我自己在python中,如果说技巧,就是教程上说的,list的那些用法, dict, defaultdict, collection, set, array, numpy, blist, event, socket, cython, __init__, __all__, __doc__, keyerror还有些常用的库。 这些标准教程上的东西,你学会了,给自己帮助很大。 也都是技巧 。
偶尔用一用lambda, map, filter, zip就足够了。 可以缩小代码量。
多用multiprocess少用thread和threading。 有时间可以研究一下stackless python, twist,它的思想很受启发。 tornado, django, jinja2等都需要学一下,简单实用,强大。
最近流行的openstack也要看一下。zope这东西太古老了,如果你真有时间还是可以借鉴一下。
python2, python3都要学习。 不能一味抵制python3, 其中有很多好的思想。
GIL不是不可逾越的。
如果喜欢windows就学一学win32 api, 反之QT, wxwindows, gtk都可以看一看。 html5, node.js, javascript, bootstrap都是好的GUI工具。 要想快速开发, 很失望的说,只有在windows平台下可以找到可视化的快速开发工具。 C#, delphi, 都是很难超越的东西。
python用得人多, 用好的人也多。 但是水准高,思想又好,编程也强大,可以创造性的做产品的人不多。 相反,模仿别人, 借鉴其它的库,拿来主义,这是python的特点。 swig这个东西学一学。
以后你还会依赖其它的语言,需要的时候就多学一学其它的语言,甚至 ruby也是必要的。 java, c++, haskell, go, lua, javascript, php, scala.
但是有一天,python一定会衍生出不一样的版本。越超所有的语言,我相信。 但是不是最近。 这些所有的语言都没有挑战性。新的语言会带来新的设计模式。
消息类型:
1. Twist - 线速度角速度
通常被用于发送到/cmd_vel话题,被base controller节点监听,控制机器人运动
geometry_msgs/Twist
geometry_msgs/Vector3 linear
float64 x
float64 y
float64 z
geometry_msgs/Vector3 angular
float64 x
float64 y
float64 z
linear.x指向机器人前方,linear.y指向左方,linear.z垂直向上满足右手系,平面移动机器人常常linear.y和linear.z均为0
angular.z代表平面机器人的角速度,因为此时z轴为旋转轴
示例:
#! /usr/bin/env python'''Author: xushangnjlh at gmail dot com
Date: 2017/05/22
@package forward_and_back'''import rospyfrom geometry_msgs.msg import Twistfrom math import piclass ForwardAndBack(): def __init__(self):
rospy.init_node('forward_and_back', anonymous=False)
rospy.on_shutdown(self.shutdown) # this "forward_and_back" node will publish Twist type msgs to /cmd_vel
# topic, where this node act like a Publisher
self.cmd_vel = rospy.Publisher('/cmd_vel', Twist, queue_size=5)
# parameters
rate = 50
r = rospy.Rate(rate)
linear_speed = 0.2
goal_distance = 5
linear_duration = goal_distance/linear_speed
angular_speed = 1.0
goal_angular = pi
angular_duration = goal_angular/angular_speed
# forward->rotate->back->rotate
for i in range(2): # this is the msgs variant, has Twist type, no data now
move_cmd = Twist()
move_cmd.linear.x = linear_speed #
# should keep publishing
ticks = int(linear_duration*rate) for t in range(ticks): # one node can publish msgs to different topics, here only publish
# to /cmd_vel self.cmd_vel.publish(move_cmd)
r.sleep # sleep according to the rate
# stop 1 ms before ratate
move_cmd = Twist()
self.cmd_vel.publish(move_cmd)
rospy.sleep(1)
move_cmd.angular_speed.z = angular_speed
ticks = int(angular_duration*rate) for t in range(ticks):
self.cmd_vel.publish(move_cmd)
r.sleep()
self.cmd_vel.publish(Twist())
rospy.sleep(1)
def shutdown(self):
rospy.loginfo("Stopping the robot")
self.cmd_vel.publish(Twist())
rospy.sleep(1)
if __name__=='__main__': try:
ForwardAndBack() except:
rospy.loginfo("forward_and_back node terminated by exception")
2. nav_msgs/Odometry - 里程计(位姿+线速度角速度+各自协方差)
通常,发布到/cmd_vel topic然后被机器人(例如/base_controller节点)监听到的Twist消息是不可能准确地转换为实际的机器人运动路径的,误差来源于机器人内部传感器误差(编码器),标定精度,以及环境影响(地面是否打滑平整);因此我们可以用更准确的Odometry消息类型类获取机器人位姿(/odom到/base_link的tf变换)。在标定准确时,该消息类型与真实的机器人位姿误差大致在1%量级(即使在rviz中显示的依然误差较大)。
还包括
参考系信息,Odometry使用/odom作为parent frame id,/base_link作为child frame id;也就是说世界参考系为/odom(通常设定为起始位姿,固定不动),移动参考系为/base_link(这里还有点不理解,后面来填坑)
时间戳,因此不仅知道运动轨迹,还可以知道对应时间点
Header headerstring child_frame_id
geometry_msgs/PoseWithCovariance pose
geometry_msgs/TwistWithCovariance twist
# 展开Header header
uint32 seq
time stamp
string frame_id
string child_frame_id
geometry_msgs/PoseWithCovariance pose
geometry_msgs/Pose pose
geometry_msgs/Point position
float64 x
float64 y
float64 z
geometry_msgs/Quaternion orientation
float64 x
float64 y
float64 z
float64 w
float64[36] covariance
geometry_msgs/TwistWithCovariance twist
geometry_msgs/Twist twist
geometry_msgs/Vector3 linear
