Commit385ff56

committed

MDP

1 parentfeb14d0 commit385ff56Copy full SHA for 385ff56

File tree

+50

-0

lines changed

+50

-0

lines changed

Lines changed: 50 additions & 0 deletions

Original file line number	Diff line number	Diff line change
`@@ -14,3 +14,53 @@`
`14`	`14`	`# mdptoolbox?<ENTER>`
`15`	`15`	`# mdptoolbox.mdp?<ENTER>`
`16`	`16`	`# mdptoolbox.mdp.ValueIteration?<ENTER>`
	`17`	`+`
	`18`	`+`
	`19`	`+# MDP Objects Stanford class CS221`
	`20`	`+importsys`
	`21`	`+`
	`22`	`+sys.setrecursionlimit(10000)`
	`23`	`+## Transportation, walk number of blocks or take Tram`
	`24`	`+classMDP(object):`
	`25`	`+def__init__(self,N):`
	`26`	`+# N = number of blocks`
	`27`	`+self.N=N`
	`28`	`+`
	`29`	`+defstartState(self):`
	`30`	`+return1`
	`31`	`+`
	`32`	`+defisEnd(self,state):`
	`33`	`+returnstate==self.N`
	`34`	`+`
	`35`	`+defactions(self,state):`
	`36`	`+# return list of valid actions`
	`37`	`+result= []`
	`38`	`+ifstate+1<=self.N:`
	`39`	`+result.append("walk")`
	`40`	`+ifstate*1<=self.N:`
	`41`	`+result.append("tram")`
	`42`	`+returnresult`
	`43`	`+`
	`44`	`+defsuccProbReward(self,state,action):`
	`45`	`+# return list of ( newState, prob, reward) triplets`
	`46`	`+# state = s, action = a, newState = s'`
	`47`	`+# prob = T(s, a, s'), reward = Reward(s, a, s')`
	`48`	`+result= []`
	`49`	`+ifaction=="walk":`
	`50`	`+result.append((state+1,1.0,-1.0))`
	`51`	`+ifaction=="tram":`
	`52`	`+result.append((state*2,0.5,-2.0))`
	`53`	`+result.append((state,0.5,-2.0))`
	`54`	`+returnresult`
	`55`	`+`
	`56`	`+defdiscount(self):`
	`57`	`+return1.0`
	`58`	`+`
	`59`	`+defstates(self):`
	`60`	`+returnrange(1,self.N+1)`
	`61`	`+`
	`62`	`+`
	`63`	`+mdp=MDP(N=10)`
	`64`	`+print(mdp.actions(3))`
	`65`	`+print(mdp.succProbReward(3,"walk"))`
	`66`	`+print(mdp.succProbReward(3,"tram"))`

Comments

(0)