episode_reward_plot(1) (1).py

# coding=utf-8

# DDPG TD3 SAC 训练曲线
import numpy as np
import sys
import matplotlib.pyplot as plt
import os

plt.rc('font',family='Times New Roman')
# matplotlib.rcParams['font.sans-serif']=['SimHei']   # 用黑体显示中文
# matplotlib.rcParams['axes.unicode_minus']=False     # 正常显示负号

plt.figure(figsize=(5,3))
plt.rcParams['savefig.dpi'] = 600 #图片像素
plt.rcParams['figure.dpi'] = 600 #分辨率

ax = plt.gca()

ax.tick_params(bottom=False,top=False,left=False,right=False)

# 设置网格颜色
ax.grid(color='grey', linestyle='-', linewidth=1, alpha=0.5, zorder=0)

# 地址
address = "./"

# 加载数据
def load(policy, num=3):
    url = os.path.dirname(os.path.realpath(__file__))
    result = []
    for i in range(num):
        # temp = np.load(address+policy+"/step_Q_values.npy")
        # temp = np.load(address+policy+"/step_Q_values.npy")
        temp = np.load(address+"/sum_reward.npy")
        result.append(temp)
    return result

def smooth(arr, fineness):
    result = arr[:]
    for i in range(fineness, arr.size):
        temp = 0
        for j in range(fineness):
            temp += result[i-j]
        result[i] = temp/fineness
    return np.array(result)

def get_mean_max_min(data_list, smooth_flag, fineness):
    n = sys.maxsize
    for data in data_list:
        n = min(n, data.size)
    max_data = np.zeros((n))
    min_data = np.zeros((n))
    mean_data = np.zeros((n))

    for i in range(n):
        temp = []
        for data in data_list:
            temp.append(data[i])
        temp = np.array(temp)
        max_data[i] = temp.max()
        min_data[i] = temp.min()
        mean_data[i] = temp.mean()

    data = [mean_data, max_data, min_data]
    if smooth_flag:
        for i in range(len(data)):
            for j in range(2, fineness):
                data[i] = smooth(data[i], j)
    return data[0], data[1], data[2]



SAC_data = load("DADDPG", 3)
TD3_data = load("original", 3)
DDPG_data = load("fd_replay", 3)

fineness = 20
fineness_pad = 15

SAC_mean_data, SAC_max_data, SAC_min_data = get_mean_max_min(SAC_data, True, fineness)
TD3_mean_data, TD3_max_data, TD3_min_data = get_mean_max_min(TD3_data, True, fineness)
DDPG_mean_data, DDPG_max_data, DDPG_min_data = get_mean_max_min(DDPG_data, True, fineness)

SAC_x = range(SAC_mean_data.size)
plt.fill_between(SAC_x, SAC_mean_data+fineness_pad, SAC_mean_data-fineness_pad, alpha=0.2, zorder=2, color="blue")
plt.plot(SAC_x, SAC_mean_data, linewidth=2, label="SAC", zorder=3, color="blue")

TD3_x = range(TD3_mean_data.size)
plt.fill_between(TD3_x, TD3_mean_data+fineness_pad, TD3_mean_data-fineness_pad, alpha=0.2, zorder=2, color="red")
plt.plot(TD3_x, TD3_mean_data, linewidth=2, label="TD3", zorder=3, color="red")

DDPG_x = range(DDPG_mean_data.size)
plt.fill_between(DDPG_x, DDPG_mean_data+fineness_pad, DDPG_mean_data-fineness_pad, alpha=0.2, zorder=4, color='green')
plt.plot(DDPG_x, DDPG_mean_data, linewidth=2, label="DDPG", zorder=3, color="green")

# plt.title("Training curve", pad=15)
plt.xlabel("Episode", labelpad=8.5)
plt.ylabel("Accumulated Reward", labelpad=8.5)

plt.tight_layout()

plt.legend(loc="upper right", frameon=True)

plt.savefig('episode_reward_plot.jpg')
plt.show()