FPGA-Based High-Performance Embedded Systems for Adaptive Edge Computing in Cyber-Physical Systems: The ARTICo³...

Area of Science:

• Computer Engineering
• Embedded Systems
• Cyber-Physical Systems

Background:

• Cyber-Physical Systems (CPS) are shifting towards distributed processing at the sensing layer (Edge Computing).
• Edge Computing requires hardware platforms balancing high computing performance, energy efficiency, and adaptability.
• SRAM-based FPGAs offer reconfigurability but often lack user accessibility and ease of development.

Purpose of the Study:

• To present an integrated framework for developing FPGA-based high-performance embedded systems for Edge Computing in CPS.
• To address the challenges of user accessibility and development ease in FPGA-based edge solutions.
• To enable dynamic adaptation of computing resources for performance, energy, and fault tolerance trade-offs.

Main Methods:

• Development of a hardware-based processing architecture.
• Creation of an automated toolchain for transparent system generation and management.
• Implementation of a runtime system for dynamic resource adaptation without user intervention.
• High-level system descriptions used for generating reconfigurable systems.

Main Results:

• The framework enables transparent generation and management of reconfigurable systems from high-level descriptions.
• Users can dynamically adapt computing resources to optimize performance, energy consumption, and fault tolerance at runtime.
• The proposed framework is a competitive alternative to software-based edge computing platforms.
• Demonstrated faster solutions and higher energy efficiency for computing-intensive algorithms with data-level parallelism.

Conclusions:

• The integrated framework successfully facilitates the development of FPGA-based edge computing systems for CPS.
• The framework allows for runtime exploration of the solution space defined by performance, energy, and fault tolerance.
• FPGA-based solutions developed with this framework offer significant advantages over software-based alternatives for specific workloads.