def __init__(self, rs, sizeOfTarget, saveTraj, foldername, thetafile,
popSize, period):
'''
Initializes parameters used to run functions below
Inputs:
'''
self.rs = rs
self.name = "Experiments"
self.call = 0
self.numfeats = rs.numfeats
self.dimState = rs.inputDim
self.dimOutput = rs.outputDim
self.numberOfRepeat = rs.numberOfRepeatEachTraj
self.foldername = foldername
self.tm = TrajMaker(rs, sizeOfTarget, saveTraj, thetafile)
self.posIni = np.loadtxt(pathDataFolder + rs.experimentFilePosIni)
self.costStore = []
self.CMAESCostStore = []
self.CMAESTimeStore = []
self.trajTimeStore = []
self.bestCost = -10000.0
self.lastCoord = []
self.popSize = popSize
self.period = period
def __init__(self, rs, sizeOfTarget, saveTraj, foldername, thetafile, popSize, period, estim="Inv"):
'''
Initializes parameters used to run functions below
Inputs:
'''
self.rs = rs
self.name = "Experiments"
self.call = 0
self.dimState = rs.inputDim
self.dimOutput = rs.outputDim
self.numberOfRepeat = rs.numberOfRepeatEachTraj
self.foldername = foldername
self.tm = TrajMaker(rs, sizeOfTarget, saveTraj, thetafile, estim)
self.posIni = np.loadtxt(pathDataFolder + rs.experimentFilePosIni)
if(len(self.posIni.shape)==1):
self.posIni=self.posIni.reshape((1,self.posIni.shape[0]))
self.costStore = []
self.cost12Store=[]
self.CMAESCostStore = []
self.CMAESTimeStore = []
self.trajTimeStore = []
self.bestCost = -10000.0
self.lastCoord = []
self.popSize = popSize
self.period = period
class Experiments:
def __init__(self, rs, sizeOfTarget, saveTraj, foldername, thetafile,
popSize, period):
'''
Initializes parameters used to run functions below
Inputs:
'''
self.rs = rs
self.name = "Experiments"
self.call = 0
self.numfeats = rs.numfeats
self.dimState = rs.inputDim
self.dimOutput = rs.outputDim
self.numberOfRepeat = rs.numberOfRepeatEachTraj
self.foldername = foldername
self.tm = TrajMaker(rs, sizeOfTarget, saveTraj, thetafile)
self.posIni = np.loadtxt(pathDataFolder + rs.experimentFilePosIni)
self.costStore = []
self.CMAESCostStore = []
self.CMAESTimeStore = []
self.trajTimeStore = []
self.bestCost = -10000.0
self.lastCoord = []
self.popSize = popSize
self.period = period
def printLastCoordInfo(self):
vec = np.array(self.lastCoord)
print "moyenne :", np.mean(vec)
print "min :", np.min(vec)
print "max :", np.max(vec)
def initTheta(self, theta):
'''
Input: -theta: controller ie vector of parameters, numpy array
'''
self.theta = np.asarray(theta).reshape(
(self.dimOutput, self.numfeats**self.dimState))
self.tm.setTheta(self.theta)
def saveCost(self):
filename = findDataFilename(self.foldername + "Cost/", "traj", ".cost")
filenameTime = findDataFilename(self.foldername + "TrajTime/", "traj",
".time")
filenameX = findDataFilename(self.foldername + "finalX/", "x", ".last")
np.savetxt(filename, self.costStore)
np.savetxt(filenameTime, self.trajTimeStore)
np.savetxt(filenameX, self.lastCoord)
def runOneTrajectory(self, x, y):
#self.tm.saveTraj = True
cost, trajTime, lastX = self.tm.runTrajectory(x, y, self.foldername)
#print "Exp local x y cost : ", x, y, cost
if lastX != -1000:
self.lastCoord.append(lastX)
return cost, trajTime
def runRichTrajectories(self, repeat):
globCost = []
xy = np.loadtxt(pathDataFolder + "PosCircu540")
#xy = np.loadtxt(pathDataFolder + "PosSquare")
for el in xy:
costAll, trajTimeAll = np.zeros(repeat), np.zeros(repeat)
for i in range(repeat):
costAll[i], trajTimeAll[i] = self.runOneTrajectory(
el[0], el[1])
meanCost = np.mean(costAll)
meanTrajTime = np.mean(trajTimeAll)
self.costStore.append([el[0], el[1], meanCost])
self.trajTimeStore.append([el[0], el[1], meanTrajTime])
globCost.append(meanCost)
return np.mean(globCost)
def runTrajectoriesForResultsGeneration(self, repeat):
globCost = []
globTime = []
for xy in self.posIni:
costAll, trajTimeAll = np.zeros(repeat), np.zeros(repeat)
for i in range(repeat):
costAll[i], trajTimeAll[i] = self.runOneTrajectory(
xy[0], xy[1])
meanCost = np.mean(costAll)
meanTrajTime = np.mean(trajTimeAll)
self.costStore.append([xy[0], xy[1], meanCost])
self.trajTimeStore.append([xy[0], xy[1], meanTrajTime])
globCost.append(meanCost)
globTime.append(meanTrajTime)
#self.printLastCoordInfo()
return np.mean(globCost), np.mean(globTime)
def runTrajectoriesCMAES(self, theta):
'''
Generates all the trajectories of the experimental setup and return the mean cost. This function is used by cmaes to optimize the controller.
