AI for Hardware Verification

Hardware Verification（RTL）with Artificial Intelligence(AI) and Machine Learning(ML)

My current design focuses on the application of Artificial Intelligence and Machine Learning (Artificial Neural Networks, Reinforcement Learning etc) to hardware RTL verification before netlist (Physical) verification.

AI/ML based verification technology is now one of the key areas of concerns for the verification engineers. My aim is the “Hardware RTL Verification Optimization”, that is to improve the verification process and to reduce the test time by
optimizing stimulus generation to minimize test time
optimizing stimulus generation to maximize coverage
&
accelerating coverage closure (accelerating coverage directed test generation)

using popular AI/ML algorithms such as Artificial Neural Networks and Reinforcement Learning.

Python will be my primary AI/ML language, because of its dominant position in the world of AI/ML applications.

Some of my favorite Python libraries for AI/ML include:

TensorFlow
PyTorch
Scikit-learn
Keras
NumPy
OpenAI Gymnasium
Stable Baselines3

Verilog/SystemVerilog/VHDL are my primary HDL/HVL languages, as they are the most popular hardware description and verification languages in the industry and research communities.

UVM is my first choice verification methodology but others such as OSVVM, UVVM, cocotb, ABV (SVA, PSL) etc can also be complimented as the alternatives.

Several communication schemes between Python and HDLs/HVLs will be employed to handle the interaction between them, I call these as Python – HDL bridges, details will be discussed later on.

Custom AI/ML modules can be developed for your testbenches developed in the above VMs and languages or in some of the other HLS (High Level Synthesis) languages such as SystemC,Python and others.

My ultimatum aim is to develop comprehensive,and even general and universal AI/ML systems which can carry out benchmarking operations of various AI/ML algorithms for various instances of RTL test cases against each other and also against non-AI, common verification plans. If possible the systems will be extended to other phases of the IC design and verification and even to non-IC applications such as medical and robotic systems. These systems or modules may take various forms such as source codes, standallone apps with or without a GUI, an online platform etc. Specialized AI/ML based verification acceleration and improvement services will also be provided based on these technologies and systems.

##############################################
Earlier IC Research, Design and Development
##############################################

I started designing ICs since 2013.

I have been involving in AI/ML well also since 2013, when I was studying the graph-theoretical aspects of IC (integrated circuits) design, where AI is also key to developing EDA algorithms for VLSI design. On those early days my IC design activities have also been involving lots of AI/ML techniques and algorithms, as well as my quantum circuits design for example qubit routing which employs deep reinforcement learning to improve the routing efficiency.

However, my IC career has been undergoing a long path, telling the whole story is not the purpose of this introductory message. However, I can simply summarize them in the following:

As fa as design is considered, I have been concentrated on design and verification of ICs and IP cores for wireless communications (modulation and demodulation) during 2018-2022, particularly for satellite broadband Internet.

As far as authoring is concerned, I have been authoring a huge volume of books containing all aspects of the verification processes - from methodologies to languages, from platforms to testing systems. My first book in this direction was related to verification languages with Python, and afterwards SystemC, VHDL, (System)Verilog and a bunch of other verification languages.；then come the verification methodologies - UVM, UVVM, OSVVM, OVM etc.

I used also to write the gigantic book series titled “Silicon IP – Not just Design”.

Other books I have drafted include the following:

Protecting Your IP Cores
Review of Verification IP & IP Core Verification – An Abstract
Verification Methodologies - A Concise Introduction
Comprehensive Review of Hardware Verification Languages (Except Python)
Hardware Verification Planning - A Concise Introduction
Hardware Verification Planning Tools
Hardware Verification in Python

So far as web development is considered, I used to put huge efforts in developing a web portal about IP core development, but now this site is reoriented to be my personal site with my information about hardware verification, see www.ipcoredesign.net for reference.

