def step(self, agent, action):
"""
2A step for an agent
"""
# make sure RL agents cannot learn in battle mode.
agent.alpha = 0.0
reward = NERO.NeroEnvironment.NeroEnvironment.step(self, agent, action)
live_agents = sum(len(t) for t in self.teams.itervalues())
# wait until all agents have spawned before we keep track of damage.
if self.print_damage < 0 and live_agents < 2 * constants.pop_size:
return reward
# print out damages every loop through the entire population.
self.print_damage += 1
if self.print_damage >= live_agents:
self.print_damage = 0
else:
return reward
damages = {}
winner = None
for team in constants.TEAMS:
damages[team] = 0
agents = self.teams.get(team, set())
agent_damages = [self.get_state(a).total_damage for a in agents]
if any(d < self.hitpoints for d in agent_damages):
damage = sum(agent_damages)
damage += (constants.pop_size - len(agents)) * self.hitpoints
damages[team] = damage
else:
damages[team] = constants.pop_size * self.hitpoints
winner = constants.OBJECT_TYPE_TEAM_0
if team == winner:
winner = constants.OBJECT_TYPE_TEAM_1
if len(damages) == 2:
ss = []
for t, d in sorted(damages.iteritems()):
s = '%s: %d' % (constants.TEAM_LABELS[t], d)
ss.append(s)
ss = tuple(ss)
if self.last_damage != ss:
print 'damages sustained by: ' + ' '.join(ss)
self.last_damage = ss
if not constants.getDisplayHint():
OpenNero.setWindowCaption('Damage sustained: ' +
' '.join(ss))
if winner is not None:
s = constants.TEAM_LABELS[winner] + ' team wins!!!'
print s
try:
if OpenNero.getAppConfig().rendertype != 'null':
import tkMessageBox
tkMessageBox.showinfo('NERO Battle Results', s)
except:
pass
if not constants.getDisplayHint():
OpenNero.setWindowCaption(s)
OpenNero.disable_ai()
for a in self.teams[winner]:
self.set_animation(a, self.get_state(a), 'jump')
return reward
def showDisplayHint():
hint = constants.nextDisplayHint()
if hint is None:
OpenNero.setWindowCaption("")
else:
OpenNero.setWindowCaption("Displaying: " + hint)
def showDisplayHint():
hint = constants.nextDisplayHint()
if hint is None:
OpenNero.setWindowCaption("")
else:
OpenNero.setWindowCaption("Displaying: " + hint)
def step(self, agent, action):
"""
2A step for an agent
"""
# make sure RL agents cannot learn in battle mode.
agent.alpha = 0.0
reward = NERO.NeroEnvironment.NeroEnvironment.step(self, agent, action)
live_agents = sum(len(t) for t in self.teams.itervalues())
# wait until all agents have spawned before we keep track of damage.
if self.print_damage < 0 and live_agents < 2 * constants.pop_size:
return reward
# print out damages every loop through the entire population.
self.print_damage += 1
if self.print_damage >= live_agents:
self.print_damage = 0
else:
return reward
damages = {}
winner = None
for team in constants.TEAMS:
damages[team] = 0
agents = self.teams.get(team, set())
agent_damages = [self.get_state(a).total_damage for a in agents]
if any(d < self.hitpoints for d in agent_damages):
damage = sum(agent_damages)
damage += (constants.pop_size - len(agents)) * self.hitpoints
damages[team] = damage
else:
damages[team] = constants.pop_size * self.hitpoints
winner = constants.OBJECT_TYPE_TEAM_0
if team == winner:
winner = constants.OBJECT_TYPE_TEAM_1
if len(damages) == 2:
ss = []
for t, d in sorted(damages.iteritems()):
s = '%s: %d' % (constants.TEAM_LABELS[t], d)
ss.append(s)
ss = tuple(ss)
if self.last_damage != ss:
print 'damages sustained by: ' +' '.join(ss)
self.last_damage = ss
if not constants.getDisplayHint():
OpenNero.setWindowCaption('Damage sustained: ' + ' '.join(ss))
if winner is not None:
s = constants.TEAM_LABELS[winner] + ' team wins!!!'
print s
try:
if OpenNero.getAppConfig().rendertype != 'null':
import tkMessageBox
tkMessageBox.showinfo('NERO Battle Results', s)
except:
pass
if not constants.getDisplayHint():
OpenNero.setWindowCaption(s)
OpenNero.disable_ai()
for a in self.teams[winner]:
self.set_animation(a, self.get_state(a), 'jump')
return reward
