# Tonwork — Guofang Tang (Tim)'s Engineering Blog

> Engineering depth for autonomous driving and intelligent systems: BEV perception, VLA, sensor fusion, C++ toolchains, and open-source contributions.

Written by Guofang Tang (Tim), this blog connects model internals with the realities of building and deploying software. Its strengths include source-level analysis, explicit tensor-shape reasoning, runnable C++ examples, and grounded discussion of deployment trade-offs.

Articles are primarily written in Simplified Chinese. This English guide helps readers and agents navigate the original content; Markdown pages preserve the original language and include canonical article URLs, publication dates, and update dates where available. Interpret performance figures and product claims within their cited sources, versions, dates, and measurement conditions.

## Author and Site

- [Guofang Tang (Tim)](https://tonwork.fun/#person): Author profile, engineering interests, and contact information.
- [Topic Guides](https://tonwork.fun/topics/): Curated reading paths for BEV perception, autonomous driving and VLA, and C++ tooling.
- [Open-Source Projects and Contributions](https://tonwork.fun/project/): Public repositories, merged pull requests, and GitHub contribution activity, including contributions to other maintainers' projects.

## BEV Perception Engineering

- [BEV Perception Engineering Guide](https://tonwork.fun/topics/bev/): A guided path from training configuration and vectorized mapping to deployment latency and sensor fusion.
- [SimpleBEV Training Notes](https://tonwork.fun/blog/simple_bev%E5%AD%A6%E4%B9%A0/index.md): Read the training entry point, camera configuration, and gradient accumulation settings behind a BEV baseline.
- [MapTR Hierarchical Queries: From Source Code to Tensor Shapes](https://tonwork.fun/blog/transformermaptr-%E6%A0%B8%E5%BF%83%E6%80%9D%E6%83%B3%E4%B8%8E%E6%B5%81%E7%A8%8B%E6%80%BB%E7%BB%93/index.md): Trace instance-point broadcasting, the 512-dimensional embedding split, deformable attention, and map outputs against a pinned upstream implementation.
- [Deploying BEV Perception: Model Latency versus System Real-Time Performance](https://tonwork.fun/blog/%E9%BB%91%E7%AE%B1%E6%8F%AD%E7%A7%98-%E8%B7%91%E6%BB%A130fps%E7%9A%84%E9%87%8F%E4%BA%A7bev%E6%84%9F%E7%9F%A5%E7%B3%BB%E7%BB%9F/index.md): Analyze BEVPoolv2, TensorRT, quantization, and pipeline trade-offs with clear boundaries between operator speed, model FPS, and end-to-end latency.
- [Sensor Fusion and Perception Optimization for Autonomous Driving](https://tonwork.fun/blog/%E6%8B%A8%E4%BA%91%E8%A7%81%E6%97%A5%E6%B4%9E%E5%AF%9F%E7%A7%8B%E6%AF%AB-%E6%99%BA%E8%83%BD%E9%A9%BE%E9%A9%B6%E6%84%9F%E7%9F%A5%E7%B3%BB%E7%BB%9F%E7%9A%84%E8%9E%8D%E5%90%88%E4%B8%8E/index.md): Connect early and late fusion, sensor calibration, BEV representations, detection, and tracking in a vehicle perception pipeline.

## Autonomous Driving and VLA

- [Autonomous Driving and VLA Guide](https://tonwork.fun/topics/autonomous-driving-vla/): Evaluate vision-language-action models in the context of driving data, product availability, vehicle architecture, and safety responsibilities.
- [Can VLA Models Work in Autonomous Driving?](https://tonwork.fun/blog/%E8%87%AA%E5%8A%A8%E9%A9%BE%E9%A9%B6%E7%9A%84%E8%AE%A4%E7%9F%A5%E9%87%8D%E6%9E%84-vla-%E6%9E%B6%E6%9E%84%E7%9A%84%E5%B7%A5%E7%A8%8B%E7%9C%9F%E7%9B%B8/index.md): Examine open-model availability, driving data requirements, and the engineering questions that stand between VLA research and vehicle deployment.
- [Tesla FSD in China and XPeng VLA: What Is Actually Available?](https://tonwork.fun/blog/tesla-fsd-china-xpeng-vla2/index.md): Distinguish official availability, reported developments, and competing technical approaches; product status is tied to the article date.
- [How Autonomous-Driving Hardware and Software Architectures Evolve](https://tonwork.fun/blog/%E9%80%A0%E8%BD%A6%E9%AA%A8%E6%9E%B6%E5%B8%88-%E6%88%91%E7%9A%84%E6%99%BA%E8%83%BD%E9%A9%BE%E9%A9%B6%E6%BC%94%E8%BF%9B%E5%BC%8F%E6%9E%B6%E6%9E%84%E5%93%B2%E5%AD%A6/index.md): An engineering perspective on SoC selection, middleware, model adaptation, and interface contracts across project stages.
- [The MCU Role in Autonomous-Driving Control and Safety Supervision](https://tonwork.fun/blog/%E5%AE%89%E5%85%A8%E5%BA%95%E7%BA%BF-mcu/index.md): Understand control interfaces, gateways, and health monitoring alongside high-performance perception compute; concrete safety requirements belong at vehicle-system level.

## C++ and Developer Toolchains

- [C++ and Developer Toolchains Guide](https://tonwork.fun/topics/cpp-toolchain/): Connect object lifetime, compilation databases, build systems, and editor diagnostics through concrete examples.
- [C++ weak_ptr: Observation, Shared Ownership, and Reference Cycles](https://tonwork.fun/blog/c-%E6%99%BA%E8%83%BD%E6%8C%87%E9%92%88-weak_ptr-%E5%85%A5%E9%97%A8%E6%8C%87%E5%8D%97/index.md): Use runnable C++17 examples to see exactly when lock() extends object lifetime and how weak links break shared_ptr ownership cycles.
- [compile_commands.json: Fields, Compiler Flags, and Generation](https://tonwork.fun/blog/compile_commands-json-%E7%AE%80%E6%98%93%E7%A7%91%E6%99%AE%E4%B8%8E%E7%BC%96%E8%AF%91%E9%80%89%E9%A1%B9%E8%A7%A3%E9%87%8A/index.md): Understand arguments and directory semantics, CMake generation, and the compilation database checks that make editor diagnostics match the build.
- [clangd and VS Code: Minimal Setup and Practical Troubleshooting](https://tonwork.fun/blog/%E4%BB%A3%E7%A0%81%E8%87%AA%E5%8A%A8%E6%8F%90%E7%A4%BA-clangd-vscode%E5%85%A5%E9%97%A8/index.md): Start with real compiler flags, then diagnose missing headers, incorrect macros, and fallback commands using clangd --check.
- [Generating a Compilation Database for Bazel C++ Projects](https://tonwork.fun/blog/bazel-%E7%94%9F%E6%88%90-compile_commands-json-%E7%AE%80%E6%98%93%E7%89%88%E6%95%99%E7%A8%8B/index.md): Follow the Hedron extractor integration path while checking dependency and configuration compatibility with the project's Bazel version.
- [C++ std::atomic Notes](https://tonwork.fun/blog/c-atomic/index.md): Explore atomic reads, writes, and common concurrent operations as a companion to object lifetime and ownership management.