Considering my earlier researches, apart from my researches many years ago, which you may find in my CV or my personal site, I can cite the one that is the mathematics for development and design of quantum computer (particularly Artificial Intelligence (Machine Learning and others) & heuristics). I started from qubit routing with a comprehensive review of that area together with an indepth study of a specific topic, one no one has tried sofar. I have initiated the writing of the following articles but stopped 2 or 3 years ago:

1) Math (AI & Heuristics) for Qubit Routing – A Survey
2) Qubit Routing with Machine Learning (Reinforcement Learning etc)

This research and writing was just the start of my huge research and writing plan in the quantum area. In fact, I have done a number of researches on quantum computer before, notably on graph theoretical applications for quantum circuits design, in line with the D-Wave quantum computer developed by a Canadian company a couple of years ago. I don't know what is its status now, because since 2016 my quantum research is stopped.

#####################################
Some Background
#####################################

My IC career has been undergoing a long path, telling the whole story is not the purpose of this introductory message. However, I can simply summarize them in the following:

I started designing ICs since 2013.

At first,I have been concentrated on design and verification of ICs and IP cores for wireless communications (modulation and demodulation) during 2018-2022, particularly for satellite broadband Internet.

On the other hand, I have been involving in AI well almost for the same time span as my IC career, since 2013, when I was studying the graph-theoretical aspects of IC (integrated circuits) design, where AI is also key to developing EDA algorithms for VLSI design. A couple of years ago my IC verification activities have also been involving lots of AI/ML techniques and algorithms, as well as my quantum circuits design for example qubit routing which employs deep reinforcement learning to improve the routing efficiency.

My current design focuses on hardware verification, particularly verification with UVM, OSVVM, UVVM, cocotb, ABV (SVA, PSL).

My current research activity in this area is the application of Artificial Intelligence and Machine Learning to hardware verification with my current focus on Neural Networks.

################################################
Additioonal: Information about my Career and Writing
################################################

CV attached:
==============================
MarkChen_ResumeEN2025Academic_short.pdf
http://1596825.xyz/MarkChen_ResumeEN2025Academic_short.pdf

AllWritingStudies.html:
==============================
Layout views of collections and file listing of books, reports, documents, essays, notes, records I have prepared during my researches, studies, business activities and other day to day activities over the period starting from 2003. Incomplete, but capable of estimating the scale of my writing strength. Overall tens of thousands of files are produced with millions of pages and hundreds of millions of words, mostly in English. Dare say few can be as productive as me.
http://1596825.xyz/0AllWritingStudies1.html

WhatsWritten2003to2017.pdf:
==============================
Academic, technological, engineering, business and routine writing during 2003-2017.
http://1596825.xyz/WhatsWritten2003to2017.pdf

writing2018uptonow.pdf:
==============================
Academic, technological, engineering, business and routine writing during sine 2018.
http://1596825.xyz/writing2018uptonow.pdf

Websites see:
WebsitesOnlineDrives.pdf
==============================
Websites I have developed since 2000 for various purposes, now not in operations except a few.
http://1596825.xyz/WebsitesOnlineDrives.pdf

Angelia Technologies / IPCore Design, Hardware Verification with Artificial Intelligence
https://www.ipcoredesign.net

Plutus Business Consultants (PBC) - Artificial Intelligence for Health, my current business focus
https://www.ai4healthcare.uk/

Github
https://github.com/worldsoft998

基于人工智能（AI）和机器学习（ML）的硬件验证（RTL）

我目前的设计重点是将人工智能和机器学习（人工神经网络、强化学习等）应用于硬件 RTL 验证，然后再进行网表（物理）验证。

基于人工智能/机器学习的验证技术如今已成为验证工程师关注的重点领域之一。我的目标是“硬件RTL验证优化”，即通过以下方式改进验证流程并缩短测试时间。
优化输入生成以最大限度地缩短测试时间
优化输入生成以最大程度地实现全覆盖
&
加速覆盖率收敛（加速覆盖率导向测试生成）

