外文翻譯--水下運載工具模糊邏輯控制器的簡單設(shè)計方法(英文）

1、A simplified approach to design fuzzy logic controller for an underwater vehicleK. Ishaque n, S.S. Abdullah, S.M. Ayob, Z. SalamFaculty of Electrical Engineering, Universiti Teknologi Malaysia, UTM 81310, Skudai, Johor B

2、ahru, Malaysiaa r t i c l e i n f oArticle history:Received 14 September 2009Accepted 23 October 2010Editor-in-Chief: A.I. Incecik Available online 13 November 2010Keywords:Fuzzy logic controllerSigned distance methodSin

3、gle input fuzzy logic controlUnderwater vehiclea b s t r a c tFuzzy logic controller (FLC) performance is greatly dependent on its inference rules. In most cases, themore rules being applied to an FLC, the accuracy of th

4、e control action is enhanced. Nevertheless, a large setof rules requires more computation time. As a result, an FLC implementation requires fast and highperformance processors. This paper describes a simplified control s

5、cheme to design a fuzzy logiccontroller (FLC) for an underwater vehicle namely, deep submergence rescue vehicle (DSRV). Theproposed method, known as the single input fuzzy logic controller (SIFLC), reduces the convention

6、al two-input FLC (CFLC) to a single input FLC. The SIFLC offers significant reduction in rule inferences andsimplifies the tuning process of control parameters. The performance of the proposed controller isvalidated via

7、simulation by using the marine systems simulator (MSS) on the Matlab/Simulinks platform.During simulation, the DSRV is subjected to ocean wave disturbances. The results indicate that the SIFLC,Mamdani and Sugeno type CFL

8、C give identical response to the same input sets. However, an SIFLC requiresvery minimum tuning effort and its execution time is in the orders of two magnitudes less than CFLC.this is primarily due to the difficult and u

9、npredictable environmental conditions that exist in the sea. During operation, the UUV under- goes complex multi-axis motion trajectories that are highly non- linear, because the subsystems in the vehicle are ill-defined

10、 and are strongly coupled with each other Goheen and Jefferys (1990). Furthermore, the vehicle dynamics can change considerably with the changes in surrounding conditions and external disturbances, such as wind velocity

11、and sea current. These hydrodynamic coefficients are normally difficult to measure or predict accurately (Humphreys and Watkinson, 1982; Abkowitz, 1969; Lewis et al., 1984). There have been various efforts to develop the

12、 controller for the UUV, which include both the (conventional) linear and the (modern) intelligent control schemes. Given the complexities of its control requirements, it is clear that the linear controller is unable to

13、control the vehicle satisfactorily (Yoerger and Slotine, 1985). Intelligent control methods which include neural networks (NN), sliding mode (SLC) and fuzzy logic controllers (FLC) are morerobust and are able to adopt th

14、e hydrodynamics uncertainties. In addition, they exhibit excellent immunity to disturbances. An automatic guidance system is proposed by Russel and Bugge (1981) and Yuh (1990) to compensate the parameters changes in UUVs

15、. Yoerger and Slotine (1985), Yoerger et al. (1991) successfully developed a sliding mode controller for the UUVs. The sliding mode approach was further improved by Fossen and Sagatun (1991) and Healey and David (1993).

16、Other intelligent controls include the work of Yuh (1990), Ishiii et al. (1998) and Kim and Yuh (2001). Wang et al. (2000), attempted to apply the neuro-fuzzy controller called the self-adaptive neuro-fuzzy inference sys

17、tem (SANFIS). This is followed by a different approach using fuzzy membership function-based neural networks (FMFNN). It is based on the work proposed by Suh and Kim (1994, 2000) for nonlinear control or nonlinear functi

18、on approximation. The FMFNN combines the advantage of both FLC and NN. Some other works that apply an FLC in UUV are due to DeBitetto (1994) and Kato (1995). Although intelligent control is very promising for UUV applica

19、- tion, it requires substantial computational power, due to the complex decision making processes. For example, an FLC has to deal with fuzzification, rule base storage, inference mechanism and defuzzification operations

20、. Despite these issues, it is known that an FLC has a simple control structure and offers higher degree of freedom in tuning its control parameters compared to other nonlinear controllers Liu and Lewis (1993). To take fu

21、ll advantage of an FLC, its computational requirements need to be reduced. In this paper, a simplification of the conventional fuzzy con- troller (CFLC) is proposed. The method is called the single input fuzzy controller

22、 (SIFLC). The simplification is achieved by applyingContents lists available at ScienceDirectjournal homepage: www.elsevier.com/locate/oceanengOcean Engineering0029-8018/$ - see front matter fax: +60 7 5566272.E-mail ad

23、dresses: kashif@fkegraduate.utm.my (K. Ishaque),shahrum@utm.my (S.S. Abdullah).Ocean Engineering 38 (2011) 271–284approximated as a piecewise linear surface (PWL), if the following conditions hold:(a) the input membershi

24、p function is triangular shape (b) the output membership function is singleton shape (c) the fuzzification and defuzzification process uses Centre of Gravity (CoG) method.Using these operating conditions, the output equa

25、tion _ uo of the SIFLC can be derived as follows. Consider the triangular shape input and singleton shape output memberships as Fig. 3(a) and (b), respectively. If L?2, L?1, L0, L1 and L2 are the input membership functio

26、ns and S?2, S?1, S0, S?1 and S?2 are the output singleton membership functions, its rules inference can be written as in Table 3. Referring to Fig. 3(a), X?3, X?2, X?1, X0, X1, X2 and X3 are denoted as the peak locations

27、 of L?3, L?2, L?1, L0, L1, L2 and L3 membership functions, respectively. Consider x as the measured distance d input within the Universe of Discourse (UoD) and is also a member to L0 and L1 membership functions. The outp

28、ut _ uo can be calculated using the Centre of Gravity (CoG) operator as_ uo ¼P 2i ¼ ?2miSiP ni ¼ 1mi: ð4ÞIn Eq. (4), mi is the membership degree value for ith membership function. Since x is a me

29、mber of L0 and L1 membership functions,then L?3, L?2, L?1, L2 and L3 will have zero membership degree value. Hence, Eq. (4) can be written as_ u0 ¼ m0S0 þm1S1 m0 þm1 ð5Þ_ uo ¼ S1?S0 X1?X0? ?

30、xþ X1S0?X0S1 X1?X0? ?: ð6ÞEq. (6) demonstrates that the output equation of an SIFLC is a linear function. It can be rewritten in a more generalized form as_ uo ¼ adþg: ð7ÞIn Eq. (7), d

31、is the input distance variable and a is the slope of the line. The variable g is the output value when input d is zero, i.e. g ¼ _ u09d ¼ 0. Both a and g parameters are defined asa ¼ S1?S0 X1?X0 and g 

32、8; X1S0?X0S1 X1?X0 : ð8ÞAs can be observed from Eq. (8), the output equation is a function of peak locations of input and output membership func- tions. A piecewise linear (PWL) control surface can be obtained.

33、 The PWL control surface can be simply constructed using a look-up++ re(k)e(k)d(k) u(k)1Signed-distance methodFLCinference p rulesDefuzzification1+?21+?2 level p fuzzification?uo (k)Fig. 2. The SIFLC control structure.In

34、put d X1 d = x1 L0 L1 L2X3L3 L-3 L-2 L-1X-3 X-2 X-1 X0 X2Output uo1???L1 ?L0S-3 S-2 S-1 S0 S1 S2 S3Fig. 3. (a) Input membership function and (b) output membership function.Table 3Rule table for the above example.d L?3 L?