SPEEDY: SystemC-Based Design Space Exploration Framework for Embedded Systems

The modern embedded systems have become ubiquitous today with ever-increasing capabilities. Designing an application-specific embedded system has thus become a complex exercise, since the designer needs to optimize hardware resources like computing platform and types of sensors required as well as application-specific metrics like power, performance etc. Increasingly, embedded devices integrate sensors that capture certain external and internal events/conditions with significant adaptation based on their values which is referred to as the context. The designer needs to also consider different possible contexts and the performance of the system in those contexts. In this paper we present our novel contribution which is a framework that explores the design space comprehensively through simulations. SPEEDY is a systemC based framework which simulates the system under consideration allowing for its hardware, software and sensors to be modeled. SPEEDY also calculates the energy consumption of the system as well as for each individual task carried out by the system. SPEEDY supports various sensors where the designer can study different performance metrics for different sensor data. The functions and capabilities of the framework are illustrated by its application to the Mobility Assistance for Visually Impaired (MAVI) system. We address different aspects of the design metrics discussed earlier and show the validity of the framework with experiments.