def __init__(self):
# print("Initialising")
m= "Test,10,10,4,0,4,9,1111211111100000000111100110011000000001100001011111011000111000000001100101010110000000011111311111"
#"Test,10,10,4,0,4,9,1111211111100000000111100110011000000001100001011111011000111000000001100101010110000000011111311111"
#"Test,6,6,4,5,1,0,131111100051105001100051150001111121" #"Test,10,10,4,0,4,9,1111211151100000000111100110011000000001100001511111011000111000000001100501050110000000011111311111"
#"Test,4,4,2,0,2,3,1125100110011531" MazeGenerator()
self.maze= Maze(m)
self.s=Simulator(self.maze)
self.span=6
self.number=2
self.pList=[]
self.stateList=[]
self.finishedP=[]
self.history=m+"|"+"0"
self.finished=0
self.count=0
for j in range(self.number):
p=Worker(self.maze)
self.s.add(p)
self.pList.append(p)
state= np.asarray(p.getView(p.getPos(),self.span))
self.stateList.append(state)
self.history+="#"+p.getName()+"-"+p.getPos().CordToString()
self.history+="|"
action_space=[]
for i in range(0,len(Action)):
action_space.append(i)
self.action_space_worker=np.asarray(action_space)
self.observation_space_worker= math.pow(2*self.span+1,2)
self.shortestRoute=len(self.maze.GetOptimalRoute()[0])
self.maze.printMaze()
self.dqn_solver_worker = DQNSolver(int(self.observation_space_worker), len(self.action_space_worker))
self.reward=0
def __init__(self):
# print("Initialising")
m = "Test,6,6,4,5,1,0,131111100051105001100051150001111121" #"Test,10,10,4,0,4,9,1111211151100000000111100110011000000001100001511111011000111000000001100501050110000000011111311111"
#"Test,4,4,2,0,2,3,1125100110011531" MazeGenerator()
self.maze = Maze(m)
self.s = Simulator(self.maze)
self.span = 6
self.number = 2
self.pList = []
self.stateList = []
self.finishedP = []
self.history = m + "|" + "0"
self.finished = 0
for j in range(self.number):
p = Worker(self.maze)
self.s.add(p)
self.pList.append(p)
state = np.asarray(p.getView(p.getPos(), self.span))
self.stateList.append(state)
self.history += "#" + p.getName() + "-" + p.getPos().CordToString()
self.history += "|"
action_space = []
for i in range(0, len(Action)):
action_space.append(i)
import itertools
possible_actions = [action_space] * self.number
print(action_space)
self.action_space = np.asarray(
list(itertools.product(*possible_actions)))
#print(self.action_space)
#self.action_space=np.asarray(action_space)
self.observation_space = math.pow(2 * self.span + 1, 2) * self.number
self.shortestRoute = len(self.maze.GetOptimalRoute()[0])
self.maze.printMaze()
def resetNewMaze(self):
m = MazeGenerator()
self.maze = Maze(m)
self.s = Simulator(self.maze)
self.pList = []
self.stateList = []
self.history = m + "|" + "0"
self.finished = 0
for j in range(self.number):
p = Worker(self.maze)
self.pList.append(p)
self.s.add(p)
state = np.asarray(p.getView(p.getPos(), self.span))
self.stateList.append(state)
self.history += "#" + p.getName() + "-" + p.getPos().CordToString()
self.history += "|"
self.shortestRoute = len(self.maze.GetOptimalRoute()[0])
self.maze.printMaze()
return self.stateList
def resetNewMaze(self):
mazesizeh = self.maze.height + 10
mazesizew = self.maze.width + 10
if (self.maze.height == 5):
mazesizeh = 10
mazesizew = 10
m = MazeGenerator(mazesizeh, mazesizew)