Input: -theta: vector of parameters, one dimension normalized numpy array
Ouput: -meanAll: the mean of the cost of all trajectories generated, float
'''
if (self.call == 0):
self.localBestCost = -1000000.0
self.localWorstCost = 1000000.0
self.localBestTime = -1000000.0
self.localWorstTime = 1000000.0
self.periodMeanCost = 0.0
self.periodMeanTime = 0.0
c = Chrono()
self.initTheta(theta)
#print "theta avant appel :", theta
#compute all the trajectories x times each, x = numberOfRepeat
meanCost, meanTime = self.runTrajectoriesForResultsGeneration(
self.numberOfRepeat)
#cma.plot()
#opt = cma.CMAOptions()
#print "CMAES options :", opt
c.stop()
print("Indiv #: ", self.call, "\n Cost: ", meanCost)
if meanCost > self.localBestCost:
self.localBestCost = meanCost
if meanTime > self.localBestTime:
self.localBestTime = meanTime
if meanCost < self.localWorstCost:
self.localWorstCost = meanCost
if meanTime < self.localWorstTime:
self.localWorstTime = meanTime
if meanCost > self.bestCost:
self.bestCost = meanCost
if meanCost > 0:
extension = ".save" + str(meanCost)
filename = findDataFilename(self.foldername + "Theta/",
"theta", extension)
np.savetxt(filename, self.theta)
self.periodMeanCost += meanCost
self.periodMeanTime += meanTime
self.call += 1
self.call = self.call % self.period
if (self.call == 0):
self.periodMeanCost = self.periodMeanCost / self.period
self.periodMeanTime = self.periodMeanTime / self.period
self.CMAESCostStore.append(
(self.localWorstCost, self.periodMeanCost, self.localBestCost))
self.CMAESTimeStore.append(
(self.localWorstTime, self.periodMeanTime, self.localBestTime))
costfoldername = self.foldername + "Cost/"
checkIfFolderExists(costfoldername)
np.savetxt(costfoldername + "cmaesCost.log", self.CMAESCostStore
) #Note: inefficient, should rather add to the file
np.savetxt(costfoldername + "cmaesTime.log", self.CMAESTimeStore
) #Note: inefficient, should rather add to the file
return 10.0 * (self.rs.rhoCF - meanCost) / self.rs.rhoCF
class Experiments:
def __init__(self, rs, sizeOfTarget, saveTraj, foldername, thetafile, popSize, period):
'''
Initializes parameters used to run functions below
Inputs:
'''
self.rs = rs
self.name = "Experiments"
self.call = 0
self.numfeats = rs.numfeats
self.dimState = rs.inputDim
self.dimOutput = rs.outputDim
self.numberOfRepeat = rs.numberOfRepeatEachTraj
self.foldername = foldername
self.tm = TrajMaker(rs, sizeOfTarget, saveTraj, thetafile)
self.posIni = np.loadtxt(pathDataFolder + rs.experimentFilePosIni)
self.costStore = []
self.CMAESCostStore = []
self.CMAESTimeStore = []
self.trajTimeStore = []
self.bestCost = -10000.0
self.lastCoord = []
self.popSize = popSize
self.period = period
def printLastCoordInfo(self):
vec = np.array(self.lastCoord)
print "moyenne :", np.mean(vec)
print "min :", np.min(vec)
print "max :", np.max(vec)
def initTheta(self, theta):
'''
Input: -theta: controller ie vector of parameters, numpy array
'''
self.theta = np.asarray(theta).reshape((self.dimOutput, self.numfeats**self.dimState))
self.tm.setTheta(self.theta)
def saveCost(self):
filename = findDataFilename(self.foldername+"Cost/","traj",".cost")
filenameTime = findDataFilename(self.foldername+"TrajTime/","traj",".time")
filenameX = findDataFilename(self.foldername+"finalX/","x",".last")
np.savetxt(filename, self.costStore)
np.savetxt(filenameTime, self.trajTimeStore)
np.savetxt(filenameX, self.lastCoord)
def runOneTrajectory(self, x, y):
