def setPlaces(self, waypoints):
i = 1
while i <= len(self.order):
pivot = str(i)
neighbor = str(i+1)
#check if there is a right neighbor and distance is calculable
if i + 1 <= len(self.order) and self.order[pivot].score < len(waypoints) and self.order[neighbor].score < len(waypoints):
#see if right neighbor surpassed i in the placement
pivotDist = utils.getDistance(self.order[pivot].pos, waypoints[self.order[pivot].score].pos)
neighborDist = utils.getDistance(self.order[neighbor].pos, waypoints[self.order[neighbor].score].pos)
if (self.order[pivot].score < self.order[neighbor].score) or (self.order[pivot].score == self.order[neighbor].score and neighborDist < pivotDist):
#swap
ent = self.order[pivot]
self.order[pivot] = self.order[neighbor]
self.order[pivot].place = i
self.order[neighbor] = ent
self.order[neighbor].place = i+1
#decrement i to see if the right neighbor surpassed anyone else -2
i -= 2
if i < 1:
i = 1
i += 1
# adjust ai max speed
if self.ai[0].place > 1:
self.ai[0].maxSpeed += 1
else:
self.ai[0].maxSpeed -= 1
# cap off max speed
if self.ai[0].maxSpeed > 250:
self.ai[0].maxSpeed = 250
elif self.ai[0].maxSpeed < 150:
self.ai[0].maxSpeed = 150
def isValidDetections(target_coordinates, pred_coordinates):
bestDistance = 100000000
for permutation in itertools.permutations(pred_coordinates.reshape(-1, 2)):
distance = 0
for i in range(len(permutation)):
distance += getDistance(permutation[i][0], permutation[i][1],
target_coordinates[i * 2],
target_coordinates[i * 2 + 1])
if distance < bestDistance:
bestPermutation = permutation
bestDistance = distance
valid = []
invalid = []
limit = getDistance(0, 0, 256, 256) * 0.05
for i in range(len(bestPermutation)):
distance = getDistance(bestPermutation[i][0], bestPermutation[i][1],
target_coordinates[i * 2],
target_coordinates[i * 2 + 1])
if distance < limit:
valid.append(bestPermutation[i][0])
valid.append(bestPermutation[i][1])
else:
invalid.append(bestPermutation[i][0])
invalid.append(bestPermutation[i][1])
return np.reshape(np.array(valid),
(-1, 2)), np.reshape(np.array(invalid), (-1, 2))
def gridToBalls(grid, nballs, otherHand=None):
gridSurface = grid[:, :, 0]
gridWidth = grid.shape[0]
gridHeight = grid.shape[1]
minDistance = getDistance(0, 0, 256, 256) * 0.04
boxWidth = 256 // gridWidth
boxHeight = 256 // gridHeight
ballList = []
for i in range(nballs):
while True: # While ball candidate not accepted
x, y = np.unravel_index(np.argmax(gridSurface),
(gridWidth, gridHeight))
ballX = int(boxWidth * x + boxWidth * grid[x, y, 1])
ballY = int(boxHeight * y + boxHeight * grid[x, y, 2])
gridSurface[x, y] = -1
acceptable = True
for j in range(0, len(ballList), 2):
if getDistance(ballX, ballY, ballList[j],
ballList[j + 1]) < minDistance:
acceptable = False
if otherHand is not None:
if getDistance(ballX, ballY, otherHand[0],
otherHand[1]) < minDistance:
acceptable = False
if acceptable:
ballList.append(ballX)
ballList.append(ballY)
break
return np.array(ballList)
def tiebian():
global sidePoints
sidePoints = np.array(sidePoints)
yawAim = getYawAim(sidePoints)
log.info('已加载导航点，共有%d个点' % len(sidePoints))
index = 0
pos = Point(g.getValue('ship').lat, g.getValue('ship').lng)
# 判断刚开始时，是否需要原地转
state = State.TURN if utils.getDistance(
pos, Point(sidePoints[0, 0], sidePoints[0, 1])) > 3 else State.GO
try:
while index < len(sidePoints):
yaw = g.getValue('ship').yaw
lat = g.getValue('ship').lat
lng = g.getValue('ship').lng
gz = g.getValue('ship').gz
speed = g.getValue('ship').speed
pShip = Point(lat, lng)
pAim = Point(sidePoints[index, 0], sidePoints[index, 1])
isFirst = True if index == 0 else False
if state == State.TURN:
thrust = 0
dire = PID_Turn(pShip, pAim, yaw, -gz, yawAim[index], isFirst)
if dire == 0:
# 转弯完毕，进入直行阶段
state = State.GO
elif state == State.GO:
# 判断直线段或是曲线段，之后五个点期望角度最大误差
slowSpeed = True if angleArray_ptp(
yawAim[index:index + 5]) > 15 or index == 0 else False
log.info('Go:当前期望方向角 %.1f 度，slowSpeed是 %s' % yawAim[index],
slowSpeed)
dire, thrust = PID_Go(pShip, pAim, yaw, -gz, yawAim[index],
slowSpeed, isFirst)
dist = utils.getDistance(pShip, pAim)
if state == State.GO and index > 0:
# 将最近点序号作为当前序号
index = index_cal(pShip, index,
SearchLength=10) #函数中使用了global量sidePoints
log.info('当前处于第 %d 个点' % index)
if dist < 1 and index == 0:
state = State.TURN
index += 1
if index == len(sidePoints):
# 跑完
thrust = 0
dire = 0
data = struct.pack("hhb", int(thrust), int(dire), 0)
g.getValue("can").send(0x544, data)
log.info("%s\t%d\t%.1f\t%.1f\t%d" %
(state, index, thrust, dire, dist))
time.sleep(1)
except Exception:
log.error('贴边Error', exc_info=True)
log.info('导航结束')
def updateHover(self):
self.invalid = False
self.hover_edge = None
self.hover_node = None
self.hover_pos = None
if self.last_node:
# Lock the hover position to an angle from last node
locked_x, locked_y = utils.getAngleLockedPosition(8, mouse.x - self.last_node.x, mouse.y - self.last_node.y)
angle_x, angle_y = self.last_node.x + locked_x, self.last_node.y + locked_y
self.hover_pos = angle_x, angle_y
else:
angle_x, angle_y = mouse.x, mouse.y
# Snap to node
snap_node = None
for node in loop.graph.nodes:
dist = utils.getDistance((angle_x, angle_y), (node.x, node.y))
if node.type is Signal.type:
continue
if dist == 0:
if snap_node is None:
snap_node = (dist, node)
elif dist < snap_node[0]:
snap_node = (dist, node)
if snap_node:
self.hover_pos = snap_node[1].x, snap_node[1].y