#"Test,10,10,4,0,4,9,1111211111100000000111100110011000000001100001011111011000111000000001100101010110000000011111311111" #MazeGenerator()
self.maze = Maze(m)
self.s = Simulator(self.maze)
self.pList = []
self.qList = []
self.mList = []
self.pPos = []
self.qPos = []
self.mPos = []
self.pStateList = []
self.qStateList = []
self.mStateList = []
self.history = m + "|" + "0"
self.finished = 0
self.finishedP = []
self.finishedQ = []
self.count = 0
self.eaten = 0
self.queenEaten = False
self.queenLeft = False
#rewards
self.wReward_not_possible = -50
self.wReward_wall = -50
self.wReward_entrance = -20
self.wReward_finished_before = 0
self.wReward_exit = self.maze.height * self.maze.width * 20
self.wReward_towards_exit = -1
self.wReward_toQueen = -1
self.wReward_atQueen = self.maze.height * self.maze.width * 20
self.wReward_repeat_pos = -20
self.wReward_else = -3
self.qReward_not_possible = -50
self.qReward_wall = -50
self.qReward_entrance = -20
self.qReward_finished_before = 0
self.qReward_exit = self.maze.height * self.maze.width * 20
self.qReward_towards_exit = -1
self.qReward_repeat_pos = -20
self.qReward_else = -3
self.sReward_not_possible = -50
self.sReward_wall = -50
self.sReward_eat = 1000
self.sReward_eatQueen = 5000
self.sReward_towards_prey = -1
self.sReward_repeat_pos = -20
self.sReward_else = -3
#for book-keeping
#self.wNumber=self.wNumber+1
self.folderName = "Span_" + str(self.spanP) + "Dim_" + str(
self.maze.height
) + "_" + str(
self.maze.width
) #m+"_maze_"+str(self.wNumber)+"_workers_"+str(self.spanP)+"_span_"+str(self.wReward_not_possible)+"_not poss_"+str(self.wReward_wall)+"_wall_"+str(self.wReward_entrance)+"_ent_"+str(self.wReward_finished_before) +"_finished already_"+str(self.wReward_exit)+"_exit_"+str(self.wReward_towards_exit)+"_to exit_"+str(self.wReward_repeat_pos) +"_rep pos_"+str(self.wReward_else)+"_else "
self.maxIter = 10 * self.maze.height * self.maze.width
self.completeStop = False
if (self.wReward_exit == 0):
self.completeStop = True
for j in range(self.wNumber):
p = Worker(self.maze, self.spanP)
self.s.add(p)
self.pList.append(p)
self.pPos.append(p.getPos())
for j in range(self.qNumber):
q = Queen(self.maze)
self.s.add(q)
self.qList.append(q)
self.qPos.append(q.getPos())
#self.history+="|"
action_space = []
for j in range(self.sNumber):
s = Spider(self.maze)
self.s.add(s)
self.mList.append(s)
self.mPos.append(s.getPos())
for j in range(self.wNumber):
state = np.asarray(
p.getAugView(p.getPos(), self.spanP, self.pPos, self.qPos,
self.mPos))
self.pStateList.append(state)
self.history += "#" + p.getName() + "-" + p.getPos().CordToString()
for j in range(self.qNumber):
#state= np.asarray(q.getView())
#self.qStateList.append(state)
self.history += "#" + q.getName() + "-" + q.getPos().CordToString()
for j in range(self.sNumber):
state = np.asarray(
s.getAugView(s.getPos(), self.spanM, self.pPos, self.qPos,
self.mPos))
self.mStateList.append(state)
self.history += "#" + s.getName() + "-" + s.getPos().CordToString(
) #fix writeup
self.history += "|"
self.shortestRoute = len(self.maze.GetOptimalRoute()[0])
self.maze.printMaze()
self.shortestRoute = len(self.maze.GetOptimalRoute()[0])
self.dqn_solver_worker = DQNSolver(int(self.observation_space_worker),
len(self.action_space_worker))
self.dqn_solver_queen = DQNSolver(int(self.observation_space_queen),
