CAREER: System Synthesis for Self-adaptive Reconfigurable Embedded Systems

为了应对时变的输入特性以及新一代通信系统和多媒体应用中对可升级性的需求，嵌入式系统正通过广泛的配置来提供巨大的灵活性。因此，除了性能、电源和可靠性需求外，嵌入式系统现在还需要硬件重新配置和自适应系统方法。为了提供从一种配置到另一种配置的有效的、负担得起的和无缝的（从应用层）转换，系统综合起着至关重要的作用。要实现这一点，需要在静态编译和动态配置管理之间取得适当的平衡。该项目开发适应性驱动的系统综合工具，并为具有硬件配置支持的自适应嵌入式系统提供动态配置管理层。研究重点是为支持自适应系统的设计工具设计新的算法和优化方法。本课题1)开发了具有自适应意识和考虑重构开销的整体任务调度和资源管理方法；2)开发了具有资源可升级性和配置序列意识的系统级布局规划方法；3)开发了基于当前系统状态调用正确配置的配置管理层高效动态设计方法。这项研究的更广泛的影响是，开发的设计工具可以部署在未来的自适应系统中，用于软件定义的无线电和多媒体应用，其中可编程性和灵活性与计算能力一样重要。PI计划开发新的课程，招募代表性不足的群体参与她的研究，并通过提供在线访问开发的设计工具，促进与其他研究人员和行业的联系。

In order to cope with time-varying input characteristics and demand for upgradability in new-generation communication systems and multimedia applications, embedded systems are being driven to provide tremendous flexibility through extensive configurations. Hence, along with performance, power, and reliability requirements, embedded systems now require hardware reconfiguration and self-adaptive system methodologies. In order to provide an effective, affordable, and seamless (from the application layer) transition from one configuration to another, system synthesis plays a crucial role. To achieve this, there is a need for the right balance between static compilation and dynamic configuration management. This project develops adaptivity-driven system synthesis tools coupled with a dynamic configuration management layer for adaptive embedded system with hardware configuration support. The research focuses on design of novel algorithms and optimization methods for design tools supporting self-adaptive systems. This project 1) develops holistic task scheduling and resource management methods which are adaptive-aware and consider the reconfiguration overhead, 2)develops system level layout planning for resource upgradability and configuration sequence awareness, and 3)develops efficient dynamic design making methods for configuration management layer which evokes the right configuration based on the current system state. The broader impact of this research is that the developed design tools can be deployed in future self-adaptive systems for software-defined radio and multimedia applications where programmability and flexibility are as crucial as computation power. The PI plans to develop new courses, recruit under-represented groups into her research, and facilitate outreach to other researchers and industry by providing online access to the developed design tools.