Page 64 - IJOCTA-15-3

M. A. Touat / IJOCTA, Vol.15, No.3, pp.435-448 (2025)
While adaptive control can update the knowledge This conversion optimizes FPGA hardware re-
base, it may yield inconsistent results for identi- source utilization and minimizes power consump-
cal uncertainties or disturbances. 5–8 To overcome tion.
this limitation, learning adaptive control strate- Several implementation works on FPGA
gies have been proposed, equipping control mech- based on fuzzy logic have been carried out.
anisms with memory. These methods ensure con- 22–31 In, 22,23 the implementation of a fuzzy con-
sistent results for recurrent situations, thereby re- troller on FPGA using the FIL technique on
ducing computation time. 9–12 the MATLAB-Simulink environment is presented.
The design, simulation, and FPGA implementa-
Although fuzzy controllers demonstrate excel-
tion of a fuzzy controller to regulate the speed of
lent performance in simulation, their practical im- 27 28
a DC motor in real time is presented in. In,
plementation demands high-performance control
genetic and pipeline algorithms are implemented
boards due to significant computational require-
in FPGA to optimize fuzzy controller parameters.
ments. Field Programmable Gate Arrays (FP- 29
In ref., an FPGA design approach with partial
GAs) have emerged as a powerful solution in con-
trol applications, 13 enabling high sampling fre- reconfiguration (PR) using a fuzzy controller is
used to control an electromechanical system.
quencies, parallel data processing, and low power
consumption. 14 Their integration accelerates con- This study presents the implementation of
an FMRLC-based position control strategy for a
trol algorithms, ensuring rapid convergence to- DC motor, incorporating a memory-based con-
ward desired performance. 15 Several studies have
trol mechanism to ensure consistent responses to
explored FPGA-based control implementations. recurring conditions, thereby reducing computa-
For instance, Huerta et al. 16 examined FPGA im- tional overhead. In the first phase, the fuzzy con-
plementation of PID and sliding mode controllers
for DC-DC buck converters. Wang et al. 17 de- troller is designed in MATLAB/Simulink and im-
veloped a neural network-based PID controller plemented on a Zedboard Zynq-7000 FPGA us-
for motion control systems. Similarly, Li et al. 18 ing the FIL technique. In this setup, the FMRLC
investigated FPGA-accelerated predictive control controller runs on the FPGA, while the remain-
for autonomous driving. Meghwal et al. 19 pre- ing control system components operate within
sented a robust fuzzy controller implemented on Simulink to assess performance and robustness
FPGA for matrix converter-based induction mo- under various disturbances. The second phase in-
tor control, while Attafi et al. 20 focused on real- volves deploying the algorithm onto the FPGA
board using HDL Coder’s ”IP Core Generation”
time FPGA implementation of robust control for
technique for real-time implementation. This
wind energy conversion systems.
method allows direct integration of the control
In this paper, a learning adaptive control strategy onto the Zedboard by generating an IP
strategy, namely the Fuzzy Model Reference block via Vivado software, eliminating the need
Learning Control (FMRLC), is implemented on for manual VHDL coding and significantly reduc-
a Zedboard Zynq-7000 FPGA for DC motor posi- ing development time. ”IP Core Generation” au-
tion control. The proposed controller is first vali- tomatically generates the VHDL code, compiles
dated in a real-time simulation environment using the bitstream programming file, and transfers it
the FPGA-in-the-Loop (FIL) approach, a tech- to the Zedboard.
nique widely employed in laboratories to assess The structure of this paper is as follows: Sec-
prototype controllers under real operating condi- tion 2 provides an overview of the FMRLC con-
tions. Compared to conventional numerical simu- trol strategy. Section 3 details the mathematical
lations, FIL provides a more realistic validation by modeling of the DC motor in both theoretical and
incorporating real-world factors such as noise, dis- experimental contexts. Section 4 describes the de-
turbances, and implementation challenges, which sign and FPGA implementation of the FMRLC
may not be accounted for in idealized simulations. controller. Section 5 presents simulation results
in FIL mode along with experimental validation.
The proposed methodology leverages MAT-
LAB/Simulink for controller design and FPGA Finally, the conclusions are summarized in Sec-
implementation. Simulink, in conjunction with tion 6.
HDL Coder, enables direct FPGA deployment 2. Architecture of the control strategy
without requiring VHDL programming. 21 How-
ever, since floating-point data types in Simulink The Fuzzy Model Reference Learning Controller
are not supported by all FPGA platforms, the (FMRLC) is shown in Figure 1. As illustrated , it
32
Fixed-Point Tool is utilized to convert floating- has three main parts as defined by : Fuzzy Direct
point designs into fixed-point representations. Controller (FC) synthesized to control the DC
436