def reset(self, env=None):
self._dissociate()
if env:
self.set_env(env=env)
self._action = None
self._proposal_offer = None
self._current_offer_offer = None
if self.init_proposal:
self._ip = Issue.sample(issues=[Issue(self._ip_range)],
n_outcomes=1,
astype=tuple)[0]
self._rp = Issue.sample(issues=[Issue(self._rp_range)],
n_outcomes=1,
astype=tuple)[0]
if "ip" in self.ufun.__dict__:
self.ufun.ip = self._ip
self.set_proposal_offer(self._ip)
if "rp" in self.ufun.__dict__:
self.ufun.rp = self._rp
self.end_time = 1
def test_drl_negotiator():
from drl_negotiation.negotiator import MyDRLNegotiator, DRLNegotiator
from drl_negotiation.utility_functions import MyUtilityFunction
from negmas import SAOMechanism, Issue, ResponseType, Outcome
from drl_negotiation.env import DRLNegotiationEnv, NEnv
name = "drl_negotiator"
issues = [
Issue(values=10, name="quantity"),
Issue(values=100, name="delivery_time"),
Issue(values=100, name="unit_price")
]
# buyer
weights = (0, -0.5, -0.8)
# set the utility function with MyUtilityFunction
mechanism = SAOMechanism(issues=issues, n_steps=100)
ufun = MyUtilityFunction(weights=weights, )
drl_negotiator = MyDRLNegotiator(name=name, ufun=ufun)
mechanism.add(drl_negotiator)
drl_negotiator_two = MyDRLNegotiator(name=name + "_two")
mechanism.add(drl_negotiator_two,
ufun=MyUtilityFunction(weights=(0, -0.2, -0.6)))
assert drl_negotiator.name == name
assert drl_negotiator.ufun == ufun
# Test the basic attributes and functions
# None
assert drl_negotiator.env is None
assert drl_negotiator.action is None
#TODO: initial value and reserved value
# Test the respond and propose of negotiator
offer = Issue.sample(issues, 1)
# Test acceptance strategy
# Action: None, Env: None, Train: True
# env is None, so the respond is always ResponseType.REJECT_OFFER
drl_negotiator.train = True
respond = drl_negotiator.respond(mechanism.state, offer[0])
assert type(respond) == ResponseType
if respond == ResponseType.REJECT_OFFER:
proposal_offer = drl_negotiator.propose(mechanism.state)
if proposal_offer is not None:
assert proposal_offer in drl_negotiator.ami.outcomes
else:
assert False, "Reponse must be ResponseType.Reject_OFFER!"
# Action: None, Env: not None, Train: True
# Env will automatically set the issues, games and so on attributes.
drl_negotiator.train = True
name = "drl_negotiation_env"
n_env = DRLNegotiationEnv(name=name)
drl_negotiator.reset(env=n_env)
assert isinstance(drl_negotiator.env, NEnv)
respond = drl_negotiator.respond(mechanism.state, offer[0])
assert type(
respond) == ResponseType, "Type of response must be ResponseType!"
# Action: not None, Env: not None
drl_negotiator.reset(env=n_env)
action = drl_negotiator.env.action_space.sample()
drl_negotiator.set_current_action(action)
assert type(drl_negotiator.action) == ResponseType
assert type(
respond) == ResponseType, "Type of response must be ResponseType!"
respond = drl_negotiator.respond(mechanism.state, offer[0])
assert respond == drl_negotiator.action