float64 x
float64 y
float64 z
geometry_msgs/Vector3 angular
float64 x
float64 y
float64 z
float64[36] covariance
示例:
运动路径和位姿通过内部的Odometry获取,该Odemetry的位姿通过监听tf坐标系变换获取(/odom和/base_link)
#! /usr/bin/env python'''Author: xushangnjlh at gmail dot comDate: 2017/05/23
@package odometry_forward_and_back'''import rospyfrom geometry_msgs.msg import Twist, Point, Quaternionimport tffrom rbx1_nav.tranform_utils import quat_to_angle, normalize_anglefrom math import pi, radians, copysign, sqrt, powclass Odometry_forward_and_back: def __init__(self):
rospy.ini_node('odometry_forward_and_back', anonymous=False)
rospy.on_shutdown(self.shutdown)
self.cmd_vel = rospy.Publisher('/cmd_vel', Twist, queue_size=5)
rate = 20
r = rospy.Rate(rate)
linear_speed = 0.2
goal_distance =1.0
angular_speed = 1.0
goal_angle = pi
angular_tolerance = radians(2.5)
# Initialize tf listener, and give some time to fill its buffer
self.tf_listener = tf.TransformListener()
rospy.sleep(2)
# Set odom_frame and base_frame
self.odom_frame = '/odom'
try:
self.tf_listener.waitForTransform(self.odom_frame,
'/base_footprint',
rospy.Time(),
rospy.Duration(1.0))
self.base_frame = '/base_footprint'
except(tf.Exception, tf.ConnectivityException, tf.LookupException): try:
self.tf_listener.waitForTransform(self.odom_frame,
'/base_link',
rospy.Time(),
rospy.Duration(1.0))
self.base_frame = '/base_link'
except(tf.Exception, tf.ConnectivityException, tf.LookupException):
rospy.loginfo("Cannot find base_frame transformed from /odom")
rospy.signal_shutdown("tf Exception")
position = Point()
for i in range(2):
move_cmd = Twist()
move_cmd.linear.x = linear_speed # Initial pose, obtained from internal odometry
(position, rotation) = self.get_odom()
x_start = position.x
y_start = position.y
distance = 0
# Keep publishing Twist msgs, until the internal odometry reach the goal
while distance <goal_distance and not rospy.is_shutdown():
self.cmd_vel.publish(move_cmd)
r.sleep()
(position, rotation) = self.get_odom()
distance = sqrt( pow( (position.x-x_start), 2 ) + \
pow( (position.y-y_start), 2 ) )
# Stop 1 ms before rotate
move_cmd = Twist()
self.cmd_vel.publish(move_cmd)
rospy.sleep(1)
move_cmd.angular.z = angular_speed # should be the current ration from odom
angle_last = rotation
angle_turn = 0 while abs(angle_turn+angular_tolerance) <abs(goal_angle) \ and not rospy.is_shutdown():
self.cmd_vel.publish(move_cmd)
r.sleep()
(position, rotation) = self.get_odom
delta_angle = normalize_angle(rotation - angle_last)
angle_turn += delta_angle
angle_last = rotation
move_cmd = Twist()
self.cmd_vel.publish(move_cmd)
rospy.sleep(1)
self.cmd_vel.publish(Twist())
def get_dom(self): try:
(trans, rot) = self.tf_listener.lookupTransfrom(self.odom_frame,
self.base_frame,
rospy.Time(0)) except(tf.Exception, tf.ConnectivityException, tf.LookupException):
rospy.loginfo("TF exception, cannot get_dom()!") return
# angle is only yaw : RPY()[2]
return (Point(*trans), quat_to_angle(*rot))
def shutdown(self):
rospy.loginfo("Stopping the robot...")
self.cmd_vel.publish(Twist(0))
rospy.sleep(1)
if __name__=='__main__': try:
Odometry_forward_and_back() except:
rospy.loginfo("Odometry_forward_and_back node terminated!")
注意这里存在tf操作:
self.tf_listener = tf.TransformListener()rospy.sleep(2)
创建TransformListener对象监听坐标系变换,这里需要sleep 2ms用于tf缓冲。
可以通过以下API获取tf变换,保存在TransformListener对象中,通过lookupTransform获取:
# TransformListener.waitForTransform('ref_coordinate', 'moving_coordinate', rospy.Time(), rospy.Duration(1.0))self.tf_listener.waitForTransform(self.odom_frame, '/base_footprint', rospy.Time(), rospy.Duration(1.0))(trans, rot) = self.tf_listener.lookupTransform(self.odom_frame, self.base_frame, rospy.Time(0))
你写了那么多, 其实就是
# dict为内建函数, 不建议作为变量名def f(d):
return max(d.values())
这里返回的只是年份的最大值