self.hover_edge = None
if self.last_node and snap_node in self.last_node.nbors: # No back-tracking
self.invalid = True
else:
self.hover_node = snap_node[1]
if self.hover_node.isBusy():
self.invalid = True
return
else:
# Snap to edge
snap_edge = utils.getPointClosestToEdge(loop.graph.edges.keys(), angle_x, angle_y)
if snap_edge:
distance, n_edge, point = snap_edge
if distance == 0:
self.hover_pos = angle_x, angle_y
self.hover_edge = n_edge
# If too close to a signal or edge is busy, it's invalid
if any(utils.getDistance(point, (node.x, node.y)) < self.signal_spacing
for node in n_edge if node.type is Signal.type) or loop.graph.getEdge(*n_edge).isBusy():
self.invalid = True
return
self.hover_pos = angle_x, angle_y
def PID_Go(pShip, pAim, yawShip, gz): # 直行段PID
global thrust_last # 上一时刻油门值
PID_P = 0.5
# PID_D = 0 在下面if结构中定义
yawAim = utils.getAimYaw(pShip, pAim) # 期望方向角
error = utils.angleDiff(yawAim, yawShip) # 方向角误差
dist = utils.getDistance(pShip, pAim) # 距离误差
if error * gz > 0:
PID_D = -0.8
else:
PID_D = 0.8
dire = PID_P * error + gz * PID_D
dire = max(dire, -100)
dire = min(dire, 100)
thrust = 100 - abs(error) * 3
if dist < 10: # 接近段
tar_speed = dist / 8 - abs(error) / 20 # 目标速度
tar_speed = max(tar_speed, 0.2) # 目标速度限幅在0.2到1之间
tar_speed = min(tar_speed, 1)
err_speed = tar_speed - speed
thrust = thrust_last + err_speed * 50
thrust = abs(thrust)
thrust = max(thrust, 10) # 下限是10
thrust = min(thrust, 100) # 上限100
thrust_last = thrust
return int(dire), thrust
def exploreGetData(self):
print("CURRENT STATE: EXPLORE")
print("CURRENT SUBSTATE: DATA ACQUISITION")
self.distance = utils.getDistance()
self.lastIMU = self.currentIMU
self.currentIMU = self.imu.getImuRawData()
self.exploreState = mbl_bots.PROCESSDATA
def min_distance(mc):
"""
:param mingcheng: 用户输入的景区名称
:return: 返回距离从近到远排序的字典
"""
#先得到用户输入景点的经纬度
NODE_des = N_selector.match("名称", name=mc).first()
R_lng1 = R_selector.match((NODE_des,), "经度").first()
R_lat1 = R_selector.match((NODE_des,), "纬度").first()
lng1 = R_lng1.end_node['name']
lat1 = R_lat1.end_node['name']
R_province = R_selector.match((NODE_des,), "省份").first()
province = R_province.end_node
R_same_province_des=R_selector.match((province,), "省2名")
des={ }
for i in R_same_province_des: #i是省份和景点的关系
#print(i)
d=i.end_node
R_lng2 = R_selector.match((d,), "经度").first()
R_lat2 = R_selector.match((d,), "纬度").first()
lng2 = R_lng2.end_node['name']
lat2 = R_lat2.end_node['name']
dis=utils.getDistance(lat1, lng1, lat2, lng2)
des[d['name']]=dis
print(sorted(des.items(), key=lambda kv: (kv[1], kv[0])))
def updateHover(self):
self.invalid = False
self.hover_edge = None
self.hover_node = None
self.hover_pos = None
# Snap to node
snap_node = None
for node in loop.graph.nodes:
if not node.type is Signal.type:
continue
dist = utils.getNodeDistance(mouse, node)
if dist < self.snap_distance:
if snap_node is None:
snap_node = (dist, node)
elif dist < snap_node[0]:
snap_node = (dist, node)
if snap_node is not None:
self.hover_node = snap_node[1]
self.hover_pos = snap_node[1].x, snap_node[1].y
else:
# Snap to edge
snap_edge = utils.getPointClosestToEdge(loop.graph.edges.keys(), mouse.x, mouse.y)
if snap_edge:
distance, n_edge, point = snap_edge
if distance < self.snap_distance:
self.hover_edge = n_edge
self.hover_pos = point
# If too close to a signal or edge is busy, it's invalid
if any(utils.getDistance(point, (node.x, node.y)) < self.signal_spacing
for node in n_edge if node.type is Signal.type) or loop.graph.getEdge(*n_edge).isBusy():
self.invalid = True
return
def evaluate(agent_net):
n = config.Map.Height
m = config.Map.Width
mazemap = []
for i in range(n):
mazemap.append([])
for j in range(m):
mazemap[i].append(np.zeros(4))
utils.setCellValue(mazemap, i, j, np.random.binomial(1, config.Map.WallDense))
cell_value_memory = {}
open(config.StrongMazeEnv.EvaluateFile, 'w')
for distance in range(1, n + m):
sum_score = 0
for sx in range(n):
for sy in range(m):
if utils.equalCellValue(mazemap, sx, sy, utils.Cell.Wall):
continue
utils.setCellValue(mazemap, sx, sy, utils.Cell.Source)
score = 0
count = 0
output = ''
for tx in range(n):
for ty in range(m):
if utils.equalCellValue(mazemap, tx, ty, utils.Cell.Empty) and utils.getDistance(sx, sy, tx, ty) <= distance:
count += 1
utils.setCellValue(mazemap, tx, ty, utils.Cell.Target)
memory_id = str(sx) + '_' + str(sy) + '_' + str(tx) + '_' + str(ty)
if memory_id in cell_value_memory:
dir_id = cell_value_memory[memory_id]
else:
dir_id = np.array(agent_net.predict(np.array([[mazemap]]))).argmax()
cell_value_memory[memory_id] = dir_id
output += utils.dir_symbols[dir_id]
utils.setCellValue(mazemap, tx, ty, utils.Cell.Empty)
if utils.getDistance(sx, sy, tx, ty) > utils.getDistance(sx + utils.dirs[dir_id][0], sy + utils.dirs[dir_id][1], tx, ty):
score += 1
sum_score += float(score) / count
utils.setCellValue(mazemap, sx, sy, utils.Cell.Empty)
sum_score /= n * m
print [distance, sum_score]
f = open(config.StrongMazeEnv.EvaluateFile, 'a')
f.write(str(distance) + '\t' + str(sum_score) + '\n')
f.close()
def findEndPos(self, city, agent_id):
start_pos = self.agent_pos[agent_id]
min_distance = -1
if city == -1:
for i in range(len(self.source_pos)):
distance = utils.getDistance(self.source_pos[i], start_pos)
if min_distance == -1 or distance < min_distance:
min_distance = distance
end_pos = self.source_pos[i]
else:
