Automated Model-Based Optimization of Data-Adaptable Embedded Systems

Abstract

Dynamic data-driven applications such as object tracking, surveillance, and other sensing and decision applications are largely dependent on the characteristics of the data streams on which they operate. The underlying models and algorithms of data-driven applications must continually adapt at runtime to changes in data quality and availability to meet both functional and designer-specified performance requirements. Given the dynamic nature of these applications, point solutions produced by traditional design tools cannot be expected to perform adequately across varying execution scenarios. Additionally, the increasing diversity and interdependence of application requirements complicates the design and optimization process. To assist designers of data-driven applications, we present a modeling and optimization framework that enables developers to model an application's data sources, tasks, and exchanged data tokens; specify application requirements through high-level design metrics and fuzzy logic--based optimization rules; and define an estimation framework to automatically optimize the application at runtime. We demonstrate the modeling and optimization process via an example application for video-based vehicle tracking and collision avoidance. We analyze the benefits of runtime optimization by comparing the performance of static point solutions to dynamic solutions over five distinct execution scenarios, showing improvements of up to 74% for dynamic over static configurations. Further, we show the benefits of using fuzzy logic--based rules over traditional weighted functions for the specification and evaluation of competing high-level metrics in optimization.