CoSub/GenerateUtilityTable.py at master · hudlass/CoSub · GitHub

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
#  =============================================================================
#  ========================== GenerateUtilityTable.py ==========================
#  ============================ Luke Hudlass-Galley ============================
#  =============================================================================

#  Imports and dependencies ====================================================
import math, random, pprint, numpy as np
from pathlib import Path
from GenerateTables import *

#  Generate utility table  =====================================================
def GenerateUtilityTable(puzzleType, puzzleSize):

    # Load relevant files, if they do not exist then create them ===============
    if puzzleType == "SlidingTile":
        #  Check if the SSI and state-action tables exist, if not create them ==
        puzzleName = str(puzzleSize) + "Puzzle"
        saFile     = "GeneratedFiles/" + puzzleName + "StateActionTable.csv"

        if Path(saFile).is_file():
            stateActions = np.loadtxt(saFile, delimiter = ",", unpack = False)
        else:
            GenerateStateSpaceIndex("SlidingTile", puzzleSize)
            GenerateStateActionTable("SlidingTile", puzzleSize)
            stateActions = np.loadtxt(saFile, delimiter = ",", unpack = False)

        maxIterations  = 2000
        discountFactor = 1
        goalReward     = 100
        punishment     = 1
        learningRate   = 0.5

        entireUtilityTable = []

        #  Define utility and rewards table ====================================
        stateSpaceSize  = puzzleSize * (puzzleSize + 1)
        tableDimensions = np.array([stateSpaceSize, 4])

        for goalStateIndex in range(stateSpaceSize):
            print(goalStateIndex)
            utilityTable    = np.zeros(tableDimensions)
            utilityTable[goalStateIndex, :] = goalReward
            #  Generate rewards table ==========================================
            rewardsTable    = np.zeros(tableDimensions) - punishment
            rewardsTable[goalStateIndex, :] = goalReward

            for counter in range(maxIterations):
                currentStateIndex = counter % stateSpaceSize
                while currentStateIndex != goalStateIndex:
                    randomAction = random.randint(0, 3)
                    rewardValue  = rewardsTable[currentStateIndex, randomAction]
                    maxUtility   = utilityTable[int(stateActions[currentStateIndex, randomAction]),:].max()
                    newUtility   = (1 - learningRate) * utilityTable[currentStateIndex, randomAction]
                    newUtility  += learningRate * (rewardValue + discountFactor * maxUtility)
                    utilityTable[currentStateIndex, randomAction] = newUtility
                    currentStateIndex = int(stateActions[currentStateIndex, randomAction])
            entireUtilityTable.extend(utilityTable)

        fileName = "GeneratedFiles/" + str(puzzleSize) + 'PuzzleUtilityTable.csv'
        np.savetxt(fileName, entireUtilityTable, delimiter = ',')

    if puzzleType == "Twisty":

        maxIterations  = 2000
        discountFactor = 1
        goalReward     = 100
        punishment     = 1
        learningRate   = 0.5

        puzzleName = str(puzzleSize) + "x" + str(puzzleSize) + "x" + str(puzzleSize)

        fileNameSSIEdges   = "GeneratedFiles/" + puzzleName + "EdgesSSI.csv"
        fileNameSSICorners = "GeneratedFiles/" + puzzleName + "CornersSSI.csv"

        fileNameStateActionsEdges   = "GeneratedFiles/" + puzzleName + "StateActionsEdges.csv"
        fileNameStateActionsCorners = "GeneratedFiles/" + puzzleName + "StateActionsCorners.csv"

        if Path(fileNameSSIEdges).is_file():
            stateSpaceIndexEdges = np.loadtxt(fileNameSSIEdges, delimiter=",", unpack = False)
        else:
            GenerateStateActionTable(puzzleType, puzzleSize)
            stateSpaceIndexEdges = np.loadtxt(fileNameSSIEdges, delimiter=",", unpack = False)

        if Path(fileNameSSICorners).is_file():
            stateSpaceIndexCorners = np.loadtxt(fileNameSSICorners, delimiter=",", unpack = False)
        else:
            GenerateStateActionTable(puzzleType, puzzleSize)
            stateSpaceIndexCorners = np.loadtxt(fileNameSSICorners, delimiter=",", unpack = False)

