Research notes · engineering essays

Zihao Han

Research notes on autonomy, machine learning, and technical systems.

这里收录了我关于自动飞行系统、生成式模型与底层工具链的长期技术笔记。它并非传统的“开发者博客”,而是作为一份持续演进的研究与工程工作台存在。

Mode

Long-form notes

Method

Read · test · document

Archive

7 indexed notes

Focus areas

Recurring threads in my research and engineering work.

The archive currently clusters around three domains: flight-control systems, modern generative models, and the tooling required to make technical work reproducible.

Focus

Autonomous Flight & Systems

Fixed-wing workflows, ground-station operation, simulation environments, and the practical steps required to make an airframe actually fly under software control.

PX4MAVSDKRobotics

Focus

Generative Models & Theory

Careful breakdowns of diffusion models, architecture intuition, and the sort of technical notes that sit between paper reading and real implementation.

DDPMMultimodalArchitecture

Focus

Systems, Tooling & Infrastructure

Linux repair logs, remote development setups, and the small-but-important infrastructure work that keeps larger projects moving.

LinuxCloudTooling

Publications

Scholar-facing work, integrated into the site.

Instead of leaving papers as an external profile only, the site now keeps a publication archive, cached Scholar metrics, and room for explanatory blog notes on first-author work.

Selected writing

Entry points into the archive.

Rather than surfacing every post chronologically, these notes best represent the core themes of the collection.

publications

AsyncDSB:从同步调度到像素异步恢复

重新梳理 AsyncDSB 的问题意识、方法结构与真正的贡献边界:它并不是简单“加一个异步策略”,而是在 image inpainting 的 diffusion Schrödinger bridge 里重写了调度与恢复过程的对应关系。

paper notediffusion modelsimage inpainting

Archive

Recent writing

Every note remains accessible in a cleaner archive view designed for long-form reading rather than blog-style browsing.

publications

AsyncDSB:从同步调度到像素异步恢复

重新梳理 AsyncDSB 的问题意识、方法结构与真正的贡献边界:它并不是简单“加一个异步策略”,而是在 image inpainting 的 diffusion Schrödinger bridge 里重写了调度与恢复过程的对应关系。

paper notediffusion modelsimage inpaintingresearch
南工御风

QA01

围绕固定翼比赛任务,汇总自主飞行、目标识别与学习路径的答疑与资料入口。

px4ubuntumavsdkyolo
南工御风

PX4仿真环境的搭建

通过 Docker 与原生 Linux 两种路径,梳理 PX4 仿真环境的搭建流程与注意事项。

px4ubuntudocker