Volume 50 Issue 7
Jul.  2021
Turn off MathJax
Article Contents

Zou Weiwen, Ma Bowen, Xu Shaofu, Zou Xiuting. Optoelectronic harmonized integration technology for intelligent processing of signal and information (Invited)[J]. Infrared and Laser Engineering, 2021, 50(7): 20211043. doi: 10.3788/IRLA20211043
Citation: Zou Weiwen, Ma Bowen, Xu Shaofu, Zou Xiuting. Optoelectronic harmonized integration technology for intelligent processing of signal and information (Invited)[J]. Infrared and Laser Engineering, 2021, 50(7): 20211043. doi: 10.3788/IRLA20211043

Optoelectronic harmonized integration technology for intelligent processing of signal and information (Invited)

doi: 10.3788/IRLA20211043
  • Received Date: 2021-04-11
  • Rev Recd Date: 2021-05-10
  • Publish Date: 2021-07-25
  • Traditional signal and information processing technologies are relatively independent and complicated. Artificial intelligence (AI) technology introduces a processing scheme of signal conversion plus information recognition to improve the level of intelligence in system processing. However, a high intensity of signals and information in future applications demand for more efficient systems and more flexible decision-making capabilities. It was proposed that optoelectronic integration technology was promising to realize the processing of signal and information as a new processing paradigm. Taking the complementary advantages of photonic and electronic technologies in electromagnetic scales, physical advantages, and practical implementations, overall and direct processing of signal and information was achieved and had the potential to integrate deeper levels of intelligence technology. Emerging signal and information processing paradigms enabled by optoelectronic integration were reviewed. The supportive significance of optoelectronic hybrid integration on optoelectronic integration processing technology was demonstrated.

  • 加载中
通讯作者: 陈斌, bchen63@163.com
  • 1. 

    沈阳化工大学材料科学与工程学院 沈阳 110142

  1. 本站搜索
  2. 百度学术搜索
  3. 万方数据库搜索
  4. CNKI搜索

Article Metrics

Article views(629) PDF downloads(159) Cited by()

Related
Proportional views

Optoelectronic harmonized integration technology for intelligent processing of signal and information (Invited)

doi: 10.3788/IRLA20211043
  • Intelligent Microwave Lightwave Integration Innovation Center (imLic), State Key Laboratory of Advanced Optical Communication Systems and Networks, School of Electronic and Electrical Engineering, Shanghai Jiao Tong University, Shanghai 200240, China

Abstract: 

Traditional signal and information processing technologies are relatively independent and complicated. Artificial intelligence (AI) technology introduces a processing scheme of signal conversion plus information recognition to improve the level of intelligence in system processing. However, a high intensity of signals and information in future applications demand for more efficient systems and more flexible decision-making capabilities. It was proposed that optoelectronic integration technology was promising to realize the processing of signal and information as a new processing paradigm. Taking the complementary advantages of photonic and electronic technologies in electromagnetic scales, physical advantages, and practical implementations, overall and direct processing of signal and information was achieved and had the potential to integrate deeper levels of intelligence technology. Emerging signal and information processing paradigms enabled by optoelectronic integration were reviewed. The supportive significance of optoelectronic hybrid integration on optoelectronic integration processing technology was demonstrated.

  • 智能技术的兴起推动着信号与信息处理策略发生深层次的范式转变。图1(a)给出了信号与信息处理策略的发展趋势示意图。如图1(b)所示,传统的信号与信息处理体系相对独立,往往涉及到多次的初级变换与提取操作才能获得有价信息,导致综合决策的效率较低。近年来,人工智能(AI)技术凭借强大的特征学习与高效的信息识别能力,能够直接实现复杂的处理目标,大大简化了传统处理流程。图1(c)~(d)表明了“信号变换+信息识别”的顺序处理策略。尽管光子辅助的AI加速技术能够缓解部分信息域中的处理压力[1],但是AI高昂的训练成本与受限的处理灵活度使其仍难以用于复杂的决策任务。

    Figure 1.  Schematics of the trend (a) and the schemes (b)-(e) of the integrated processing for signal and information. Figures (b)-(e) correspond to the b-e stages pointed out in Fig. 1(a), respectively

