技术标签: navigation ROS costmap_2d ROS Navigation
写在最前:
尊重原创,尊重他人劳动成果。原文地址:https://blog.csdn.net/u013158492/article/details/50485418
在上一篇文章中moveBase就有关于costmap_2d的使用: planner_costmap_ros_是用于全局导航的地图, controller_costmap_ros_是局部导航用的地图,地图类型为经过ROS封装的costmap_2d::Costmap2DROS*。
move_base.cpp中planner_costmap_ros_的代码如下:
//create the ros wrapper for the planner's costmap... and initializer a pointer we'll use with the underlying map
planner_costmap_ros_ = new costmap_2d::Costmap2DROS("global_costmap", tf_);
planner_costmap_ros_->pause();
move_base.cpp中controller_costmap_ros_代码如下:
//create the ros wrapper for the controller's costmap... and initializer a pointer we'll use with the underlying map
controller_costmap_ros_ = new costmap_2d::Costmap2DROS("local_costmap", tf_);
controller_costmap_ros_->pause();
以下是这个ROS类的UML:
这个类的成员变量:LayeredCostmap* layered_costmap_; pluginlib::ClassLoader<Layer> plugin_loader_; 这两个最重要的成员变量,而LayeredCostmap类又包含了Costmap2D costmap_; 这个数据成员。
下面是这些类之间的关系:
绿色的是核心代码,从ROS用户的角度,只需要调用Costmap2DROS这个类,因为这个类已经把所有关于地图的操作都封装好了。不过我这里是分析底层算法实现,就不得不写得很长很长。
所以还是先回到对Costmap2DROS这个类的分析,然后再进一步一层一层的分析其他的类。这些类完成了对机器人地图的表示和操作,因此其数据结构和算法都很有分析的价值。
首先是构造函数Costmap2DROS::Costmap2DROS(std::string name, tf::TransformListener& tf) :
因此必须提供一个tf参数。tf参数需要提供以下两个坐标系的关系:
// get global and robot base frame names
private_nh.param("global_frame", global_frame_, std::string("map"));
private_nh.param("robot_base_frame", robot_base_frame_, std::string("base_link"));
如果没有找到这两个坐标系的关系或者超时,则构造函数会一直阻塞在这里:
// we need to make sure that the transform between the robot base frame and the global frame is available
while (ros::ok()
&& !tf_.canTransform(global_frame_, robot_base_frame_, ros::Time(), ros::Duration(0.1), &tf_error))
{
ros::spinOnce();
if (last_error + ros::Duration(5.0) < ros::Time::now())
{
ROS_WARN("Timed out waiting for transform from %s to %s to become available before running costmap, tf error: %s",
robot_base_frame_.c_str(), global_frame_.c_str(), tf_error.c_str());
last_error = ros::Time::now();
}
// The error string will accumulate and errors will typically be the same, so the last
// will do for the warning above. Reset the string here to avoid accumulation.
tf_error.clear();
}
然后加入各个层次的地图:
if (private_nh.hasParam("plugins"))
{
XmlRpc::XmlRpcValue my_list;
private_nh.getParam("plugins", my_list);
for (int32_t i = 0; i < my_list.size(); ++i)
{
std::string pname = static_cast<std::string>(my_list[i]["name"]);
std::string type = static_cast<std::string>(my_list[i]["type"]);
ROS_INFO("%s: Using plugin \"%s\"", name_.c_str(), pname.c_str());
copyParentParameters(pname, type, private_nh);
boost::shared_ptr<Layer> plugin = plugin_loader_.createInstance(type);
layered_costmap_->addPlugin(plugin);
plugin->initialize(layered_costmap_, name + "/" + pname, &tf_);
}
}
boost::shared_ptr<Layer> plugin = plugin_loader_.createInstance(type);
这行会创建一个以 type为类类型的实例变量,然后让plugin这个指针指向这个实例。
layered_costmap_->addPlugin(plugin);
然后 layered_costmap_ 将这些类型的地图都加入,addPlugin
实现:
void addPlugin(boost::shared_ptr<Layer> plugin)
{
plugins_.push_back(plugin);
}
这里的关系是:Costmap2DROS 有一个layered_costmap_ 数据成员,然后layered_costmap_ 又有一个std::vector<boost::shared_ptr<Layer> > plugins_; 成员,因此可以将各个子类的实例化对象的指针交给父类Layer 指针plugins_ 管理。
plugin->initialize(layered_costmap_, name + "/" + pname, &tf_);
这行将会对实例初始化,实际执行是plugin调用的父类Layer的方法void Layer::initialize(LayeredCostmap* parent, std::string name, tf::TransformListener *tf) 。
实际上父类Layer有一个成员变量为LayeredCostmap* layered_costmap_; 的指针,因此通过LayeredCostmap* layered_costmap_;指针指向了具体的子类,比如ObstacleLayer StaticLayer InflationLayer 等。
然后设置footprint:footprint_sub_ = private_nh.subscribe(topic, 1, &Costmap2DROS::setUnpaddedRobotFootprintPolygon, this); ,回调函数setUnpaddedRobotFootprintPolygon 实际调用的是成员函数:
