'batch_size': 32,
'state_space_dim': env.state_dim,
'action_space_dim': env.action_dim,
}
agent = Agent(**params)
score = []
mean = []
for episode in range(1000):
s0 = env.reset()
total_reward = 1
while True:
env.render()
a0 = agent.act(s0)
s1, r1, done= env.step(a0)
if done:
r1 = -1
agent.put(s0, a0, r1, s1)
if done:
break
total_reward += r1
s0 = s1
agent.learn()
score.append(total_reward)
mean.append( sum(score[-100:])/100)
###############################################
import os
import sys
import numpy as np
import pandas as pd
import matplotlib.pyplot as plt
from env import ArmEnv
from rl import DDPG
MAX_EPISODES = 500
MAX_EP_STEPS = 200
env = ArmEnv()
s_dim = env.state_dim
a_dim = env.action_dim
a_bound = env.action_bound
rl = DDPG(a_dim, s_dim, a_bound)
for i in range(MAX_EPISODES):
s = evn.reset()
for j in range(MAX_EP_STEPS):
env.render()
a = rl.choose_actions(s)
s_, r, done = env.step(a)
rl.store_transition(s, a, r, s_)
if rl.memory_full():
rl.learn()
s = s_
# Gloabel Variable
MAX_EPISOSES = 500
MAX_EP_STEPS = 500
# Set the environement
env = ArmEnv()
s_dim = env.state_dim
a_dim = env.action_dim
a_bound = env.action_bound
# set the RL method
rl = DDPG(a_dim, s_dim, a_bound)
# start Training
for i in range(MAX_EPISOSES):
s = env.reset()
for j in range(MAX_EP_STEPS):
env.render()
a = rl.choose_action(s)
s_, r, done = env.step(a)
rl.store_transitions(s, a, r, s_)
if rl.memory_full:
# start to learn once has fulfulled the memory
rl.learn()
s = s_