        if Path(fileNameStateActionsEdges).is_file():
            stateActionsEdges = np.loadtxt(fileNameStateActionsEdges, delimiter=",", unpack = False)
        else:
            GenerateStateActionTable(puzzleType, puzzleSize)
            stateActionsEdges = np.loadtxt(fileNameStateActionsEdges, delimiter=",", unpack = False)

        if Path(fileNameStateActionsCorners).is_file():
            stateActionsCorners = np.loadtxt(fileNameStateActionsCorners, delimiter=",", unpack = False)
        else:
            GenerateStateActionTable(puzzleType, puzzleSize)
            stateActionsCorners = np.loadtxt(fileNameStateActionsCorners, delimiter=",", unpack = False)

        # Q-Learning ===================================================================================================
        entireUtilityTableEdges   = []
        entireUtilityTableCorners = []
        numberOfChildStates  = 24
        numberOfActions      = 18
        tableDimensions      = np.array([numberOfChildStates, numberOfActions])

        # Generate the utilities for each possible goal state ==========================================================
        for goalStateIndex in range(numberOfChildStates):
            #print(goalStateIndex)

            # Utility tables ===========================================================================================
            utilityTableEdges   = np.zeros(tableDimensions)
            utilityTableCorners = np.zeros(tableDimensions)
            utilityTableEdges[goalStateIndex, :]   = goalReward
            utilityTableCorners[goalStateIndex, :] = goalReward

            # Rewards table ============================================================================================
            rewardsTable                    = np.zeros(tableDimensions) - punishment
            rewardsTable[goalStateIndex, :] = goalReward

            for edgeCounter in range(maxIterations):
                print("Iteration number: " + str(edgeCounter))
                currentStateIndex = (edgeCounter) % numberOfChildStates
                while currentStateIndex != goalStateIndex:
                    randomAction = random.randint(0, numberOfActions - 1)
                    rewardValue  = rewardsTable[currentStateIndex, randomAction]

                    maxUtility   = utilityTableEdges[int(stateActionsEdges[currentStateIndex, randomAction]) - 1, :].max()
                    newUtility   = (1 - learningRate) * utilityTableEdges[currentStateIndex, randomAction]
                    newUtility  += learningRate * (rewardValue + discountFactor * maxUtility)
                    utilityTableEdges[currentStateIndex, randomAction] = newUtility
                    currentStateIndex = int(stateActionsEdges[currentStateIndex, randomAction]) - 1
            print("Edge goal index " + str(goalStateIndex) + "reached with all starting states indexes")
            entireUtilityTableEdges.extend(np.asarray(utilityTableEdges))

            for cornerCounter in range(maxIterations):
                print("Iteration number: " + str(cornerCounter))
                currentStateIndex = (cornerCounter) % numberOfChildStates
                while currentStateIndex != goalStateIndex:
                    randomAction = random.randint(0, numberOfActions - 1)
                    rewardValue  = rewardsTable[currentStateIndex, randomAction]

                    maxUtility   = utilityTableCorners[int(stateActionsCorners[currentStateIndex, randomAction]) - 1, :].max()
                    newUtility   = (1 - learningRate) * utilityTableCorners[currentStateIndex, randomAction]
                    newUtility  += learningRate * (rewardValue + discountFactor * maxUtility)
                    utilityTableCorners[currentStateIndex, randomAction] = newUtility
                    currentStateIndex = int(stateActionsCorners[currentStateIndex, randomAction]) - 1
            print("Corner goal index " + str(goalStateIndex) + "reached with all starting states indexes")
            entireUtilityTableCorners.extend(utilityTableCorners)


        # Save utility tables to .csv file =============================================================================
        puzzleName = str(puzzleSize) + "x" + str(puzzleSize) + "x" + str(puzzleSize)

        fileNameUtilityTableEdges   = "GeneratedFiles/" + puzzleName + "UtilityTableEdges.csv"
        fileNameUtilityTableCorners = "GeneratedFiles/" + puzzleName + "UtilityTableCorners.csv"
        np.savetxt(fileNameUtilityTableEdges, entireUtilityTableEdges, delimiter = ',')
        np.savetxt(fileNameUtilityTableCorners, entireUtilityTableCorners, delimiter = ',')