Honor of Kings Arena

Welcome to documentation for the AI Arena Competition and its related repositories and components!

What is AI Arena Competition

AI Arena Competition, also known as Kaiwu (“开悟” in Chinese) Competition, is jointly organized by Tencent AI Lab and the Honor of Kings team. In this competition, participants develop an agent to master the game of Honor of Kings with sparse rewards, long-term decision making and cooperative ability. Honor of Kings is one of the most popular mobile Multiplayer Online Battle Arena (MOBA) games with hundreds of millions of players, whose complexity provides an ideal testing ground for game AI. In short AI Arena Competition consists of several major components:

  • Honor of Kings Environment – The test game environments authorized by the Honor of Kings, one of the world’s most popular game, includes a set of environments which highlight many of the hardest problems in modern-day Reinforcement Learning: long time horizons, partially-observable state and Multi-agent coordination and competition.

    • hok – A rich python3 package for doing artificial intelligence research in Honor of Kings. This includes two major submodules.

      • HoK1v1 – A growing set of OpenAI Gym environments in Honor of Kings. These environments leverage a synchronous, stable, and fast integration of Honor of Kings.

      • interface.so – The main interface module for interacting with the Honor of Kings game engine.

  • SAIL Framework – The supporting reinforcement learning framework which enables asynchronous training strategies and multi-machine multi-card training.

  • Kaiwu Platform – The task management platform where the competitors can submit their training tasks to the platforms through the front end of the web page.

Questions and Feedback

Got any questions?


HoK package API Reference

Indices and tables