Related Experiment Videos
Simulink-based HW/SW codesign of embedded neuro-fuzzy systems
L M Reyneri1, M Chiaberge, L Lavagno
1Dipartimento di Elettronica - Politecnico di Torino, Italy. reyneri@polito.it
International Journal of Neural Systems
|September 30, 2000
Summary
This study introduces a semi-automatic hardware/software codesign flow for creating efficient Neuro-Fuzzy embedded systems. It enables rapid prototyping and design while optimizing for low power and cost.
Area of Science:
- Embedded Systems Engineering
- Computer Architecture
- Artificial Intelligence
Background:
- Neuro-Fuzzy systems are increasingly used in embedded applications.
- Designing these systems requires balancing hardware and software trade-offs for performance, power, and cost.
- Existing design flows may lack flexibility across different implementation technologies.
Purpose of the Study:
- To propose a semi-automatic hardware/software (HW/SW) codesign flow.
- To facilitate the design of low-power and low-cost Neuro-Fuzzy embedded systems.
- To support diverse implementation technologies including software, digital, and analog.
Main Methods:
- Utilized The Mathworks' Simulink environment for functional specification and performance analysis.
- Developed a codesign flow capable of exploiting HW/SW and digital/analog implementation trade-offs.
- Generated C, VHDL, and SKILL code for selected architectures.
Main Results:
- Demonstrated a flexible codesign flow applicable to various technologies and architectures.
- Enabled high-speed simulation of Simulink models and rapid design of Neuro-Fuzzy devices.
- Facilitated analysis of critical performance criteria: timing, power, size, and cost.
Conclusions:
- The proposed semi-automatic HW/SW codesign flow effectively supports the development of optimized Neuro-Fuzzy embedded systems.
- The flow allows designers to explore and exploit design trade-offs for low-power and low-cost solutions.
- It enhances the prototyping and simulation capabilities for embedded system design.