#self.tm.saveTraj = True
cost, trajTime, lastX = self.tm.runTrajectory(x, y, self.foldername)
#print "Exp local x y cost : ", x, y, cost
if lastX != -1000:
self.lastCoord.append(lastX)
return cost, trajTime
def runRichTrajectories(self, repeat):
globCost = []
xy = np.loadtxt(pathDataFolder + "PosCircu540")
#xy = np.loadtxt(pathDataFolder + "PosSquare")
for el in xy:
costAll, trajTimeAll = np.zeros(repeat), np.zeros(repeat)
for i in range(repeat):
costAll[i], trajTimeAll[i] = self.runOneTrajectory(el[0], el[1])
meanCost = np.mean(costAll)
meanTrajTime = np.mean(trajTimeAll)
self.costStore.append([el[0], el[1], meanCost])
self.trajTimeStore.append([el[0], el[1], meanTrajTime])
globCost.append(meanCost)
return np.mean(globCost)
def runTrajectoriesForResultsGeneration(self, repeat):
globCost = []
globTime = []
for xy in self.posIni:
costAll, trajTimeAll = np.zeros(repeat), np.zeros(repeat)
for i in range(repeat):
costAll[i], trajTimeAll[i] = self.runOneTrajectory(xy[0], xy[1])
meanCost = np.mean(costAll)
meanTrajTime = np.mean(trajTimeAll)
self.costStore.append([xy[0], xy[1], meanCost])
self.trajTimeStore.append([xy[0], xy[1], meanTrajTime])
globCost.append(meanCost)
globTime.append(meanTrajTime)
#self.printLastCoordInfo()
return np.mean(globCost), np.mean(globTime)
def runTrajectoriesCMAES(self, theta):
'''
Generates all the trajectories of the experimental setup and return the mean cost. This function is used by cmaes to optimize the controller.
Input: -theta: vector of parameters, one dimension normalized numpy array
Ouput: -meanAll: the mean of the cost of all trajectories generated, float
'''
if (self.call==0):
self.localBestCost = -1000000.0
self.localWorstCost = 1000000.0
self.localBestTime = -1000000.0
self.localWorstTime = 1000000.0
self.periodMeanCost = 0.0
self.periodMeanTime = 0.0
c = Chrono()
self.initTheta(theta)
#print "theta avant appel :", theta
#compute all the trajectories x times each, x = numberOfRepeat
meanCost, meanTime = self.runTrajectoriesForResultsGeneration(self.numberOfRepeat)
#cma.plot()
#opt = cma.CMAOptions()
#print "CMAES options :", opt
c.stop()
print("Indiv #: ", self.call, "\n Cost: ", meanCost)
if meanCost>self.localBestCost:
self.localBestCost = meanCost
if meanTime>self.localBestTime:
self.localBestTime = meanTime
if meanCost<self.localWorstCost:
self.localWorstCost = meanCost
if meanTime<self.localWorstTime:
self.localWorstTime = meanTime
if meanCost>self.bestCost:
self.bestCost = meanCost
if meanCost>0:
extension = ".save" + str(meanCost)
filename = findDataFilename(self.foldername+"Theta/", "theta", extension)
np.savetxt(filename, self.theta)
self.periodMeanCost += meanCost
self.periodMeanTime += meanTime
self.call += 1
self.call = self.call%self.period
if (self.call==0):
self.periodMeanCost = self.periodMeanCost/self.period
self.periodMeanTime = self.periodMeanTime/self.period
self.CMAESCostStore.append((self.localWorstCost,self.periodMeanCost,self.localBestCost))
self.CMAESTimeStore.append((self.localWorstTime,self.periodMeanTime,self.localBestTime))
costfoldername = self.foldername+"Cost/"
checkIfFolderExists(costfoldername)
np.savetxt(costfoldername+"cmaesCost.log",self.CMAESCostStore) #Note: inefficient, should rather add to the file
np.savetxt(costfoldername+"cmaesTime.log",self.CMAESTimeStore) #Note: inefficient, should rather add to the file
return 10.0*(self.rs.rhoCF-meanCost)/self.rs.rhoCF
class Experiments:
