一个自行设计的坦克大站的游戏客户端，可从中学习到Agent多智能体的理论，并从游戏中学习编程

瓦片世界，类似于推箱子游戏，一个智能体自动将箱子推到制定地点并且计算步数，附带地图编辑代码，可以自行设定地图-Tile world, similar to Sokoban Game, an agent will automatically push the boxes to the development of sites and the calculation step, with the map editor code, you can set up their own map

使用深度强化学习中的ddpg算法学习玩游戏，让智能体学习最优策略。(The ddpg algorithm in deep reinforcement learning is used to learn to play games, so that the agent can learn the optimal strategy.)