and useful design methodologies under such a deign philosophy.

A Dynamic Hierarchical Fuzzy Neural Network for a General Continuous Function

Abstract: A serious problem limiting the applicability of the fuzzy neural networks is the ”curse of dimensionality”, especially for general continuous functions. A way to deal with this problem is to construct a dynamic hierarchical fuzzy neural network. In this paper, we propose a two-stage genetic algorithm to intelligently construct the dynamic hierarchical fuzzy neural network (HFNN) based on the merged-FNN for general continuous functions. First, we use a genetic algorithm which is popular for flowshop scheduling problems (GA_FSP) to construct the HFNN. Then, a reduced-form genetic algorithm (RGA) optimizes the HFNN constructed by GA_FSP. For a real-world application, the presented method is used to approximate the Taiwanese stock market.

Novel direct adaptive fuzzy control system design

Abstract: In our earlier study, the property of L 2 -gain robust control is incorporated into the direct adaptive fuzzy control design with the use of the compensative controller. The tracking control performance is guaranteed. In the study, we further employ the L 2 -gain property to improve the adaptive law design. Due to the L 2 -gain control property, the approximate error of the fuzzy system is estimable, and thus, an adaptive law with the approximate error feedback is proposed in the study. The novel adaptive law improves the learning speed so that the learning performances can be better. Since the L 2 -gain based compensative controller can guarantee the tracking control performance and remove the effect of the gain function in the adaptive law. The direct adaptive fuzzy control system design of the paper not only has the satisfactory tracking control performance but also has the better learning performance. Various simulations are conducted to demonstrate the effectiveness of the proposed design.

Adaptive fuzzy sliding controller design with approximate error feedback

Abstract: The research about sliding based approaches is a widely studied topic to the adaptive fuzzy control system designs. In this paper, a novel state error feedback sliding controller is proposed. An optimal feedback gain is required and in this study it is assumed to be unknown. Usually, a rudimentary feedback gain is used. Besides, in order to approximate the state error feedback sliding controller with the optimal feedback gain, an adaptive fuzzy system is employed. Thus, the proposed control scheme consists of an adaptive fuzzy system and a state error feedback sliding controller with a rudimentary feedback gain. In the system framework, the rudimentary state error feedback sliding controller can be viewed as the approximate error estimator of the adaptive fuzzy system. Therefore, such an estimated error can be fed back to the learning of the fuzzy system through a modified adaptive law. With such an approximate error feedback, it is clearly evident from our simulation that the learning speed of the proposed learning scheme is faster than that of the original scheme. Also, with the proposed controller, the system stability not only is guaranteed, but also becomes more stable.

Genetic adaptive scheme design for high initial gain problems in a L 2 -gain state feedback controller

Abstract: This paper is a study of a genetic adaptive scheme design for L 2 -gain state feedback controllers It is known that the design of the initial gain producer of the L 2 -gain state feedback controller (LC) is a difficult problem The derivative-free optimization, the genetic algorithm, is utilized to resolve the high initial gain problem of LC for a class of nonlinear systems It is a novel approach for robust control and can be considered as a special application of genetic algorithms A real-value genetic algorithm with on-line characteristics is designed to search a suitable control gain of LC under auxiliary searching conditions and a specific cost function The specific cost function is designed under Lyapunov stable theory Since the system has the L 2 -gain control properties, then the system states are bounded in an assignable region so that the stability of the initial system is guaranteed Thus, the system stability of any searched results is guaranteed Besides, due to the assignable L 2 -gain attenuation level, the search space of the genetic algorithm is definable The search target of the genetic algorithm is to find a suitable set of the initial gain so that the system can have required initial control performance The simulation results indeed demonstrate the effectiveness of the proposed approach