for i in range(len(self.hole_pos)):
if self.hole_city[i] != city:
continue
distance = utils.getDistance(self.hole_pos[i], start_pos)
if min_distance == -1 or distance < min_distance:
min_distance = distance
end_pos = self.hole_pos[i]
return end_pos
def computeFittingError(self, data, model):
totalDist = 0
for data in data:
dist = getDistance(data, model[0], model[1], model[2], model[3],
model[4])
totalDist = totalDist + dist
return totalDist
def filterIndexes(data, indexes, threshold, model) :
filteredIndexes = []
for index in indexes :
temp = data[index]
dist = getDistance(temp, model[0], model[1], model[2], model[3], model[4])
if (dist < threshold) :
filteredIndexes.append(index)
return filteredIndexes
def calc(center, cluster):
'''
'center' is a candidate for the cluster center
'cluster' has all the points in the cluster
The return value indicates how far all the points are from the candidate center
'''
distList = []
for c in range(len(cluster)):
dist = utils.getDistance(center, cluster[c])
distList.append(dist)
return np.sqrt(sum(i * i for i in distList))
def exploreGetData(self):
print("CURRENT STATE: EXPLORE")
print("CURRENT SUBSTATE: DATA ACQUISITION")
self.distance = utils.getDistance()
self.lastIMU = self.currentIMU
self.currentIMU = self.imu.getImuRawData()
start_time = time.time()
self.frame = self.camera.getFrame()
self.camera.truncateFrame()
end_time = time.time()
print('It took me {} s to get the frame'.format(start_time - end_time))
self.exploreState = mbl_bots.PROCESSDATA
def assign(data, clusterCtrs):
# Assign each point to one of the clusters
clusterAssignments={}
for idx in range(len(data)):
shortestDist=np.inf
bestCluster=None
for n in range(len(clusterCtrs)):
dist=utils.getDistance(clusterCtrs[n], data[idx])
if dist<shortestDist:
bestCluster=n
shortestDist=dist
clusterAssignments[idx]=bestCluster
return clusterAssignments
def index_cal(pos, index, SearchLength=10):
# 计算当前点之后的SearchLength个点中距离当前最近的点的序号
# 输入参数
# pos为当前位置
# index为当前点序号
# SearchLength为搜索序列长度
# 输出：距离最近的点序号
global sidePoints
if index + SearchLength > len(sidePoints):
SearchLength = len(sidePoints) - index
dist_list = list()
for ii in range(SearchLength):
dist_list.append(utils.getDistance(pos, sidePoints[ii + index]))
index_max = dist_list.index(max(dist_list))
return index_max + index
def getNearestNeighbor(data, cfg):
''' Return a dictionary with the closest neighbor of each time series '''
d = {}
for a in range(len(data)):
shortest = np.inf
neighbor = None
for b in range(len(data)):
if a == b:
pass
else:
dist = utils.getDistance(data[a], data[b])
if dist < shortest:
shortest = dist
d[a] = b
return d
def filter(data, center):
''' Remove two sets of points:
- most distant from the center
- closest to the center
Neither one will make good cluster centers '''
distances = []
for idx in range(len(data)):
d = utils.getDistance(center, data[idx])
distances.append((idx, d))
distances.sort(key=lambda n: n[1])
# remove closest 10%
closest = int(len(data) * .1)
distances = distances[closest:]
# remove furthest 10%
furthest = int(len(distances) * .9)
distances = distances[:furthest]
return distances
def process(self, position):
if self.previousPosition is not None:
if getDistance(position[0], position[1], self.previousPosition[0],
self.previousPosition[1]) < self.maxDistance:
self.streak += 1
if self.streak >= 1:
self.streakPosition = position
self.timeSinceStreak = 0
else:
self.streak = 0
self.timeSinceStreak += 1
self.previousPosition = position
if self.streakPosition is not None and self.timeSinceStreak < 2:
return self.streakPosition
else:
return position
def getPath(self, train):
if not train.end and train.destination is train.start:
return []
if train.end is train.destination:
return [train.end]
node_data = {train.end: self.DataItem(None, 0, 0)}
open_list = [train.end]
closed_list = []
current_node = train.end
last_node = train.start # For one-way rule
while True:
for nbor in current_node.nbors:
if nbor is train.destination:
path = [nbor]
node = current_node
while node is not train.end:
path.append(node)
node = node_data[node].parent
return list(reversed(path))
if nbor is last_node or nbor in closed_list: # Enforce one-way rule (nbor is last_node)
continue
edge = self.graph.getEdge(current_node, nbor)
if nbor in open_list:
g = node_data[current_node].g + edge.length
if g < node_data[nbor].g:
node_data[nbor].g = g
node_data[nbor].parent = current_node
else:
node_data[nbor] = self.DataItem(current_node,
node_data[current_node].g + edge.length,
utils.getDistance((train.destination.x, train.destination.y),
(nbor.x, nbor.y)))
open_list.append(nbor)
open_list.remove(current_node)
closed_list.append(current_node)
last_node = current_node
if not open_list:
return []
current_node = sorted([(node_data[node].g + node_data[node].h, node) for node in open_list],
key=lambda x: x[0], reverse=True)[-1][1]
def PID_Go(pShip,
pAim,
yawShip,
gz,
yawOpt,
slowSpeed=True,
isFirst=True): # 直行段PID
global thrust_last # 上一时刻油门值
PID_P = 0.5
# PID_D = 0 在下面if结构中定义
if isFirst == True:
yawAim = utils.getAimYaw(pShip, pAim) # 期望方向角
else:
yawAim = yawOpt
# *********************补偿项***************注意，距离小的时候不准
# if utils.getAimYaw(pShip, pAim)>5:
# weight=np.array([0.7,0.3])
# yawAim=angleWeightSum(np.array(yawOpt,utils.getAimYaw(pShip, pAim)),weight)
# else:
# yawAim=yawOpt
error = utils.angleDiff(yawAim, yawShip) # 方向角误差
