def main():
env = utils.setupEnv('MarLo-TrickyArena-v0')
# Get the number of available actions, minus waiting action
actionSize = env.action_space.n
epsilonDecay = 0.98
alphas = [0.8, 0.5, 0.1]
gammas = [1, 0.5]
for alpha in alphas:
for gamma in gammas:
QTableName = "QTable_Alpha_" + str(alpha).replace(
".", "_") + "_Gamma_" + str(gamma).replace(
".", "_") + "_Decay_" + str(epsilonDecay).replace(
".", "_") + ".json"
CSVName = "Results_Alpha_" + str(alpha).replace(
".", "_") + "_Gamma_" + str(gamma).replace(
".", "_") + "_Decay_" + str(epsilonDecay).replace(
".", "_") + ".csv"
myAgent = QLearningAgent(actionSize, 200, QTableName, CSVName,
False, epsilonDecay, alpha, gamma)
# Start the running of the Agent
myAgent.runAgent(env)
return
def main():
env = utils.setupEnv('MarLo-CliffWalking-v0')
# Get the number of available actions, minus waiting action
#actionSize = env.action_space.n
actionSize = 5
epsilonDecay = 0.98
#alphas = [0.8,0.5,0.1]
#gammas = [1,0.5]
alphas = [0.8]
gammas = [1]
i = 1
for alpha in alphas:
for gamma in gammas:
QTableName = "QTable_Alpha_" + str(alpha).replace(
".", "_") + "_Gamma_" + str(gamma).replace(
".", "_") + "_Decay_" + str(epsilonDecay).replace(
".", "_") + ".json"
CSVName = str(i) + "_Test_Results_Alpha_" + str(alpha).replace(
".", "_") + "_Gamma_" + str(gamma).replace(
".", "_") + "_Decay_" + str(epsilonDecay).replace(
".", "_") + ".csv"
myAgent = QLearningAgent(actionSize,
25,
QTableName,
CSVName,
True,
epsilonDecay,
alpha,
gamma,
0.00,
training=True)
print(
"\n\n -------------- Starting test run of Decay %s, Alpha %s and Gamma %s --------- \n \n"
% (epsilonDecay, alpha, gamma))
# Start the running of the Agent
myAgent.runAgent(env)
return
"""
Initialize differents Agents
SARSA
Expected SARSA
QLearning
"""
#print(f"EPSILON: {params.EPSILON}, LEARNING_RATE: {params.LEARNING_RATE}, DISCOUNT: {params.DISCOUNT}, DISCRETE_VALUE: {params.DISCRETE_VALUE}, env.observation_space.high: {params.env.observation_space.high},\
# env.action_space.n: {params.env.action_space.n}, env.action_space: {params.env.action_space}, \
# DISCRETE: {params.DISCRETE}")
qLearningAgent = QLearningAgent(params.EPSILON, params.LEARNING_RATE,
params.DISCOUNT, params.DISCRETE_VALUE,
params.env.observation_space.high,
params.env.action_space.n,
params.env.action_space, params.DISCRETE)
sarsaAgent = SarsaAgent(params.EPSILON, params.LEARNING_RATE, params.DISCOUNT,
params.DISCRETE_VALUE,
params.env.observation_space.high,
params.env.action_space.n, params.env.action_space,
params.DISCRETE)
expectedSarsaAgent = ExpectedSarsaAgent(params.EPSILON, params.LEARNING_RATE,
params.DISCOUNT, params.DISCRETE_VALUE,
params.env.observation_space.high,
params.env.action_space.n,
params.env.action_space,
params.DISCRETE)