A FUZZY LOGIC BASED MULTIPLE REFERENCE MODEL ADAPTIVE CONTROL Sukumar Kamalasadan Adel A. Ghandakly Khalid Al-Olimat Dept. of Electrical Engineering and Comp. Science Dept. of Electrical and Computer Engineering University of Toledo Ohio Northern University Toledo, Ohio, 43606, USA Ada, Ohio, 45810, USA skamalas@eng.utoledo.edu adel.ghandakly@utoledo.edu k-al-olimat@onu.edu Abstract This paper presents a fuzzy logic approach for switching multiple reference models, within the Model Reference Adaptive Control (MRAC) framework, in response to major changes in the plant operating conditions. Following a rule base, the fuzzy switching scheme effectively monitors changes in operating conditions or such drastic changes in plant. A fuzzy inference engine then fires appropriate rules, which gives a fuzzified output value. Defuzzification is then performed to switch the reference model in a predefined domain. The whole process is conducted on line, monitoring the plant auxiliary inputs at selected control intervals. The main contribution of the paper is that the proposed fuzzy switching scheme can be performed online and is very well suitable for multi modal jump systems. Unlike, static multiple model algorithms for switching (non-interacting individual model-based filters) or switching dynamic algorithms (susceptible to numerical overflow), this scheme provides an interactive multiple model environment with soft switching. The scheme is computationally feasible, effective and efficient. The proposed scheme is applied to an example system with disturbed model parameters to show its effectiveness. 1 INTRODUCTION Control of multi modal systems is often difficult and has captured great deal of attention in recent years. Such control calls for a dynamic control system, which efficiently determines the system changes, indirectly, keeping track of uncertainties and takes appropriate control action every instant. Multiple model adaptive control has become one of the effective solutions for controlling such systems [11]. Basically it designs a set of models to represent system behavior patterns at different modes, covering normal parametric variations, external disturbances, or both. Further, depending on certain schemes best suitable model at a certain instance of time is switched and a control action has been carried out based on this reference model. Research directions in MMAC has two forms; either with mathematical formulation or with heuristic technique. In a stochastic representation, reference [5] proposed a numerically robust implementation of standard multiple model estimation algorithms, which are otherwise often prone to numerical underflows. Authors highlight the computational complexity in dealing with probabilistic switching. In [10], the authors address an alternate way of computing conditional probabilities using a residual correlation kalman filter banks for each modeled hypotheses. In a deterministic fashion, reference [6] proposes different switching and tuning schemes combining fixed and adaptive models. Subsequently in reference [7] the same research group deals with the adaptive control of LTI discrete-time system using multiple models. On the application side, reconfigurable flight control system using MMAC method is proposed in [8], which demonstrates consistently effective reconfiguration capabilities when subjected to actuator and sensor failure. The work done in [9] is basically an application of MMAC algorithm to control F-8c aircraft, proving MMAC is a reasonable candidate. In the development of heuristic switching algorithm, reference [4] proposes a fuzzy logic based switching scheme and applies it to electric machine and robotic manipulator control. Though this scheme is effective, the switching scheme has been performed at predefined time interval. In this paper an alternate way of heuristic based fuzzy logic switching scheme is proposed. This scheme performs switching of the reference model structure at every time instant along with plant variations providing a smooth change in the functional relation between them. The main idea is to change the reference model so that it improves the overall performance in the form of perfect tracking without affecting the stability in the prescribed domain of interest. In section 2 we provide MMAC concept followed by the proposed fuzzy switching scheme in section 3. Further section 4 shows the simulation results for selected cases to demonstrate the effectiveness of the proposed scheme and conclusions are in section 5. In Appendix A we will show the standard set of MRAC equations which is effective for the problem of interest once the reference model can be switched wisely. 2 MULTIPLE MODEL CONCEPTS Let a system characterizes ‘n’ models, which conveniently fits in the range of system parametric space