Hi, I’m
Penguin
Notes from learning graphics, game development and game design: the books I read, the code I write, the things that finally click.
What I’m working on
Reforge
A post-apocalyptic top-down survival game built with UE5.6, with core mechanics inspired by Project Zomboid and Deep Rock Galactic.
Changjiang Journey
Changjiang Journey follows the culture of the Yangtze River as its narrative thread. Each stage explores one NPC’s cultural journey through a distinct minigame.
FPS Demo
An early-to-mid-development first-person shooter demo built with UE5.8, focused on recreating the interaction feel of Rainbow Six Siege, Delta Force, and PUBG.
How this blog is built
A personal site where the blog comes first and the wiki is a module — how notes become pages, how books are read online, and how shaders run inside posts.
AI 感知：守卫看见了什么，又以为看见了什么
距离、视野锥和视线遮挡决定 NPC 能不能发现你；发现以后写进记忆的是「信念」，不是真相。拖着鼠标当玩家，亲手调一只守卫的眼睛，再看它怎么挑一个能看见你的位置。
The library
17 sources — books, excerpts and courses, loaded page by page.
Recently written
All writing →- LessonsAI 感知：守卫看见了什么，又以为看见了什么距离、视野锥和视线遮挡决定 NPC 能不能发现你；发现以后写进记忆的是「信念」，不是真相。拖着鼠标当玩家，亲手调一只守卫的眼睛，再看它怎么挑一个能看见你的位置。
- Lessons游戏 AI 导论：感知、决策、执行游戏 AI 不是新闻里的生成式 AI，它的目标是让 NPC「看起来」理性。先把整套 AI 拆成感知、决策、执行几层，弄清每层管什么、为什么要拆开，再看这条课程线该怎么读。
- Lessons行为树：失败了，还有 Plan B用 Sequence、Selector 和三种返回值搭一棵树，让 NPC 门打不开就去走窗。逐拍跟着一次 tick 走下来，再亲手写出 Sequence 和 Selector 的核心代码。
- Lessons黑板：感知、决策、执行共用的一块记忆黑板是 AI 各模块共享的工作记忆：感知往上写，决策从上面读，任务回头更新。看它解决了什么问题，行为树的数据上下文、虚幻引擎的黑板资产和 Halo 2 的 Props 各自是什么样，以及推和拉两种模型怎么配合。
A self-compiled knowledge base
Books and articles I’ve read, compiled into interlinked notes — the key ones rewritten as interactive lessons. (Chinese only.)
Not sure where to start?
Every section has a reading order from basics to advanced — 6 roadmaps, 37 stages, each step jumping straight to its category. (Chinese only.)