def __init__(self, rs, sizeOfTarget, saveTraj, foldername, thetafile, popSize, period, estim="Inv"):
'''
Initializes parameters used to run functions below
Inputs:
'''
self.rs = rs
self.name = "Experiments"
self.call = 0
self.dimState = rs.inputDim
self.dimOutput = rs.outputDim
self.numberOfRepeat = rs.numberOfRepeatEachTraj
self.foldername = foldername
self.tm = TrajMaker(rs, sizeOfTarget, saveTraj, thetafile, estim)
self.posIni = np.loadtxt(pathDataFolder + rs.experimentFilePosIni)
if(len(self.posIni.shape)==1):
self.posIni=self.posIni.reshape((1,self.posIni.shape[0]))
self.costStore = []
self.cost12Store=[]
self.CMAESCostStore = []
self.CMAESTimeStore = []
self.trajTimeStore = []
self.bestCost = -10000.0
self.lastCoord = []
self.popSize = popSize
self.period = period
def printLastCoordInfo(self):
vec = np.array(self.lastCoord)
print ("moyenne : "+ str(np.mean(vec)))
print ("min : " + str(np.min(vec)))
print ("max :" + str(np.max(vec)))
def initTheta(self, theta):
'''
Input: -theta: controller ie vector of parameters, numpy array
'''
self.theta=theta
self.tm.setTheta(self.theta)
def saveCost(self):
'''
filename = findDataFilename(self.foldername+"Cost/","traj",".cost")
filenameTime = findDataFilename(self.foldername+"TrajTime/","traj",".time")
filenameX = findDataFilename(self.foldername+"finalX/","x",".last")
np.savetxt(filename, self.costStore)
np.savetxt(filenameTime, self.trajTimeStore)
np.savetxt(filenameX, self.lastCoord)
'''
writeArray(self.costStore,self.foldername+"Cost/","traj",".cost")
writeArray(self.cost12Store,self.foldername+"CostU12/","traj",".cost")
writeArray(self.trajTimeStore, self.foldername+"TrajTime/","traj",".time")
writeArray(self.lastCoord, self.foldername+"finalX/","x",".last")
def setNoise(self, noise):
self.tm.setnoise(noise)
def runOneTrajectory(self, x, y):
#self.tm.saveTraj = True
cost, trajTime, lastX = self.tm.runTrajectory(x, y, self.foldername)
#cost, trajTime, lastX = self.tm.runTrajectoryOpti(x, y)
#print "Exp local x y cost : ", x, y, cost
if lastX != -1000:
self.lastCoord.append(lastX)
return cost, trajTime
def runRichTrajectories(self, repeat):
globCost = []
xy = np.loadtxt(pathDataFolder + "PosCircu540")
#xy = np.loadtxt(pathDataFolder + "PosSquare")
for el in xy:
costAll, trajTimeAll = np.zeros(repeat), np.zeros(repeat)
for i in range(repeat):
costAll[i], trajTimeAll[i] = self.runOneTrajectory(el[0], el[1])
meanCost = np.mean(costAll)
meanTrajTime = np.mean(trajTimeAll)
self.costStore.append([el[0], el[1], meanCost])
self.trajTimeStore.append([el[0], el[1], meanTrajTime])
globCost.append(meanCost)
return np.mean(globCost)
def runTrajectoriesForResultsGeneration(self, repeat):
globMeanCost=0.
globTimeCost=0.
for xy in self.posIni:
costAll, trajTimeAll, costU12 = np.zeros(repeat), np.zeros(repeat), np.zeros(repeat)
for i in range(repeat):
costAll[i], trajTimeAll[i] = self.runOneTrajectory(xy[0], xy[1])
costU12[i] = self.tm.costU12
meanCost = np.mean(costAll)
meanTrajTime = np.mean(trajTimeAll)
meanCostU12=np.mean(costU12)
self.costStore.append([xy[0], xy[1], meanCost])
self.trajTimeStore.append([xy[0], xy[1], meanTrajTime])
self.cost12Store.append([xy[0], xy[1], meanCostU12])
globMeanCost+=meanCost
globTimeCost+=meanTrajTime
#self.printLastCoordInfo()
return globMeanCost/len(self.posIni), globTimeCost/len(self.posIni)
def runTrajectoriesForResultsGenerationNController(self, repeat, thetaName):
globMeanCost=0.
