def __call__(self):
if self.seed is not None:
np.random.seed(self.seed)
self.seed = self.seed + 1
xMin, xMax = self.xBoundary
yMin, yMax = self.yBoundary
initialAgentState = np.array(
[np.random.uniform(xMin, xMax),
np.random.uniform(yMin, yMax)])
targetPosition = np.array(
[np.random.uniform(xMin, xMax),
np.random.uniform(yMin, yMax)])
initialDistance = computeVectorNorm(targetPosition - initialAgentState)
while not (
(self.checkBound(initialAgentState) == initialAgentState).all() and
(self.checkBound(targetPosition) == targetPosition).all()
and initialDistance <= self.maxInitDist):
initialAgentState = np.array(
[np.random.uniform(xMin, xMax),
np.random.uniform(yMin, yMax)])
targetPosition = np.array(
[np.random.uniform(xMin, xMax),
np.random.uniform(yMin, yMax)])
initialDistance = computeVectorNorm(targetPosition -
initialAgentState)
return np.concatenate([initialAgentState, targetPosition])
def __call__(self):
point = dataTools.sampleData(self.dataSet, 1)
state, action, value = point
sheep, wolf = getEachState(state[0])
distance = computeVectorNorm(np.array(sheep) - np.array(wolf))
while not (distance >= self.minInitDist):
point = dataTools.sampleData(self.dataSet, 1)
state, action, value = point
sheep, wolf = getEachState(state[0])
distance = computeVectorNorm(np.array(sheep) - np.array(wolf))
return state[0]
def __call__(self, state):
agentState, targetPosition = getEachState(state)
relativeVector = np.array(agentState) - np.array(targetPosition)
relativeDistance = computeVectorNorm(relativeVector)
if relativeDistance <= self.minDistance:
return True
return False
def __call__(self, state):
sheepState, wolfState = getEachState(state)
relativeVector = np.array(sheepState) - np.array(wolfState)
relativeDistance = computeVectorNorm(relativeVector)
if relativeDistance <= self.minDistance:
return True
return False
def __call__(self, state, action):
oldSheepPos, oldWolfPos = getEachState(state)
# wolf
wolfAction = self.wolfPolicy(state)
wolfActionMagnitude = computeVectorNorm(np.array(wolfAction))
modifiedWolfAction = np.array(
wolfAction) * 0.95 * self.velocity / wolfActionMagnitude
newWolfPos = np.array(oldWolfPos) + modifiedWolfAction
# sheep
sheepActionMagnitude = computeVectorNorm(np.array(action))
modifiedSheepAction = np.array(
action) * self.velocity / sheepActionMagnitude
newSheepPos = np.array(oldSheepPos) + modifiedSheepAction
sheepPos = self.checkBound(newSheepPos)
wolfPos = self.checkBound(newWolfPos)
return np.concatenate([sheepPos, wolfPos])
def __call__(self, state, action):
agentState, targetPosition = getEachState(state)
actionMagnitude = computeVectorNorm(np.array(action))
modifiedAction = np.array(action) * self.velocity / actionMagnitude
newAgentState = np.array(agentState) + modifiedAction
if checkBound(newAgentState, self.xBoundary, self.yBoundary):
return np.concatenate([newAgentState, targetPosition])
return np.concatenate([agentState, targetPosition])
def __call__(self):
xMin, xMax = self.xBoundary
yMin, yMax = self.yBoundary
initialAgentState = np.array(
[np.random.uniform(xMin, xMax),
np.random.uniform(yMin, yMax)])
targetPosition = np.array(
[np.random.uniform(xMin, xMax),
np.random.uniform(yMin, yMax)])
initialDistance = computeVectorNorm(targetPosition - initialAgentState)
while not (
checkBound(initialAgentState, self.xBoundary, self.yBoundary)
and checkBound(targetPosition, self.xBoundary, self.yBoundary)
and initialDistance >= 20):
initialAgentState = np.array(
[np.random.uniform(xMin, xMax),
np.random.uniform(yMin, yMax)])
targetPosition = np.array(
[np.random.uniform(xMin, xMax),
np.random.uniform(yMin, yMax)])
initialDistance = computeVectorNorm(targetPosition -
initialAgentState)
return np.concatenate([initialAgentState, targetPosition])