dist = utils.getDistance(pShip, pAim) # 距离误差
if error * gz > 0:
PID_D = -0.8
else:
PID_D = 0.8
dire = PID_P * error + gz * PID_D
dire = max(dire, -100)
dire = min(dire, 100)
thrust = (100 - abs(error) * 3) * 0.8
tar_speed = 0
if dist < 0.8 and slowSpeed == True:
tar_speed = dist / 1.5 - abs(error) / 20 # 目标速度
tar_speed = max(tar_speed, 0.2) # 目标速度限幅在0.2到1之间
tar_speed = min(tar_speed, 1)
err_speed = tar_speed - speed
thrust = thrust_last + err_speed * 50
thrust = abs(thrust)
thrust = max(thrust, 10) # 下限是10
thrust = min(thrust, 100) # 上限100
thrust_last = thrust
log.debug("PID_Go 目标速度：{}\t实际速度: {}\tthrust: {}\tdist: {}".format(
tar_speed, speed, thrust, dist))
return int(dire), thrust
def process(self, balls):
if self.previousBalls is None:
self.previousBalls = balls
return balls
else:
bestDistance = 100000000
bestPermutation = None
for permutation in itertools.permutations(balls):
distance = 0
for i in range(len(permutation)):
distance += getDistance(permutation[i][0],
permutation[i][1],
self.previousBalls[i][0],
self.previousBalls[i][1])
if distance < bestDistance:
bestPermutation = permutation
bestDistance = distance
self.previousBalls = bestPermutation
return np.array(self.previousBalls)
def tick(self, dtime):
self.ent.AI_UP = False
self.ent.AI_DOWN = False
self.ent.AI_LEFT = False
self.ent.AI_RIGHT = False
self.ent.AI_FORWARD = False
self.ent.AI_REVERSE = False
if self.ent.goals:
self.currentGoal = self.ent.goals[0].pos
if self.finished == False:
dist = self.range
if self.currentGoal != None:
#get distane from goal
dist = utils.getDistance(self.ent.pos, self.currentGoal)
self.setControls(self.currentGoal)
#if distance is within range, complete goal and move onto the next one
if dist < self.range:
self.ent.score += 1 #increment score counter for ai to keep track of the winner and losers
self.nextGoal()
def calc_mark(self, state): #a ameliorer pour la gestion des ennemis
if self.assignedGroup.eff == 0:
print(
"Warning : calc_mark tente d'evaluer les actions d'un groupe vide"
)
return 0
if self.action_type == ActionType.attackHuman:
distance = utils.getDistance(self.assignedGroup, self.target_group)
groupe_max = Group(0, 0, 0, self.assignedGroup.species.inverse()
) #on initialise un groupe max de base, a 0
for group in state.groupes:
if group.species == self.assignedGroup.species.inverse() \
and utils.getDistance(group,self.target_group) < distance \
and group.eff >= self.target_group.eff \
and group.eff>groupe_max.eff: #on cherche les groupes ennemis plus proches que nous de la cible, plus nombreux que la cible, et on garde le plus gros d'entre eux
groupe_max = group
enemyWinner = utils.simulateBattle(
groupe_max, self.target_group
) # on simule une bataille entre ce groupe_max (cad le groupe d'enemis plus proche que nous, ou, à défaut, un groupe fictif vide, qui perdra forcément la bataille)
winnerFinal = utils.simulateBattle(
self.assignedGroup, enemyWinner
) # et on simule une bataille entre nous et le gagnant de la première bataille
#et ensuite on traite les differents cas
if winnerFinal.species == self.assignedGroup.species: #soit on gagne avec certitude la derniere bataille
self.possibleEnemyGain = -groupe_max.eff
self.possibleGain = (winnerFinal.eff) - self.assignedGroup.eff
elif winnerFinal.species == Species.human: # soit les humains remportent et nos deux groupes se sont faits bouffer (on a été cons)
self.possibleEnemyGain = -groupe_max.eff
self.possibleGain = -self.assignedGroup.eff
else: # soit les enemis gagnent et on perd
self.possibleEnemyGain = (winnerFinal.eff) - groupe_max.eff
self.possibleGain = -self.assignedGroup.eff
elif self.action_type == ActionType.attackEnemy:
distance = utils.getDistance(self.assignedGroup, self.target_group)
groupe_max = Group(
0, 0, 0,
Species.human) #on initialise un groupe max de base, a 0
for group in state.groupes:
if group.species == Species.human \
and utils.getDistance(group,self.target_group) < distance \
and group.eff <= self.target_group.eff \
and group.eff>groupe_max.eff: #on cherche les groupes d'humains plus proches que nous de la cible ennemie, moins nombreuse que la cible, et on garde le plus gros d'entre eux
groupe_max = group
enemyWinner = utils.simulateBattle(
self.target_group, groupe_max
) # on simule une bataille entre ce groupe_max (cad le groupe d'humains plus proche que nous, ou, à défaut, un groupe fictif vide, qui perdra forcément la bataille)
winnerFinal = utils.simulateBattle(
self.assignedGroup, enemyWinner
) # et on simule une bataille entre nous et le gagnant de la première bataille
if winnerFinal.species == self.assignedGroup.species:
self.possibleGain = (winnerFinal.eff) - self.assignedGroup.eff
self.possibleEnemyGain = -self.target_group.eff
else:
self.possibleGain = -self.assignedGroup.eff
self.possibleEnemyGain = (winnerFinal.eff -
self.target_group.eff)
elif self.action_type == ActionType.run: #todo a implementer en cas de plateau vide à la fin si on est plus nombreux que les ennemis
self.possibleGain = 0
else:
print("type d'action non reconnu par calc_mark", self.action_type)
intuitive_mark = (self.possibleGain - self.possibleEnemyGain)
#enfin, on divise les notes intuitives par la distance pour prendre en compte des incertitudes
if intuitive_mark >= 0:
self.mark = intuitive_mark / float(
utils.getDistance(
self.assignedGroup,
self.target_group)) #todo diviser par distance au carre ?