len(self.action_space_queen))
self.dqn_solver_spider = DQNSolver(int(self.observation_space_spider),
len(self.action_space_spider))
self.pReward = 0
self.qReward = 0
self.mReward = 0
return self.pStateList
def reset(self):
#print("Resetting")
self.maze = Maze(self.maze.mazeString)
self.pList = []
self.qList = []
self.mList = []
self.pPos = []
self.qPos = []
self.mPos = []
self.pStateList = []
self.mStateList = []
self.qStateList = []
self.finishedP = []
self.finishedQ = []
self.history = self.maze.mazeString + "|" + "0"
self.finished = 0
self.finishedP = []
self.count = 0
self.queenLeft = False
self.maxIter = 10 * self.maze.height * self.maze.width
for j in range(self.wNumber):
p = Worker(self.maze, self.spanP)
self.pList.append(p)
for k in range(self.qNumber):
q = Queen(self.maze)
self.qList.append(q)
for h in range(self.sNumber):
s = Spider(self.maze)
self.mList.append(s)
#print(len(self.pList),len(self.qList),len(self.mList))
self.eaten = 0
self.queenEaten = False
for p in self.pList:
p.setInitPos(Cord(self.maze.getInitialX(),
self.maze.getInitialY()))
self.pPos.append(p.getPos())
for q in self.qList:
q.setInitPos(Cord(self.maze.getInitialX(),
self.maze.getInitialY()))
self.qPos.append(q.getPos())
for q in self.mList:
q.setInitPos(Cord(q.start.X, q.start.Y))
self.mPos.append(q.getPos())
for p in self.pList:
state = np.asarray(
p.getAugView(p.getPos(), self.spanP, self.pPos, self.qPos,
self.mPos))
self.pStateList.append(state)
self.history += "#" + p.getName() + "-" + p.getPos().CordToString()
for q in self.qList:
#state=np.asarray(p.getAugView(p.getPos(),self.spanP,self.pPos, self.qPos, self.mPos))
#self.pStateList.append(state)
self.history += "#" + q.getName() + "-" + q.getPos().CordToString()
for q in self.mList:
state = np.asarray(
q.getAugView(q.getPos(), self.spanM, self.pPos, self.qPos,
self.mPos))
self.mStateList.append(state)
self.history += "#" + q.getName() + "-" + q.getPos().CordToString()
self.history += "|"
self.shortestRoute = len(self.maze.GetOptimalRoute()[0])
self.pReward = 0
self.qReward = 0
self.mReward = 0
#print(self.maze.mazeString)
#self.maze.printMaze()
return self.pStateList
def __init__(self):
# print("Initialising")
m = MazeGenerator(
10, 10
) #"Test,10,10,4,0,4,9,1111211111100000000111100110011000000001100001011111011000111000000001100101010110000000011111311111"
#"Test,10,10,4,0,4,9,1111211111100000000111100110011000000001100001011111011000111000000001100101010110000000011111311111"
#"Test,6,6,4,5,1,0,131111100051105001100051150001111121" #"Test,10,10,4,0,4,9,1111211151100000000111100110011000000001100001511111011000111000000001100501050110000000011111311111"
#"Test,4,4,2,0,2,3,1125100110011531" MazeGenerator()
self.maze = Maze(m)
self.s = Simulator(self.maze)
self.spanP = 5
self.spanM = 5
self.wNumber = 5
self.qNumber = 0
self.sNumber = 0
self.pList = []
self.qList = []
self.mList = []
self.pPos = []
self.qPos = []
self.mPos = []
self.pStateList = []
self.qStateList = []
self.mStateList = []
self.finishedP = []
self.finishedQ = []
self.history = m + "|" + "0"
self.finished = 0
self.count = 0
self.eaten = 0
self.queenEaten = False
self.queenLeft = False
#rewards
self.wReward_not_possible = -50
self.wReward_wall = -50
self.wReward_entrance = -20
self.wReward_finished_before = 0
