深度增强学习射击类游戏(vizdoom)

前端时间搞一个airsim的学习，通过ddqn的方式，然后这两天在尝试一些增强学习的训练；

在Papers With Code上看到如下游戏

FPS Games | Papers With Code

GitHub - mwydmuch/ViZDoom: Doom-based AI Research Platform for Reinforcement Learning from Raw Visual Information.

然后在增强学习页面reinforcement-learning | Papers With Code看到项目 DLR-RM/stable-baselines3,

搭建环境如下：

首先要安装pytorch，1.8版本以上

然后安装stable-baselines3

pip install stable-baselines3

然后安装游戏

sudo apt install cmake libboost-all-dev libsdl2-dev libfreetype6-dev libgl1-mesa-dev libglu1-mesa-dev libpng-dev libjpeg-dev libbz2-dev libfluidsynth-dev libgme-dev libopenal-dev zlib1g-dev timidity tar nasm
pip install vizdoom

当然安装过程肯定有很多依赖，相关依赖可从上面链接中看到

然后开始训练

import gym
from vizdoom import gym_wrapper

from stable_baselines3 import PPO

# env = gym.make("CartPole-v1")
env = gym.make("VizdoomHealthGatheringSupreme-v0")

model = PPO("MultiInputPolicy", env, verbose=1)
model.learn(total_timesteps=10_000)

obs = env.reset()
for i in range(10000):
    action, _states = model.predict(obs, deterministic=True)
    obs, reward, done, info = env.step(action)
    env.render()
    if done:
      obs = env.reset()

env.close()

啊——，到现在位置，看起来没有语法错误，至于是否可以训练出预期结果，尚不可知(我在虚拟机里调试的)