else: #mais on garde les notes négatives telles quelles pour leur donner plus de poids negatif
self.mark = intuitive_mark
def test_cityDistance():
london = [-0.118092, 51.509865]
paris = [2.352222, 48.856614]
assert (utils.getDistance(paris[0], paris[1], london[0], london[1]) == 340)
def getNearestVehicles(latitude, longitude, route=None):
vehicles = getVehiclesByRoute(route)
vehicles.sort(key=lambda v: utils.getDistance(v.latitude, v.longitude, latitude, longitude))
return vehicles
def __init__(self):
self.previousPosition = None
self.streak = 0
self.streakPosition = None
self.timeSinceStreak = 0
self.maxDistance = getDistance(0, 0, 256, 256) * 0.1
thread.start_new_thread(serial_recv_thread, ())
# 开启RTK线程
thread.start_new_thread(ntrip.fun, ())
# 等待定位成功
while g.getValue('lat') == None or g.getValue('lat') == 0:
time.sleep(1)
pass
logging.info('已定位')
# 获取导航点
getPoints('gps.csv')
if len(sidePoints) > 1:
logging.info('已加载导航点，共有%d个点' % len(sidePoints))
index = 0
pos = Point(g.getValue('lat'), g.getValue('lng'))
# 判断刚开始时，是否需要原地转
state = State.TURN if utils.getDistance(
pos, sidePoints[0]) > 3 else State.GO
while index < len(sidePoints):
pos = Point(g.getValue('lat'), g.getValue('lng'))
if state == State.TURN:
thrust = 0
dire = PID_Turn(pos, sidePoints[index], yaw, -gz)
if dire == 0:
# 转弯完毕，进入直行阶段
state = State.GO
elif state == State.GO:
dire, thrust = PID_Go(pos, sidePoints[index], yaw, -gz)
# 已到达
dist = utils.getDistance(pos, sidePoints[index])
if dist < 1:
if index == 0:
def enumerate_possible_actions(state, group, specie, number_my_groups, max_split_rate):
groups_human = state.getMembers(Species.human)
groups_enemy = state.getMembers(specie.inverse())
actions_total = []
len_group_me = group.eff
actions_simple_per_group = []
actions_split_per_group = []
doublets = []
groups_targets = []
#on elague les groupes d'humains
humanDistances = []
for humangroup in groups_human:
humanDistances.append(utils.getDistance(group, humangroup))
groups_human.sort(key=dict(zip(groups_human, humanDistances)).get, reverse=False)
groups_human=groups_human[:len_group_me+1]
#de même pour les ennemis
enemyDistances = []
for enemy in groups_enemy:
enemyDistances.append(utils.getDistance(group, enemy))
groups_enemy.sort(key=dict(zip(groups_enemy, enemyDistances)).get, reverse=False)
groups_enemy=groups_human[:len_group_me+1]
# actions sans split
for group_human in groups_human:
action = Action(ActionType.attackHuman, group_human, group)
action.calc_mark(state)
actions_simple_per_group.append(action)
groups_targets.append(group_human)
for group_enemy in groups_enemy:
action = Action(ActionType.attackEnemy, group_enemy, group)
action.calc_mark(state)
actions_simple_per_group.append(action)
groups_targets.append(group_enemy)
# actions avec splits
if number_my_groups <= max_split_rate: #on évite de trop se splitter
for i in range(1, int(len_group_me/2)+1):
doublets.append([i, len_group_me-i])
for doublet in doublets:
group1 = Group(group.x, group.y, doublet[0], specie)
group2 = Group(group.x, group.y, doublet[1], specie)
for target_group_1 in groups_targets:
action_type_1 = specie.determine_action_type(target_group_1.species)
for target_group_2 in groups_targets:
action_type_2 = specie.determine_action_type(target_group_2.species)
# si les deux targets sont différentes :
if (target_group_1.x != target_group_2.x) or (target_group_1.y != target_group_2.y):
action1 = Action(action_type_1, target_group_1, group1)
action2 = Action(action_type_2, target_group_2, group2)
action1.calc_mark(state)
action2.calc_mark(state)
action1.parent_group = group
action2.parent_group = group
actions_split_per_group.append([action1, action2])
actions_total.append(actions_simple_per_group)
actions_total.append(actions_split_per_group)
return actions_total
def serial_recv_thread():
ser = serial.Serial("/dev/ttyAMA0", 115200)
gz = 0
flag_ang_control = 0
flag_finish = False
flag_back = False
while True:
if True or g.getValue('mode') != 1 and g.getValue('mode') != 0:
try:
msg = ser.readline().strip()
if msg.startswith('$') and msg.endswith('#'):
datas = msg.replace('$', '').replace('#', '').split(';')
if len(datas) >= 10:
yaw = float(datas[3])
gps_speed = float(datas[7])
if flag_finish:
angle_error = 279 - yaw
if angle_error > 180:
angle_error -= 360
elif angle_error <= -180:
angle_error += 360
if flag_back:
dire = angle_control_5(angle_error, -gz)
dire = 0
ser.write(utils.parityGenerate(
"$E;-25,%d#" % dire))
dist = utils.getDistance(Point(lat_rev1, lng_rev1), Point(
g.getValue('lat'), g.getValue('lng')))
print(dist, angle_error)
if dist < 2:
time.sleep(1)
ser.write(utils.parityGenerate("$E;0,0#"))
break
continue
if abs(angle_error) < 10:
ser.write(utils.parityGenerate("$E;-25,0#"))
flag_back = True
dire = angle_control_4(angle_error, -gz)
print("角度差、方向油门：%d\t%d" % (angle_error, dire))
ser.write(utils.parityGenerate(
"$E;%d,%d#" % (0, dire)))
continue
(thrust, dire, flag_ang_control) = control_reverse_point(
g.getValue('lat'), g.getValue('lng'), yaw, gps_speed, -gz, lat_rev, lng_rev, flag_ang_control)
# print(g.getValue('lat'), g.getValue(
# 'lng'), g.getValue('mode'), yaw, gps_speed, thrust, dire)
ser.write(utils.parityGenerate(
"$E;%d,%d#" % (thrust, dire)))
if thrust == 0 and dire == 0:
flag_finish = True
elif msg.startswith('&') and msg.endswith('#'):
datas = msg.replace('&', '').replace('#', '').split(';')
gz = float(datas[0])
except KeyboardInterrupt:
break
except Exception:
continue
ser.close()
print('Finish')
with open(tmp_input, 'w') as f:
f.write('Host1\tHost2\tDistance\n')
for i in xrange(len(species)):
for j in xrange(len(species)):
if i != j:
host1 = species[i]
host2 = species[j]
sum = float(0)
elms_compared = 0
for elm in utils_graph.unionLists([host2elmFreqs[host1],
host2elmFreqs[host2]]):
for h in (host1, host2):
if elm not in host2elmFreqs[h]:
host2elmFreqs[h][elm] = {}
if len(host2elmFreqs[host1][elm].keys()) != 0 and len(host2elmFreqs[host2][elm].keys()) != 0:
sum += utils.getDistance(host2elmFreqs[host1][elm],
host2elmFreqs[host2][elm])
elms_compared += 1
f.write('%s\t%s\t%.10f\n'
% (short_names[host1],
short_names[host2],
float(sum)/float(elms_compared)))
out_file = 'plots/for_aydin_2/cos_dis_sum_host' + suffix + '.png'
tmp_r = 'tmp_r' + str(random.randint(0,100))
