from Env.CustomEnv.StablizerOneD import StablizerOneD
import random
import matplotlib.pyplot as plt
env = StablizerOneD()
state = env.reset()
print(state)
stateSet = []
stateSet.append(state)
for i in range(1000):
action = random.randint(0, env.nbActions - 1)
state, reward, done, _ = env.step(action)
#print('step ' + str(i))
#print(state)
#print(reward)
#print(done)
stateSet.append(state)
if done:
break
print(stateSet)
fig = plt.figure(2)
ax = plt.axes()
env.render_traj(stateSet, ax)
plt.show()
config['epsThreshold'] = 0.1
config['targetNetUpdateStep'] = 100
config['memoryCapacity'] = 200
config['trainBatchSize'] = 32
config['gamma'] = 0.9
config['learningRate'] = 0.001
config['netGradClip'] = 1
config['logFlag'] = True
config['logFileName'] = 'StabilizerOneDLog/traj'
config['logFrequency'] = 100
config['priorityMemoryOption'] = False
config['netUpdateOption'] = 'doubleQ'
config['netUpdateFrequency'] = 1
config['priorityMemory_absErrUpper'] = 5
env = StablizerOneD()
N_S = env.stateDim
N_A = env.nbActions
netParameter = dict()
netParameter['n_feature'] = N_S
netParameter['n_hidden'] = [100]
netParameter['n_output'] = N_A
policyNet = MultiLayerNetRegression(netParameter['n_feature'],
netParameter['n_hidden'],
netParameter['n_output'])
targetNet = deepcopy(policyNet)
optimizer = optim.Adam(policyNet.parameters(), lr=config['learningRate'])
def _thunk():
env = StablizerOneD(config, i)
return env