Hardware Implementation of Embedded Fuzzy Controllers on FPGAs and ASICs - Fuzzy Modeling and Control: Theory and Applications - page 226

this limitation, a considerable number of dedicated hardware structures, known as
fuzzy accelerators or fuzzy coprocessors, were developed at the nineties. Among the
most popular fuzzy coprocessors available on the market deserve mention the SAE
81C99 and SAE 81C991 families from Siemens (Eichfeld et al. 1996 ), the FP1000,
FP3000, and FP5000 fuzzy coprocessors from Omron (Shimizu et al. 1992 ), and
the W.A.R.P. architecture from ST Microelectronics (Pagni 1998 ). Fuzzy coproces-
sors disappeared from the market at the end of the nineties as a consequence of
the increasing speed of conventional processors. Dedicated hardware realization of
fuzzy controllers allows achieving the requirements of size, power, and inference
speed imposed by complex embedded controllers operating under real-time condi-
tions. Since the first fuzzy hardware proposals from Togai and Watanabe ( 1986 ) and
Yamakawa and Miki ( 1986 ) in the mid-eighties, many microelectronic implementa-
tions of fuzzy controllers have been described in the literature (Basterretxea and del
Campo 2009 ; Zavala and Nieto 2012 ). Most of these controllers were implemented
as application specific integrated circuits (ASICs) using analog or digital design
techniques. More recently, the advances in field programmable gate arrays (FPGAs)
have promoted the use of this kind of devices both for rapid prototyping and as final
implementation of fuzzy controllers (Kim 2000 ; McKenna and Wilamowski 2001 ;
Li et al. 2003 ; Sánchez-Solano et al. 2007 ; Kung et al. 2009 ; Taeed et al. 2012 ).
The success of fuzzy logic in practical control applications has also encouraged
the development of specific tools dedicated to the design of fuzzy inference systems.
Most of these tools are focused on the creation of fuzzy systems from a data set,
the tuning of system parameters using learning algorithms, and the comparison of
different fuzzy operators and inference approaches (Nurnberger et al. 1999 ; Alonso
et al. 2004 ; Guillaume and Charnomordic 2011 ). In addition, tools for automatic syn-
thesis of fuzzy systems using specific hardware based on analog and digital design
techniques have been also proposed (Hollstein et al. 1996 ; Carvajal et al. 1997 ;Kim
and Cho 1997 ;Reetal. 2000 ). First digital approaches were based on the generation
of general-purpose or device-specific VHDL code. However, in the last years a great
number of proposals use the facilities provided by the /Simulink environ-

ment to apply a model-based methodology for the design and hardware implementa-
tion of fuzzy controllers (Bakhti and Benbaouche 2006 ; Altas and Sharaf 2007 ;Lu
and Zhang 2010 ; Sánchez-Solano et al. 2013 ).
This chapter presents two hardware synthesis tools for fuzzy controllers provided
by the Xfuzzy environment (Xfuzzy 2013 ). The main characteristics of this develop-
ment environment are introduced in Sect. 13.2 , which also includes the description
of the hardware architecture supported by both synthesis tools. xfsg uses the Xilinx's
SystemGenerator (XSG) tool (SysGen 2010 ) to provide implementations on Xilinx's
FPGAs taking advantage of the flexibility and ease of configuration offered by the
Matlab

Matlab

/Simulink environment. This tool and the associated cell library developed
in Simulink are presented in Sect. 13.3 . xfvhdl employs a strategy based on VHDL
and generates code that can be synthesized and implemented on ASICs or FPGAs.
This tool, as well as the VHDL cell library supporting it, is detailed in Sect. 13.4 .
Section 13.5 describes the design flow associated to both strategies and compares
the obtained results in terms of approximation capability and resource consump-