self.wReward_exit = self.maze.height * self.maze.width * 20
self.wReward_towards_exit = -1
self.wReward_toQueen = -1
self.wReward_atQueen = self.maze.height * self.maze.width * 20
self.wReward_repeat_pos = -20
self.wReward_else = -3
self.wReward_queenEaten = -5000
self.qReward_not_possible = -50
self.qReward_wall = -50
self.qReward_entrance = -20
self.qReward_finished_before = 0
self.qReward_exit = self.maze.height * self.maze.width * 20
self.qReward_towards_exit = -1
self.qReward_repeat_pos = -20
self.qReward_else = -3
self.sReward_not_possible = -50
self.sReward_wall = -50
self.sReward_eat = 1000
self.sReward_eatQueen = 5000
self.sReward_towards_prey = -1
self.sReward_repeat_pos = -20
self.sReward_else = -3
#for book-keeping/need to update with spiders
self.folderName = "Span_" + str(self.spanP) + "Dim_" + str(
self.maze.height) + "_" + str(self.maze.width)
#m+"_maze_"+str(self.wNumber)+"_workers_"+str(self.spanP)+"_span_"+str(self.wReward_not_possible)+"_not poss_"+str(self.wReward_wall)+"_wall_"+str(self.wReward_entrance)+"_ent_"+str(self.wReward_finished_before) +"_finished already_"+str(self.wReward_exit)+"_exit_"+str(self.wReward_towards_exit)+"_to exit_"+str(self.wReward_repeat_pos) +"_rep pos_"+str(self.wReward_else)+"_else " ####
self.maxIter = 10 * self.maze.height * self.maze.width
self.completeStop = False
for j in range(self.wNumber):
p = Worker(self.maze, self.spanP)
self.s.add(p)
self.pList.append(p)
self.pPos.append(p.getPos())
for j in range(self.qNumber):
q = Queen(self.maze)
self.s.add(q)
self.qList.append(q)
self.qPos.append(q.getPos())
#self.history+="|"
action_space = []
for j in range(self.sNumber):
s = Spider(self.maze)
self.s.add(s)
self.mList.append(s)
self.mPos.append(s.getPos())
for j in range(self.wNumber):
state = np.asarray(
p.getAugView(p.getPos(), self.spanP, self.pPos, self.qPos,
self.mPos))
self.pStateList.append(state)
self.history += "#" + p.getName() + "-" + p.getPos().CordToString()
for j in range(self.qNumber):
#state= np.asarray(q.getView())
#self.qStateList.append(state)
self.history += "#" + q.getName() + "-" + q.getPos().CordToString()
for j in range(self.sNumber):
state = np.asarray(
s.getAugView(s.getPos(), self.spanM, self.pPos, self.qPos,
self.mPos))
self.mStateList.append(state)
self.history += "#" + s.getName() + "-" + s.getPos().CordToString(
) #fix writeup
self.history += "|"
action_space = []
for i in range(0, len(Action)):
action_space.append(i)
self.action_space_worker = np.asarray(action_space)
self.observation_space_worker = math.pow(2 * self.spanP + 1, 2)
self.action_space_queen = np.asarray(action_space)
self.observation_space_queen = (self.maze.height + 2 * self.spanP) * (
self.maze.width + 2 * self.spanP) #math.pow(2*self.spanP+1,2)
self.action_space_spider = np.asarray(action_space)
self.observation_space_spider = math.pow(2 * self.spanM + 1, 2)
self.shortestRoute = len(self.maze.GetOptimalRoute()[0])
self.maze.printMaze()
self.dqn_solver_worker = DQNSolver(int(self.observation_space_worker),
len(self.action_space_worker))
self.dqn_solver_queen = DQNSolver(int(self.observation_space_queen),
len(self.action_space_queen))
self.dqn_solver_spider = DQNSolver(int(self.observation_space_spider),
len(self.action_space_spider))
self.pReward = 0
self.qReward = 0
self.mReward = 0