with open(tmp_r, 'w') as f:
f.write('library(ggplot2)\n')
f.write("d<-read.delim('"
+ tmp_input + "', header=T, sep='\\t')\n")
f.write("png('" + out_file + "')\n")
f.write("ggplot(d,aes(Host1,Host2)) + geom_tile(aes(fill=Distance),colour='white') + scale_fill_gradient(low='green',high='steelblue')\n")
f.write('dev.off()\n')
(elm, fq) = line.strip().split('\t')
freqs_single[elm] = float(fq)
freqs[host] = freqs_single
tmp_input = 'tmp_i' + str(random.randint(0,100))
tmp_r = 'tmp_r' + str(random.randint(0,100))
out_file = 'plots/for_aydin_2/elm_count_dis_redo.png'
with open(tmp_input, 'w') as f:
f.write('Host1\tHost2\tDistance\n')
for i in xrange(len(species)):
for j in xrange(len(species)):
if i != j:
host1 = species[i]
host2 = species[j]
f.write('%s\t%s\t%.10f\n'
% (short_names[host1],
short_names[host2],
utils.getDistance(freqs[host1],
freqs[host2])))
with open(tmp_r, 'w') as f:
f.write('library(ggplot2)\n')
f.write("d<-read.delim('"
+ tmp_input + "', header=T, sep='\\t')\n")
f.write("png('" + out_file + "')\n")
f.write("ggplot(d,aes(Host1,Host2)) + geom_tile(aes(fill=Distance),colour='white') + scale_fill_gradient(low='white',high='steelblue')\n")
f.write('dev.off()\n')
os.system('R < ' + tmp_r + ' --no-save')
#os.system('rm ' + tmp_r + ' ' + tmp_input)
def enumerate_possible_actions(state, group, specie, number_my_groups,
max_split_rate):
groups_human = state.getMembers(Species.human)
groups_enemy = state.getMembers(specie.inverse())
actions_total = []
len_group_me = group.eff
actions_simple_per_group = []
actions_split_per_group = []
doublets = []
groups_targets = []
#on elague les groupes d'humains
humanDistances = []
for humangroup in groups_human:
humanDistances.append(utils.getDistance(group, humangroup))
groups_human.sort(key=dict(zip(groups_human, humanDistances)).get,
reverse=False)
groups_human = groups_human[:len_group_me + 1]
#de même pour les ennemis
enemyDistances = []
for enemy in groups_enemy:
enemyDistances.append(utils.getDistance(group, enemy))
groups_enemy.sort(key=dict(zip(groups_enemy, enemyDistances)).get,
reverse=False)
groups_enemy = groups_human[:len_group_me + 1]
# actions sans split
for group_human in groups_human:
action = Action(ActionType.attackHuman, group_human, group)
action.calc_mark(state)
actions_simple_per_group.append(action)
groups_targets.append(group_human)
for group_enemy in groups_enemy:
action = Action(ActionType.attackEnemy, group_enemy, group)
action.calc_mark(state)
actions_simple_per_group.append(action)
groups_targets.append(group_enemy)
# actions avec splits
if number_my_groups <= max_split_rate: #on évite de trop se splitter
for i in range(1, int(len_group_me / 2) + 1):
doublets.append([i, len_group_me - i])
for doublet in doublets:
group1 = Group(group.x, group.y, doublet[0], specie)
group2 = Group(group.x, group.y, doublet[1], specie)
for target_group_1 in groups_targets:
action_type_1 = specie.determine_action_type(
target_group_1.species)
for target_group_2 in groups_targets:
action_type_2 = specie.determine_action_type(
target_group_2.species)
# si les deux targets sont différentes :
if (target_group_1.x != target_group_2.x) or (
target_group_1.y != target_group_2.y):
action1 = Action(action_type_1, target_group_1, group1)
action2 = Action(action_type_2, target_group_2, group2)
action1.calc_mark(state)
action2.calc_mark(state)
action1.parent_group = group
action2.parent_group = group
actions_split_per_group.append([action1, action2])
actions_total.append(actions_simple_per_group)
actions_total.append(actions_split_per_group)
return actions_total
def main(self,img):
transform = transforms.Compose([transforms.ToTensor()])
with torch.no_grad():
height, width = img.shape[:2]
img_det = cv2.cvtColor(img, cv2.COLOR_BGR2RGB)
det = self.mtcnn.detect_face(img_det)
# '''
# [{'box': (160, 48, 322, 270), 'cls': array([0.9666081], dtype=float32),
# 'pts': {'leye': (214, 136), 'reye': (276, 121), 'nose': (257, 159), 'lmouse': (238, 208),
# 'rmouse': (288, 197)}}]
# '''
for i in range(len(det)): #单张图片人脸数量
box = det[i]['box'] #人脸框tuple
#cls = result[i]['cls'] #置信度ndarry
pts = det[i]['pts'] #五官坐标dict
x1, y1, x2, y2 = box #左上右下
dis = y2 - y1
#cv2.rectangle(img, (x1, y1), (x2, y2), (255, 255, 25)) #天蓝色人脸框
w = x2 - x1 + 1
h = y2 - y1 + 1
size_w = int(max([w, h])*0.9)
size_h = int(max([w, h]) * 0.9)
cx = x1 + w//2
cy = y1 + h//2
x1 = cx - size_w//2
x2 = x1 + size_w
y1 = cy - int(size_h * 0.4)
y2 = y1 + size_h
left = 0
top = 0
bottom = 0
right = 0
if x1 < 0:
left = -x1
if y1 < 0:
top = -y1
if x2 >= width:
right = x2 - width
if y2 >= height:
bottom = y2 - height
x1 = max(0, x1)
y1 = max(0, y1)
x2 = min(width, x2)
y2 = min(height, y2)
cropped = img[y1:y2, x1:x2] #裁剪出的人脸
# print(cropped.shape)
# np_img = img[int(y1/1):y2+int(y1/1),int(x1/1):x2+int(x1/1)]
# cv2.imshow(str(numa),np_img)
cropped = cv2.copyMakeBorder(cropped, top, bottom, left, right, cv2.BORDER_CONSTANT, 0)
input = cv2.resize(cropped, (112, 112))
input = cv2.cvtColor(input, cv2.COLOR_BGR2RGB)
input = transform(input).unsqueeze(0).cuda()
pose, landmarks = self.pfldmodel(input)
#poses = pose.cpu().detach().numpy()[0] * 180 / np.pi
# 长度3 pitch是围绕X轴旋转,也叫做俯仰角。 yaw是围绕Y轴旋转,也叫偏航角。 roll是围绕Z轴旋转,也叫翻滚角
pre_landmark = landmarks[0]
pre_landmark = pre_landmark.cpu().detach().numpy().reshape(-1, 2) * [size_w, size_h] # 长度98
# cv2.rectangle(img,(x1, y1), (x2, y2),(255,0,0)) #蓝色正方形
fatigue = []
for num, (x, y) in enumerate(pre_landmark.astype(np.int32)):
#cv2.circle(img, (x1 - left + x, y1 - bottom + y), 1, (255, 255, 0), 1)
#if 59 < num < 76 or num in [96,97]: #眼眶坐标
#眼镜轮廓坐标
# 62 70
# 60 64 68 72
# 66 74
if num in [60, 62, 64, 66, 68, 70, 72, 74]:
cv2.circle(img, (x1 - left + x, y1 - bottom + y), 1, (255, 255, 0), 1)
fatigue.append((x, y))
# print(fatigue)
rightrow = getDistance(fatigue[0], fatigue[2])
rightcol = getDistance(fatigue[1],fatigue[3])
leftrow = getDistance(fatigue[4],fatigue[6])
leftcol = getDistance(fatigue[5],fatigue[7])
numerator = rightcol+leftcol
denominator = rightrow+leftrow
distance = numerator/denominator
print('dis:'+str(distance))
text ,color= 'eyes closed!' if distance < 0.17 else 'eyes opened',\
(0, 0, 255) if distance < 0.17 else (0, 255, 0)
img = cv2.putText(img, text, (50, 50), cv2.FONT_HERSHEY_SIMPLEX, 1.2, color, 2)