globTimeCost=0.
for enum,xy in enumerate(self.posIni):
try :
costAll, trajTimeAll, costU12 = np.zeros(repeat), np.zeros(repeat), np.zeros(repeat)
controllerFileName = thetaName.replace("*",str(enum))
self.tm.controller.load(controllerFileName)
for i in range(repeat):
costAll[i], trajTimeAll[i] = self.runOneTrajectory(xy[0], xy[1])
costU12[i] = self.tm.costU12
meanCost = np.mean(costAll)
meanTrajTime = np.mean(trajTimeAll)
meanCostU12=np.mean(costU12)
self.costStore.append([xy[0], xy[1], meanCost])
self.trajTimeStore.append([xy[0], xy[1], meanTrajTime])
self.cost12Store.append([xy[0], xy[1], meanCostU12])
globMeanCost+=meanCost
globTimeCost+=meanTrajTime
except IOError:
pass
#self.printLastCoordInfo()
return globMeanCost/len(self.posIni), globTimeCost/len(self.posIni)
def runTrajectoriesForResultsGenerationOnePoint(self, repeat, point):
xy = self.posIni[point]
costAll, trajTimeAll = np.zeros(repeat), np.zeros(repeat)
for i in range(repeat):
costAll[i], trajTimeAll[i] = self.runOneTrajectory(xy[0], xy[1])
meanCost = np.mean(costAll)
meanTrajTime = np.mean(trajTimeAll)
return meanCost, meanTrajTime
def runTrajectoriesForResultsGenerationOpti(self, repeat):
globMeanCost=0.
globTimeCost=0.
#pool=Pool()
costAll, trajTimeAll = np.zeros(repeat), np.zeros(repeat)
for xy in self.posIni:
for i in range(repeat):
costAll[i], trajTimeAll[i] = self.runOneTrajectoryOpti(xy[0], xy[1])
meanCost = np.mean(costAll)
meanTrajTime = np.mean(trajTimeAll)
self.costStore.append([xy[0], xy[1], meanCost])
self.trajTimeStore.append([xy[0], xy[1], meanTrajTime])
globMeanCost+=meanCost
globTimeCost+=meanTrajTime
#self.printLastCoordInfo()
size=len(self.posIni)
return globMeanCost/size, globTimeCost/size
def runTrajectoriesForResultsGenerationEstim(self, repeat):
globMeanCost=0.
globTimeCost=0.
#pool=Pool()
costAll, trajTimeAll = np.zeros(repeat), np.zeros(repeat)
for xy in self.posIni:
for i in range(repeat):
costAll[i], trajTimeAll[i] = self.runOneTrajectoryEstim(xy[0], xy[1])
meanCost = np.mean(costAll)
meanTrajTime = np.mean(trajTimeAll)
self.costStore.append([xy[0], xy[1], meanCost])
self.trajTimeStore.append([xy[0], xy[1], meanTrajTime])
globMeanCost+=meanCost
globTimeCost+=meanTrajTime
#self.printLastCoordInfo()
size=len(self.posIni)
return globMeanCost/size, globTimeCost/size
def runMultiProcessTrajectories(self, repeat):
pool=Pool(processes=len(self.posIni))
result = pool.map(partial(self.runNtrajectory, repeat=repeat) , [(x, y) for x, y in self.posIni])
pool.close()
pool.join()
meanCost, meanTraj=0., 0.
for Cost, traj in result:
meanCost+=Cost
meanTraj+=traj
size = len(result)
return meanCost/size, meanTraj/size
def runNtrajectory(self, (x, y), repeat):
costAll, trajTimeAll = np.zeros(repeat), np.zeros(repeat)
for i in range(repeat):
costAll[i], trajTimeAll[i] = self.runOneTrajectoryOpti(x, y)
meanCost = np.mean(costAll)
meanTrajTime = np.mean(trajTimeAll)
self.costStore.append([x, y, meanCost])
self.trajTimeStore.append([x, y, meanTrajTime])
return meanCost, meanTrajTime
