光子计算关键技术研究

论文总字数：49917字

摘 要

本文利用光子计算关键技术针对通信领域的相干光PSK判决问题设计光子计算方案，并通过仿真说明设计模型的性能和特点。

首先本文介绍光子计算的发展背景和发展历史，对当前已有的光子计算方案进行总结，归纳出当前各类方案的优缺点和适用范围。接下来本文针对通信领域中相干光PSK判决问题探寻光子计算与通信相结合的方案。根据下一代移动信息系统的需求，光通信系统将朝向全光网络发展，而传统的相干光通信系统需要将光信号转换为电信号才能够完成最终判决，通过光子计算可以直接以光信号的形式输出判决结果。

当前光子计算的一个重要应用领域是机器学习，尤其是在实现神经网络的前馈运算方面光子计算可以表现出其在速度和能耗上的优势。从上述应用领域出发，本文详细探讨了将光子计算以及神经网络应用于相干光PSK信号判决问题的思路，具体对BPSK、QPSK以及8PSK判决问题进行了分析，并分别设计光子电路结构。在设计光学网络时，本文对当前的光子计算方案进行了改进使其适用于光信号处理。设计完成后，本文针对BPSK、QSPK和8PSK的判决光子电路搭建仿真模型，分别进行测试，验证各模型的性能。

关键词：光子计算，集成光子电路，相干光通信系统，PSK，人工神经网络

Abstract

In this paper, an optical computing scheme is designed for the coherent optical PSK decision problem in the communication field by using the key technology of optical computing. The performance and characteristics of the proposed model are illustrated by simulation.

This paper introduces the background and history of the research in optical computing, enumerates the existing optical computing schemes, and summarizes the advantages and disadvantages as well as the applicable scope of the current schemes. This paper explores the combination of optical computing and communication for the coherent optical PSK decision problem in the communication field. According to the requirements of the next generation mobile information system, optical communication will be developed towards all-optical networks. Traditional coherent optical communication systems need to convert optical signals into electrical signals to complete the final decision. Optical computing can directly output the decision result in the form of optical signals.

An important application area of optical computing is machine learning, especially in the implementation of neural network feedforward operations where optical computing can exhibit its advantages in speed and energy consumption. Starting from the above application field, this paper discusses in detail the application of optical computing and neural network to the decision problem of optical PSK signal. The BPSK, QPSK and 8PSK decision problems are analyzed, and the photonic circuit structures are designed separately. When designing the optical network, this paper improves the current optical computing scheme and makes it suitable for optical signal processing. After the design is completed, this paper builds simulation models for BPSK, QSPK and 8PSK decision problems, and tests them separately to verify the performance of each model.

KEY WORDS: optical computing, integrated photonic circuit, coherent optical communication system, PSK, artificial neural network

目 录

摘要......................................................................................................................................................Ⅰ

Abstract................................................................................................................................................Ⅱ

1 绪论 1

1.1 研究背景及意义 1

1.2 国内外研究分析 2

1.2.1 光学信息处理技术 2

1.2.2 光学人工神经网络 4

1.2.3 通用光学数字逻辑电路 7

2 传统相干光PSK判决电路 10

2.1 系统模型 11

2.2 90度光混频器 12

2.3 平衡接收机 13

2.4 同步电路 14

3 光子计算和光学神经网络 17

3.1 通用线性光学系统 17

3.1.1 马赫曾德干涉仪 17

3.1.2 有限维酉变换 18

3.1.3 通用线性变换 20

3.1.4 可编程纳米光子集成电路 21

3.2 光学神经网络模型 21

3.2.1 系统模型 21

3.2.2 光学干涉单元 24

3.2.3 光学非线性单元 24

3.2.4 实验验证 25

4 基于ONN的相干光PSK判决光子电路 27

4.1 系统模型 27

4.1.1 判决光子电路 27

4.1.2 同步电路 28

4.2 BPSK判决光子电路 29

4.2.1 系统模型 30

4.2.2 模型分析 31

4.2.3 光子电路仿真 34

4.2.4 线性光学的局限 35

4.3 QPSK判决光子电路 35

4.3.1 系统模型 35

4.3.2 神经网络结构 36

4.3.3 数据集构建 37

4.3.4 训练 38

4.3.5 模型分析 39

4.3.6 光子电路仿真 41

4.4 8PSK判决光子电路 43

4.4.1 神经网络结构 44

4.4.2 模型分析 46

4.5 下一步工作 48

结论 49

参考文献 50

致谢 52

剩余内容已隐藏，请支付后下载全文，论文总字数：49917字