基于python调用matlab，进而调用simulink模型并建立强化学习仿真环境项目源码.zip

# ------------------------------------------------------------------ # FileName: CoC2O4withoutT7.py # Author: 宁文彬 # Github: http://github.com/simuler # Version: V 1.0 # Description: 草酸钴合成过程gym环境搭建 # ------------------------------------------------------------------ import matlab.engine import gym from gym import spaces from gym.utils import seeding import numpy as np import os import sys class CoC2O4withoutT7Env(gym.Env): def __init__(self): self.action_space = spaces.Discrete(2) high = np.array([1e20, 1e20, 1e20, 1e20, 1e20, 1e20], dtype=np.float64) self.observation_space = spaces.Box(-high, high, dtype=np.float64) self.seed() self.state = None self.mean = np.array([2, 2, 5.6e16, 7.8e10, 1.2e5, 0.2]) self.std = np.array([5, 0.5, 1.3e16, 1.6e10, 3.5e4, 0.07]) self.time = 0 self.done = False self.eng = matlab.engine.start_matlab() self.env_name = 'CoC2O4withoutT7' self.eng.load_system(self.env_name) def seed(self, seed=None): self.np_random, seed = seeding.np_random(seed) return [seed] def step(self, action): self.time += 60 state = self.eng.CoC2O4_step(float(action), self.time, nargout=1) state = np.array(state) state = np.squeeze(state) if self.time == 660: self.done = True reward = state[3] / state[2] * 1e6 if np.isnan(reward): reward = 0 else: reward = 0 state = (state - self.mean) / self.std return np.array(state, dtype=np.float64), reward, self.done, {} def reset(self): self.state = [0.0, 1.4, 0.0, 0.0, 0.0, 0.0] self.eng.reset_file(nargout=0) # 运行初始化的m文件 self.eng.CoC2O4_step(float(1e-6), 1, nargout=1) self.time = 0 self.done = False return np.array(self.state, dtype=np.float64)