SHF: Small: Fusion of Quantum Dot/Nanowire Based Sensors and Processors in Ultra-low-energy, Distributed-Intelligence Sensing Network

该奖项的目的是开发一种新的纳米电路架构，使用量子电子器件的新型电路实现，以及用于实时电子视觉应用的全面垂直集成设计方法。该架构将纳米光学传感器和主动量子点处理元素集成到具有元级决策代理的混合、超高效、分布式智能系统中。这与传统的电子视觉系统形成鲜明对比，在传统的电子视觉系统中，所有与决策相关的处理都是在纳米线传感器输入数字化后由算法代理完成的。该方法采用平滑的模数处理转换，允许在信号传输到中央决策代理的途中，将输入转换为更简单的多维抽象输入特征的数字表示。可以实现更快速的决策，因为代理现在可以直接关联关键特征，而无需首先识别和提取这些特征或过滤掉无关的细节。这种近源处理还包括所有其他理想的好处，如高能效，低信号退化和芯片实现的小面积要求，除了高速度。具体来说，将输入分散到主动量子电子模拟功能单元的类似细胞神经网络的架构中，将提取并编码关键特征，然后将其分散到一个或多个决策代理中。这些单元包括时空滤波器、速度估计器和图像处理元素。本研究的智力优势包括构建基于纳米级量子点的细胞逻辑阵列，能够进行神经形态计算，如时空信号处理、视频和图像处理；并设计了一种新的CAD工具，用于优化量子隧道器件的3D纳米结构，同时在由首席研究员研究小组开发的增强电路模拟器中执行系统级优化。更广泛的影响包括为下一代电路工程师提供跨学科的教学培训，以及补充教学材料——两本新的、权威的纳米电子学教科书。

The purpose of the award is to develop a new nano-circuit architecture, novel circuit realization using quantum-electronic devices, and a comprehensive, vertically integrated design methodology for real-time electronic vision applications. The architecture integrates nano-optical sensors and active quantum dot processing elements into hybrid, ultra efficient, distributed intelligence systems with meta-level decision making agents. This is in contrast with conventional electronic vision systems where all decision related processing is done by an algorithmic agent only after digitization of the input from nanowire sensors. The approach employs a smooth analog-to-digital processing transition to allow en-route, hierarchical transformation of the input into simpler, digital representations of multidimensional, abstract input characteristics while the signal is on its way to the central decision making agent. A more rapid decision-making can be achieved because the agent can now directly correlate just the key features without first identifying and extracting these features or filtering out extraneous details. Such close-to-source processing also includes all other desirable benefits like high energy-efficiency, low signal degradation and small area requirement for chip implementation, in addition to high speed. Specifically, fanning-out the input to a cellular-neural-network-like architecture of active quantum-electronic analog functional units will extract and encode the key features that can then be fanned in to one or more decision agents. These units include spatio-temporal filters, velocity estimators, and image processing elements. The intellectual merits of this research include the construction of nanoscale quantum dots based cellular logic arrays capable of performing neuromorphic computation like spatiotemporal signal processing, video and image processing; and the design of a new CAD tool for optimizing the 3D nanostructures of quantum tunneling devices while performing the system-level optimization in an augmented circuit simulator developed by the principal investigator's research group. The broader impacts include development of pedagogical interdisciplinary training to the next generation of circuit engineers and supplementary didactic material---two new, definitive textbooks on nanoelectronics.