# another way 计算
# eye = rightcol + leftcol
# if eye / dis < 0.03:
# print('dis:' + str(distance))
# print('eyes closed!')
# else:
# print('ok')
# plotPosecube(img, poses[0], poses[1], poses[2], tdx=pts['nose'][0], tdy=pts['nose'][1],
# size=(x2 - x1) // 2)
cv2.imshow('example', img)
cv2.waitKey(0)
def getNearestStops(latitude, longitude, route=None):
stops = getStopsByRoute(route)
stops.sort(key=lambda s: utils.getDistance(s.latitude, s.longitude, latitude, longitude))
return stops
def serial_recv_thread():
global gps_speed
ser = serial.Serial("/dev/ttyAMA0", 115200)
gz = 0
flag_ang_control = 0
flag_finish = False
flag_back = False
while True:
if True or g.getValue('mode') != 1 and g.getValue('mode') != 0:
try:
msg = ser.readline().strip()
if msg.startswith('$') and msg.endswith('#'):
datas = msg.replace('$', '').replace('#', '').split(';')
if len(datas) >= 10:
yaw = float(datas[3])
gps_speed = float(datas[7])
if flag_finish: # 到达倒库点
angle_error = 279 - yaw
if angle_error > 180:
angle_error -= 360
elif angle_error <= -180:
angle_error += 360
if flag_back: # 在倒库点修完角，开始后退
# 边后退边修角，修角PID采用PID_Go中的
PID_P = 0.5
if angle_error * (-gz) > 0:
PID_D = -0.8
else:
PID_D = 0.8
dire = PID_P * angle_error + (-gz) * PID_D
dire = saturate(dire, -100, 100)
#考虑是否使用PID_Go然后thrust加负号
ser.write(
utils.parityGenerate("$E;-15,%d#" % dire))
dist = utils.getDistance(
Point(lat_rev1, lng_rev1),
Point(g.getValue('lat'),
g.getValue('lng'))) # 离最终点距离
print(dist, angle_error)
if dist < 2: # 离最终点距离小于2m，停止
time.sleep(1)
ser.write(utils.parityGenerate("$E;0,0#"))
break
continue
if abs(angle_error) < 10:
ser.write(utils.parityGenerate("$E;-15,0#"))
flag_back = True
dire = PID_Turn(angle_error, yaw_now,
-gz) # 在倒库点修角
# dire = angle_control_4(angle_error, -gz)
print("角度差、方向油门：%d\t%d" % (angle_error, dire))
ser.write(
utils.parityGenerate("$E;%d,%d#" % (0, dire)))
continue
(thrust, dire,
flag_ang_control) = control_reverse_point(
g.getValue('lat'), g.getValue('lng'), yaw,
gps_speed, -gz, lat_rev, lng_rev,
flag_ang_control)
# print(g.getValue('lat'), g.getValue(
# 'lng'), g.getValue('mode'), yaw, gps_speed, thrust, dire)
ser.write(
utils.parityGenerate("$E;%d,%d#" % (thrust, dire)))
if thrust == 0 and dire == 0:
flag_finish = True
elif msg.startswith('&') and msg.endswith('#'):
datas = msg.replace('&', '').replace('#', '').split(';')
gz = float(datas[0])
except KeyboardInterrupt:
break
except Exception:
continue
ser.close()
print('Finish')
def getNearestVehicles(latitude, longitude, route=None):
vehicles = getVehiclesByRoute(route)
vehicles.sort(key=lambda v: utils.getDistance(v.latitude, v.longitude,
latitude, longitude))
return vehicles
def AStar(self, agent_id, start_pos, end_pos, agent_pos):
open_set = []
close_set = []
h = [{} for i in range(self.window + 2)]
g = [{} for i in range(self.window + 2)]
path = [{} for i in range(self.window + 2)]
open_set.append([start_pos, 0])
pos = start_pos
g[0][encode(pos)] = 0
h[0][encode(pos)] = utils.getDistance(start_pos, end_pos)
path[0][encode(pos)] = [[-1, -1], -1, -1]
schedule = []
while len(open_set) > 0:
min_f = -1
idx = -1
for i in range(len(open_set)):
[pos, t] = open_set[i]
f = g[t][encode(pos)] + h[t][encode(pos)]
if min_f == -1 or f < min_f:
min_f = f
idx = i
[pos, t] = open_set[idx]
open_set.pop(idx)
close_set.append([pos, t])
#dirs = utils.dirs + [[0, 0]]
dirs = utils.dirs
for i in range(len(dirs)):
a = dirs[i]
new_pos = [pos[0] + a[0], pos[1] + a[1]]
#print new_pos
if not self.isValidPath(agent_id, new_pos, t + 1, agent_pos,
end_pos):
continue
if [new_pos, t + 1] in close_set:
continue
new_g = g[t][encode(pos)] + 1
if [new_pos, t + 1] not in open_set:
open_set.append([new_pos, t + 1])
g[t + 1][encode(new_pos)] = new_g
h[t + 1][encode(new_pos)] = utils.getDistance(
new_pos, end_pos)
path[t + 1][encode(new_pos)] = [pos, t, a]
elif new_g < g[t + 1][encode(new_pos)]:
g[t + 1][encode(new_pos)] = new_g
path[t + 1][encode(new_pos)] = [pos, t, a]
if pos == end_pos or g[t][encode(pos)] == self.window or len(
open_set) == 0:
while (path[t][encode(pos)] != [[-1, -1], -1, -1]):
[pos, t, a] = path[t][encode(pos)]
schedule.insert(0, [pos, a])
break
if len(schedule) == 0:
#print 'fail to schedule, stay'
#print path
schedule.append([start_pos, [0, 0]])
[pos, a] = schedule[-1]
new_pos = [pos[0] + a[0], pos[1] + a[1]]
schedule.append([new_pos, None])
#print 'self.AStar'
#print ['start_pos', start_pos, 'end_pos', end_pos]
#print ['schedule', schedule]
#print 'End self.AStar'
return schedule
def calc_mark(self, state): #a ameliorer pour la gestion des ennemis
if self.assignedGroup.eff==0:
print("Warning : calc_mark tente d'evaluer les actions d'un groupe vide")
return 0
if self.action_type == ActionType.attackHuman:
