Python OpenNero.set_ai方法代码示例

# 需要导入模块: import OpenNero [as 别名]
# 或者: from OpenNero import set_ai [as 别名]
 def load_rtneat(self, location , pop, team=constants.OBJECT_TYPE_TEAM_0):
     location = os.path.relpath("/") + location
     if os.path.exists(location):
         OpenNero.set_ai("rtneat-%s" % team, OpenNero.RTNEAT(
                 str(location), "data/ai/neat-params.dat",
                 constants.pop_size,
                 OpenNero.get_environment().agent_info.reward))

# 需要导入模块: import OpenNero [as 别名]
# 或者: from OpenNero import set_ai [as 别名]
 def deploy(self, ai='rtneat', team=constants.OBJECT_TYPE_TEAM_0):
     OpenNero.disable_ai()
     if ai == 'rtneat':
         OpenNero.set_ai('rtneat-%s' % team, None)
     self.environment.remove_all_agents(team)
     for _ in range(constants.pop_size):
         self.spawnAgent(ai=ai, team=team)
     OpenNero.enable_ai()

# 需要导入模块: import OpenNero [as 别名]
# 或者: from OpenNero import set_ai [as 别名]
    def __init__(self):
        """
        Create the environment
        """
        OpenNero.Environment.__init__(self)

        self.curr_id = 0
        self.max_steps = 20
        self.MAX_DIST = math.hypot(constants.XDIM, constants.YDIM)
        self.states = {}
        self.teams = {}
        self.script = 'Hw5/menu.py'

        abound = OpenNero.FeatureVectorInfo() # actions
        sbound = OpenNero.FeatureVectorInfo() # sensors
        rbound = OpenNero.FeatureVectorInfo() # rewards

        # actions
        abound.add_continuous(-1, 1) # forward/backward speed (gets multiplied by constants.MAX_MOVEMENT_SPEED)
        abound.add_continuous(-constants.MAX_TURN_RADIANS, constants.MAX_TURN_RADIANS) # left/right turn (in radians)

        # sensor dimensions
        for a in range(constants.N_SENSORS):
            sbound.add_continuous(0, 1);

        # Rewards
        # the enviroment returns the raw multiple dimensions of the fitness as
        # they get each step. This then gets combined into, e.g. Z-score, by
        # the ScoreHelper in order to calculate the final rtNEAT-fitness
        for f in constants.FITNESS_DIMENSIONS:
            # we don't care about the bounds of the individual dimensions
            rbound.add_continuous(-sys.float_info.max, sys.float_info.max) # range for reward

        # initialize the rtNEAT algorithm parameters
        # input layer has enough nodes for all the observations plus a bias
        # output layer has enough values for all the actions
        # population size matches ours
        # 1.0 is the weight initialization noise
        rtneat = OpenNero.RTNEAT("data/ai/neat-params.dat",
                                 constants.N_SENSORS,
                                 constants.N_ACTIONS,
                                 constants.pop_size,
                                 1.0,
                                 rbound, False)

        key = "rtneat-%s" % constants.OBJECT_TYPE_TEAM_0
        OpenNero.set_ai(key, rtneat)
        print "get_ai(%s): %s" % (key, OpenNero.get_ai(key))

        # set the initial lifetime
        lifetime = module.getMod().lt
        rtneat.set_lifetime(lifetime)
        print 'rtNEAT lifetime:', lifetime

        self.agent_info = OpenNero.AgentInitInfo(sbound, abound, rbound)

# 需要导入模块: import OpenNero [as 别名]
# 或者: from OpenNero import set_ai [as 别名]
 def start_rtneat(self, pop_size):
     " start the rtneat learning demo "
     OpenNero.disable_ai()
     #self.environment = RoombaEnvironment(constants.XDIM, constants.YDIM, self)
     #set_environment(self.environment)
     #self.reset_sandbox()
     # Create RTNEAT object
     rbound = OpenNero.FeatureVectorInfo()
     rbound.add_continuous(-sys.float_info.max, sys.float_info.max)
     rtneat = OpenNero.RTNEAT("data/ai/neat-params.dat", 2, 1, pop_size, 1.0, rbound, False)
     rtneat.set_weight(0,1)
     OpenNero.set_ai("rtneat",rtneat)
     OpenNero.enable_ai()
     self.distribute_bots(pop_size, "data/shapes/roomba/RoombaRTNEAT.xml")

# 需要导入模块: import OpenNero [as 别名]
# 或者: from OpenNero import set_ai [as 别名]
    def start_rtneatq(self, team=constants.OBJECT_TYPE_TEAM_0):
        # initialize the rtNEAT+Q algorithm parameters
        # input layer has enough nodes for all the observations plus a bias
        # output layer has enough values for all the wires
        # population size matches ours
        # 1.0 is the weight initialization noise
        rtneatq = OpenNero.RTNEAT("data/ai/neat-params.dat",
                                 constants.N_SENSORS+1,
                                 constants.N_ACTION_CANDIDATES * (constants.N_ACTIONS + 1),
                                 constants.pop_size,
                                 1.0,
                                 rtneat_rewards(), 
                                 False)

        key = "rtneatq-%s" % team
        OpenNero.set_ai(key, rtneatq)
        print "get_ai(%s): %s" % (key, OpenNero.get_ai(key))

        rtneatq.set_lifetime(self.environment.lifetime)