使用流行的AI/ML算法，例如人工神经网络和强化学习。

Python 将成为我的主要人工智能/机器学习语言，因为它在人工智能/机器学习应用领域占据主导地位。

我最喜欢的一些用于人工智能/机器学习的Python库包括：

TensorFlow
PyTorch
Scikit-learn
Keras
NumPy
OpenAI Gymnasium
Stable Baselines3

Verilog/ SystemVerilog /VHDL 是我的主要 HDL/HVL 语言，因为它们是业界和研究界最流行的硬件描述和验证语言。

UVM 是我的首选验证方法，但其他方法如 OSVVM、UVVM、 cocotb 、ABV（SVA、PSL）等也可以作为替代方案。

我们将采用多种 Python 与硬件描述语言 (HDL)/硬件验证语言 (HVL) 之间的通信方案来处理它们之间的交互，我将这些方案称为 Python-HDL 桥接器， 详情将专文讨论。

可以为您的测试平台开发定制的 AI/ML 模块，这些测试平台可以使用上述验证方法学和语言开发，也可以使用其他一些 HLS（高级综合）语言，例如SystemC、Python等。

我的最终目标是开发全面、通用且功能强大的AI/ML系统，该系统能够针对各种RTL测试用例，对不同的AI/ML算法进行对比测试，并与其他非AI算法以及通用的验证方案进行对比。如果可能，这些系统将扩展到芯片设计和验证的其他阶段，甚至扩展到医疗和机器人系统等非芯片应用领域。这些系统或模块可以采用多种形式，例如源代码、带或不带图形用户界面的独立应用程序、在线平台等等。此外，我们还将基于这些技术和系统提供专业的AI/ML验证加速和改进服务。

################################
早期研究、设计和开发
################################

我从2013年开始设计芯片。

自2013年以来，我一直积极参与人工智能/机器学习领域的研究。当时我正在研究芯片（IC）设计的图论方面，人工智能也是开发用于超大规模集成电路（VLSI）设计的EDA算法的关键。在早期，我的芯片设计工作也涉及了许多人工智能/机器学习技术和算法，例如，我的量子电路设计，如量子比特路由，就采用了深度强化学习来提高路由效率。

然而，我的IC职业生涯是一条漫长的道路，这篇介绍性信息的目的并非讲述全部故事。不过，我可以简单地概括如下：

就设计而言，2018-2022 年间，我一直专注于无线通信（调制和解调）芯片和 IP 核的设计和验证，特别是卫星宽带互联网方面的设计。

就写作而言，我撰写了大量书籍，涵盖验证过程的各个方面——从方法论到编程语言，从平台到测试系统。我的第一本书是关于Python验证语言的，之后又介绍了SystemC 、VHDL、(System)Verilog以及其他一些验证语言；接下来是验证方法论——UVM、UVVM、OSVVM、OVM等等。

我还写过一套名为“硅IP——不仅仅是设计”的巨著系列。

我撰写的其他书籍包括以下几本：

保护您的IP核心
验证IP和IP核验证综述——摘要
验证方法论——简要介绍
硬件验证语言（Python 除外）全面评述
硬件验证规划——简要介绍
硬件验证规划工具
Python硬件验证

就网站开发而言，我曾经投入大量精力开发一个关于 IP 核开发的门户网站，但现在这个网站已经转型为我的个人网站，用于发布我关于硬件验证的信息，详情请访问 www.ipcoredesign.net 。

考虑到我早期的研究，除了多年前的研究（您可以在我的简历或个人网站上找到这些研究）之外，我还可以重点介绍量子计算机开发和设计的数学（特别是人工智能（机器学习等）和逼近式算法）。我从量子比特路由入手，对该领域进行了全面的回顾，并深入研究了一个迄今为止无人涉足的特定主题。我曾着手撰写以下文章，但在两三年前停止了：

1) 量子比特路由的数学（人工智能与启发式算法）——综述
2) 基于机器学习（强化学习等）的量子比特路由

这项研究和写作仅仅是我庞大的量子领域研究和写作计划的开端。事实上，我之前已经做过一些关于量子计算机的研究，特别是关于图论在量子电路设计中的应用，这与几年前一家加拿大公司开发的D-Wave量子计算机的研究方向一致。我不知道它现在的进展如何，因为自2016年以来，我的量子研究就停止了。

################################
我的芯片学习与研究方面的一些背景
################################

