CAREER: Centaur: A Bio-inspired Ultra Low-Power Hybrid Embedded Computing Engine Beyond One TeraFlops/Watt

本研究的目的是创新一种名为Centaur的嵌入式计算引擎。采用仿生计算模型和先进的忆阻器技术，实现超高能效。主要的技术障碍包括三个组成部分：(1)集成了基于忆阻器的突触网络和Crosbar结构的高能效混合计算系统，分别以灵活和密集的数据处理为目标。(2)Centaur的稳健设计方法，包括电路和算法的改进以及必要的EDA流程。(3)半人马座与现代异质系统的集成及原型演示。还将充分利用Centaur的优势，探索对当今移动和嵌入式系统至关重要的创造性应用，包括模式识别和视频图像处理。这项研究将通过计算体系结构和硬件设计的革命，在功能多样性、能效和成本方面造福于嵌入式系统社区。通过刺激设备工程和计算模型的进步之间的相互作用，这些结果可以进一步造福于半导体和神经形态社会。所开发的技术将在与多个行业合作伙伴的密切合作下转化为主流实践，并直接影响未来的嵌入式系统。协作中的活动还包括在各主要会议上就项目的技术方面和新课程的开发提供指导。该教育计划将通过整合关于新兴的基于忆阻器的计算结构和相关的神经形态计算模型的新模块来加强现有的标准课程。

The objective of the research is to innovate an embedded computing engine named ?Centaur? to achieve ultra-high power efficiency by adopting the bio-inspired computation model and the advanced memristor technology. Three constituent elements are included to address the major technical obstacles: (1) The power-efficient hybrid computing system that integrates memristor-based synapse network and crossbar structure, targeting the flexible and intensive data processing, respectively. (2) The robust design methodology for Centaur, including the circuit and algorithm enhancements as well as the necessary EDA flow. (3) The integration of Centaur into modern heterogeneous systems and the prototype demonstration. Creative applications of critical importance to nowadays mobile and embedded systems by taking the full advantages of Centaur, including pattern recognition and video and image processing, will be also explored.The research can benefit the embedded system community by the revolutions in computing architecture and hardware design for functional variety, power-efficiency, and cost. The results can further benefit the semiconductor and neuromorphic societies at large by stimulating the interaction between the advances in device engineering and computing models. The developed techniques will be transferred to mainstream practices under the close collaborations with several industry partners, and directly impact the future embedded systems. The activities in the collaboration also include the tutorials in the major conferences on the technical aspects of the projects and new course development. The educational plan will enhance the existing standard curricula by integrating new modules on emerging memristor-based computing architecture and the relevant neuromorphic computing model.