    未来,在5G/6G、国防等复杂应用场景中,信号与信息相互交织且动态变化,将给处理系统带来严峻的挑战:例如,信号输入速率与处理复杂度的增长速度越发难以匹配,系统处理效率与处理功能多样性的矛盾越发突出,高级信息的获取与整合评估越发困难。因此,如图1(e)所示,信号与信息处理必将迈向兼容一体的“融合处理”策略:在不同尺度上对信号与信息进行直接处理,打破单一的处理模式并构建灵活的信号/信息通路,大幅提升系统的紧凑度与处理效率,同时,深度融入高速发展的智能技术,为全局决策目标提供实时、灵活的决策能力。然而,传统的电子学方法电磁尺度有限,并且存在固有的能效瓶颈,仅仅依靠电子处理系统难以支撑信号与信息高效兼容的融合处理。

    光电融合与集成技术有望引领信息与信号处理策略在电磁尺度、物理优势、实现方法等不同层面的融合突破。如图2所示,“集成技术”包含功能集成与芯片集成,是突破系统可靠性、稳定性、实用性的基础,“光电融合”凸显系统的灵活配置,充分发挥光子与电子技术的互补优势:首先,光子与电子融合能够带来信号尺度的广域扩展。电磁频谱应用正在向太赫兹(THz)延伸,密集的频谱通道将成为包罗万象的信息库。让信号“进得来”才能“看得清”,光与电的融合将带来全频谱覆盖的处理能力;其次,光子与电子的物理优势确保在信息尺度上的灵活调控。光子流水式处理速率高,电子精细化处理易操控,相互借力的同时融合深层的智能技术,有望满足高度复杂的处理需求;最后,光子与电子技术在芯片制备上同样具备融合潜力。芯片尺度上光子与电子技术能效互补、成熟度互补,有望突破处理效率与密度瓶颈,实现高效的智能化处理系统。

    Figure 2.  Schematic of optoelectronic integration technology

    近年来,光电融合的趋势推动了传统处理体制的变革,涌现出了全新的处理模式与功能,促进了光电混合集成技术的突破发展。利用有效的电子控制方法,传统单一功能的光子信号处理系统正在朝着可编程、多功能、高效率的方向不断发展[2],为光电融合的智能化信号处理提供支撑;将成熟的光子和电子处理系统进一步有效衔接,显著提高了光电融合下的系统处理性能:Xu等人将AI技术应用于光模数转换(PADC)系统,实现了有效比特位数的跨越式提升[3];多学科交叉下的新兴处理原理更是为光电融合提供了丰富的系统形态,通过将光电子技术深度融入神经网络计算、神经拟态处理等体制,实现了处理能效与能力的突破。光子技术能够实现流水式的零功耗神经网络运算[4],提供高速率的神经拟态动力学响应[5],而电子技术擅长灵活精细的调控,实现了非线性激活[6]、反馈控制[7]等功能,上述光电融合处理技术具有智能化提取高级信息的潜力。光电混合集成技术旨在突破光电器件的材料、工艺、结构兼容性挑战,可分为单片同质集成与异质集成路线:Atabaki等人提出基于硅基材料构建光电集成芯片的方法[8];Wang等人通过外延生长在硅基衬底上制备了单片III-V族-硅基激光器[9]。单片异质集成能够充分发挥不同材料体系的优势,优化功能器件的性能,但是带来了材料兼容性的挑战。混合异质集成的方式虽然能够避免材料间的晶格失配,降低工艺难度,但是牺牲了部分集成度。未来,不断成熟的光电混合集成技术将支撑光电融合处理的高效片上实现。

    综上,光电融合与集成技术具有独特的物理性质、实现方法优势,有望满足不同尺度下的信号与信息智能化融合处理需求。随着光电融合的新兴处理体制、光电混合集成等研究取得突破,迫切需要挖掘并建立深层次的智能融合处理原理及评估体系,同时深度聚焦规模化光电集成系统中的关键挑战,兼顾工艺方法实现与芯片整体性能。

Reference (9)
WeChat followshare

Top

Catalog

    /

    DownLoad:  Full-Size Img  PowerPoint
    Return
    Return