distance = utils.getDistance(self.assignedGroup,self.target_group)
groupe_max = Group(0,0,0,self.assignedGroup.species.inverse()) #on initialise un groupe max de base, a 0
for group in state.groupes:
if group.species == self.assignedGroup.species.inverse() \
and utils.getDistance(group,self.target_group) < distance \
and group.eff >= self.target_group.eff \
and group.eff>groupe_max.eff: #on cherche les groupes ennemis plus proches que nous de la cible, plus nombreux que la cible, et on garde le plus gros d'entre eux
groupe_max=group
enemyWinner = utils.simulateBattle(groupe_max,self.target_group) # on simule une bataille entre ce groupe_max (cad le groupe d'enemis plus proche que nous, ou, à défaut, un groupe fictif vide, qui perdra forcément la bataille)
winnerFinal = utils.simulateBattle(self.assignedGroup,enemyWinner) # et on simule une bataille entre nous et le gagnant de la première bataille
#et ensuite on traite les differents cas
if winnerFinal.species==self.assignedGroup.species: #soit on gagne avec certitude la derniere bataille
self.possibleEnemyGain = -groupe_max.eff
self.possibleGain = (winnerFinal.eff)-self.assignedGroup.eff
elif winnerFinal.species == Species.human: # soit les humains remportent et nos deux groupes se sont faits bouffer (on a été cons)
self.possibleEnemyGain = -groupe_max.eff
self.possibleGain = -self.assignedGroup.eff
else: # soit les enemis gagnent et on perd
self.possibleEnemyGain = (winnerFinal.eff)-groupe_max.eff
self.possibleGain = -self.assignedGroup.eff
elif self.action_type == ActionType.attackEnemy:
distance = utils.getDistance(self.assignedGroup,self.target_group)
groupe_max = Group(0,0,0,Species.human) #on initialise un groupe max de base, a 0
for group in state.groupes:
if group.species == Species.human \
and utils.getDistance(group,self.target_group) < distance \
and group.eff <= self.target_group.eff \
and group.eff>groupe_max.eff: #on cherche les groupes d'humains plus proches que nous de la cible ennemie, moins nombreuse que la cible, et on garde le plus gros d'entre eux
groupe_max=group
enemyWinner = utils.simulateBattle(self.target_group,groupe_max) # on simule une bataille entre ce groupe_max (cad le groupe d'humains plus proche que nous, ou, à défaut, un groupe fictif vide, qui perdra forcément la bataille)
winnerFinal = utils.simulateBattle(self.assignedGroup,enemyWinner) # et on simule une bataille entre nous et le gagnant de la première bataille
if winnerFinal.species==self.assignedGroup.species:
self.possibleGain = (winnerFinal.eff)-self.assignedGroup.eff
self.possibleEnemyGain = -self.target_group.eff
else :
self.possibleGain = -self.assignedGroup.eff
self.possibleEnemyGain = (winnerFinal.eff-self.target_group.eff)
elif self.action_type == ActionType.run: #todo a implementer en cas de plateau vide à la fin si on est plus nombreux que les ennemis
self.possibleGain = 0
else :
print("type d'action non reconnu par calc_mark", self.action_type)
intuitive_mark = (self.possibleGain-self.possibleEnemyGain)
#enfin, on divise les notes intuitives par la distance pour prendre en compte des incertitudes
if intuitive_mark >= 0:
self.mark = intuitive_mark/float(utils.getDistance(self.assignedGroup,self.target_group)) #todo diviser par distance au carre ?
else : #mais on garde les notes négatives telles quelles pour leur donner plus de poids negatif
self.mark = intuitive_mark
('H_sapiens', 'Gallus_gallus'),
('H_sapiens', 'Taeniopygia_guttata'),
('H_sapiens', 'D_rerio')
):#itertools.combinations(hosts.keys(),2):
use_dict_1 = {}
use_dict_2 = {}
if elm in host2elmFreqs[host1]:
use_dict_1 = host2elmFreqs[host1][elm]
if elm in host2elmFreqs[host2]:
use_dict_2 = host2elmFreqs[host2][elm]
if len(use_dict_1.keys()) == 0 and len(use_dict_2.keys()) == 0:
pass
# f.write('%s\t%s\t%.10f\n'
# % (str(10), elm, float(0)))
else:
dis = utils.getDistance(use_dict_2,
use_dict_1)
f.write('%s\t%s\t%.10f\n'
% (str(counter), elm, dis))
counter += 1
out_file = 'plots/for_aydin_2/cos_dis_host' + suffix + '.png'
tmp_r = 'tmp_r' + str(random.randint(0,100))
with open(tmp_r, 'w') as f:
f.write('library(ggplot2)\n')
f.write("d<-read.delim('"
+ tmp_input + "', header=T, sep='\\t')\n")
f.write("png('" + out_file + "')\n")
f.write("ggplot(d,aes(Host_Host,ELM)) + opts(axis.text.x = theme_blank()) + opts(axis.text.y = theme_blank()) + geom_tile(aes(fill=Distance),colour='white') + scale_fill_gradient(low='green',high='steelblue')\n")
f.write('dev.off()\n')
os.system('R < ' + tmp_r + ' --no-save')
def getNearestStops(latitude, longitude, route=None):
stops = getStopsByRoute(route)
stops.sort(key=lambda s: utils.getDistance(s.latitude, s.longitude,
latitude, longitude))
return stops