Automated Firmware Generation for Compressive Sensing on Heterogeneous Hardware

Area of Science:

• Embedded Systems Engineering
• Signal Processing
• Data Acquisition

Background:

• Developing firmware for complex sampling schemes, particularly for compressive sensing (CS), is challenging.
• Existing methods often require significant expertise and development time.
• There is a need for streamlined solutions to facilitate the adoption of CS in embedded applications.

Purpose of the Study:

• To present a model-based firmware generator for creating both fixed-rate and variable-rate compressive sensing acquisition schemes.
• To automate the generation of pseudo-random sampling sequences for CS firmware based on undersampling factors.
• To reduce the complexity of developing embedded CS systems and lower the barrier to entry for practical field use.

Main Methods:

• A model-based framework is used, defining acquisition sequences and target platforms.
• The framework automatically generates functional firmware from model definitions and specifications.
• For CS, it generates pseudo-random sampling schemes tailored to specified undersampling factors.

Main Results:

• Successful automatic generation of fixed-rate Shannon-compliant and variable-rate CS acquisition firmware.
• Demonstrated firmware generation for CS with pseudo-random sampling aligned with undersampling factors.
• An example use-case evaluated the generated firmware, including a synchronization strategy for CS setups.

Conclusions:

• The model-based firmware generator effectively simplifies the development of complex sampling schemes for embedded systems.
• The framework facilitates the implementation of compressive sensing with reduced complexity and improved usability.
• This approach aims to accelerate the adoption and effective deployment of embedded CS technologies in various fields.