def agent_init(self,taskSpec): self.episode = 0 self.steps = 1 print "TaskSpec:" print taskSpec print self.action = Action() self.action_types = [] self.action.intArray = [] self.action.doubleArray = [] self.action.numInts = 1 rnd.seed(0) TaskSpec = TaskSpecParser(taskSpec) self.input_ranges = TaskSpec.getDoubleObservations() print 'observations ranges',self.input_ranges #parse action self.action_ranges = TaskSpec.getIntActions()[0] print "action ranges",self.action_ranges self.actionlist = range(self.action_ranges[0],self.action_ranges[1]+1) self.nactions = len(self.actionlist) print "action list",self.actionlist #build a nearest neighbor function approximator self.Q = NNQ(nactions=self.nactions,input_ranges=self.input_ranges,nelemns=[7,7,3,3],alpha=0.5,lm=0.8) self.SelectAction = e_greedy_selection(epsilon=0.1) #set eplison for e-greedy exploration self.SelectAction.parent = self # discount factor self.gamma = 1.0#TaskSpec.getDiscountFactor() self.proving+= 1
def agent_init(self, taskSpec):
self.episode = 0
self.steps = 1
print "TaskSpec:"
print taskSpec
print
self.action = Action()
self.action_types = []
self.action.intArray = []
self.action.doubleArray = []
self.action.numInts = 1
rnd.seed(0)
TaskSpec = TaskSpecParser(taskSpec)
self.input_ranges = TaskSpec.getDoubleObservations()
print 'observations ranges', self.input_ranges
#parse action
self.action_ranges = TaskSpec.getIntActions()[0]
print "action ranges", self.action_ranges
self.actionlist = range(self.action_ranges[0],
self.action_ranges[1] + 1)
self.nactions = len(self.actionlist)
print "action list", self.actionlist
#build a nearest neighbor function approximator
self.Q = NNQ(nactions=self.nactions,
input_ranges=self.input_ranges,
nelemns=[7, 7, 3, 3],
alpha=0.5,
lm=0.8)
self.SelectAction = e_greedy_selection(
epsilon=0.1) #set eplison for e-greedy exploration
self.SelectAction.parent = self
# discount factor
self.gamma = 1.0 #TaskSpec.getDiscountFactor()
self.proving += 1