我的 IC 职业生涯经历了漫长的道路，讲述整个故事并不是这篇介绍的目的。但是，我可以简单地总结如下：

我从 2013 年开始设计 IC。

起初，我在 2018-2022 年期间专注于无线通信（调制和解调）的 IC 和 IP 核的设计和验证，特别是卫星宽带互联网。当时我的卫星宽带互联网业务有了一点起色，谈了几个上规模的项目，但没有成功，所以该业务就很快销声匿迹。

另一方面，自 2013 年以来，几乎与从事 IC 研究设计的同时，一直在研究 AI，当时我正在研究 IC（集成电路）后端（物理）设计（尤其是布线）的图论基础，其中 AI 也是开发 VLSI 设计的 EDA 算法的关键。几年前，我的 IC 验证活动也涉及许多 AI/ML 技术和算法，以及我的量子电路设计，例如量子比特路由(qubit routing)，它采用深度强化学习(deep reinforcement learning)来提高路由效率。

我目前的设计侧重于硬件验证，特别是使用 UVM、OSVVM、UVVM、cocotb、ABV（SVA、PSL）进行验证。

我目前在该领域的研究活动是将人工智能和机器学习应用于芯片设计验证，目前重点是神经网络。

#####################################
附加信息：关于我的职业生涯和写作的信息
#####################################

简历附上：
==============================
MarkChen_ResumeEN2025Academic_short.pdf
http://1596825.xyz/MarkChen_ResumeEN2025Academic_short.pdf

AllWritingStudies.html：
==============================
这是自2003年以来，我在研究、学习、商务活动及其他日常活动中整理的著述。论文、报告、文档、文章、笔记和记录的合集及文件列表。虽然并不完整，但足以体现我的写作能力。总计产生了数万份文件，数百万页，数亿字，大部分为英文。可以说，很少有人能像我一样高产。
http://1596825.xyz/0AllWritingStudies1.html

WhatsWritten2003to2017.pdf：
==============================
2003-2017 年期间的学术、技术、工程、商业和日常写作。
http://1596825.xyz/WhatsWritten2003to2017.pdf

writing2018uptonow.pdf：
==============================
2018年以来从事学术、技术、工程、商业和日常写作。
http://1596825.xyz/writing2018uptonow.pdf

网站显示：
WebsitesOnlineDrives.pdf
==============================
我自 2000 年以来开发过各种用途的网站，现在除了少数几个之外，都已停止运营。
http://1596825.xyz/WebsitesOnlineDrives.pdf

Angelia Technologies / IPCore设计、硬件验证与人工智能
https://www.ipcoredesign.net

Plutus Business Consultants (PBC)——人工智能在医疗健康领域的应用，这是我目前专注于的业务领域。
https://www.ai4healthcare.uk/

我的代码托管网站
Github
https://github.com/worldsoft998

#####################################
简介
#####################################

我叫陈明华，同济大学1989年硕士，及1991-1992年间获得几个国外名校（法国巴黎大学，普罗旺斯大学，英国格拉斯哥大学，德国ESSEN大学,加拿大Laval大学等）博士资格。

我从1983年开始工作，最初的工作是位于北京三里河的机械工业部总部，接着在各种形态的公司企业工作，从事多种类型的工作，简单列出如下（按最近工作先列出，时间短的忽略）：

1997 - 1999  德国巴高克公司北京办事处  全球采购经理，
1995 - 1997  法国阿尔斯通公司上海办事处 （上海）电力/造船/其它行业，市场开拓，电力工程师，
1994 - 1996  德国莱茵技术（上海）有限公司 TÜV Rheinland  工程师，质量体系认证兼市场营销
1992 - 1995  西门子-海曼光电有限（深圳）公司 （深圳）光电业，质量保证，部门经理，
1990 – 1992  广东核电合营有限公司 （深圳大亚湾）核电业，质量保证，工程师，
1989 – 1990  广东省电力设计研究院 （广州）电力业，设计，工程师，
1983 - 1986  机械工业部电器工业局电站处/中国电工设备总公司 （北京）电力业，行业管理，助理工程师，

本人会英语，德语，法语，俄语等。

本人也熟悉许多IT技术。