void Costmap2DROS::setUnpaddedRobotFootprint(const std::vector<geometry_msgs::Point>& points)
{
unpadded_footprint_ = points;
padded_footprint_ = points;
padFootprint(padded_footprint_, footprint_padding_);
layered_costmap_->setFootprint(padded_footprint_);
}
以下是在footprint.cpp中的定义:
void padFootprint(std::vector<geometry_msgs::Point>& footprint, double padding)
{
// pad footprint in place
for (unsigned int i = 0; i < footprint.size(); i++)
{
geometry_msgs::Point& pt = footprint[ i ];
pt.x += sign0(pt.x) * padding;
pt.y += sign0(pt.y) * padding;
}
}
然后声明了一个timer,定时检测机器人是否在移动:
// Create a time r to check if the robot is moving
robot_stopped_ = false;
timer_ = private_nh.createTimer(ros::Duration(.1), &Costmap2DROS::movementCB, this);
这里回调函数movementCB
实现,是通过比较前后两个pose的差,判断机器人是否在移动:
void Costmap2DROS::movementCB(const ros::TimerEvent &event)
{
// don't allow configuration to happen while this check occurs
// boost::recursive_mutex::scoped_lock mcl(configuration_mutex_);
geometry_msgs::PoseStamped new_pose;
if (!getRobotPose(new_pose))
{
ROS_WARN_THROTTLE(1.0, "Could not get robot pose, cancelling reconfiguration");
robot_stopped_ = false;
}
// make sure that the robot is not moving
else
{
old_pose_ = new_pose;
robot_stopped_ = (tf2::Vector3(old_pose_.pose.position.x, old_pose_.pose.position.y,
old_pose_.pose.position.z).distance(tf2::Vector3(new_pose.pose.position.x,
new_pose.pose.position.y, new_pose.pose.position.z)) < 1e-3) &&
(tf2::Quaternion(old_pose_.pose.orientation.x,
old_pose_.pose.orientation.y,
old_pose_.pose.orientation.z,
old_pose_.pose.orientation.w).angle(tf2::Quaternion(new_pose.pose.orientation.x,
new_pose.pose.orientation.y,
new_pose.pose.orientation.z,
new_pose.pose.orientation.w)) < 1e-3);
}
}
在构造函数末尾,开启参数动态配置:
dsrv_ = new dynamic_reconfigure::Server<Costmap2DConfig>(ros::NodeHandle("~/" + name));
dynamic_reconfigure::Server<Costmap2DConfig>::CallbackType cb = boost::bind(&Costmap2DROS::reconfigureCB, this, _1,
_2);
dsrv_->setCallback(cb);
回调函数reconfigureCB 除了对一些类成员的配置值做赋值以外,还会开启一个更新map的线程
void Costmap2DROS::reconfigureCB(costmap_2d::Costmap2DConfig &config, uint32_t level)
{
transform_tolerance_ = config.transform_tolerance;
if (map_update_thread_ != NULL)
{
map_update_thread_shutdown_ = true;
map_update_thread_->join();
delete map_update_thread_;
}
map_update_thread_shutdown_ = false;
double map_update_frequency = config.update_frequency;
double map_publish_frequency = config.publish_frequency;
if (map_publish_frequency > 0)
publish_cycle = ros::Duration(1 / map_publish_frequency);
else
publish_cycle = ros::Duration(-1);
// find size parameters
double map_width_meters = config.width, map_height_meters = config.height, resolution = config.resolution, origin_x =
config.origin_x,
origin_y = config.origin_y;
if (!layered_costmap_->isSizeLocked())
{
layered_costmap_->resizeMap((unsigned int)(map_width_meters / resolution),
(unsigned int)(map_height_meters / resolution), resolution, origin_x, origin_y);
}
// If the padding has changed, call setUnpaddedRobotFootprint() to
// re-apply the padding.
if (footprint_padding_ != config.footprint_padding)
{
footprint_padding_ = config.footprint_padding;
setUnpaddedRobotFootprint(unpadded_footprint_);
}
readFootprintFromConfig(config, old_config_);
old_config_ = config;
map_update_thread_ = new boost::thread(boost::bind(&Costmap2DROS::mapUpdateLoop, this, map_update_frequency));
}
mapUpdateLoop() 的线程函数定义:
void Costmap2DROS::mapUpdateLoop(double frequency)
{
// the user might not want to run the loop every cycle
if (frequency == 0.0)
return;
ros::NodeHandle nh;
ros::Rate r(frequency);
while (nh.ok() && !map_update_thread_shutdown_)
{
struct timeval start, end;
double start_t, end_t, t_diff;
gettimeofday(&start, NULL);
updateMap();
gettimeofday(&end, NULL);
start_t = start.tv_sec + double(start.tv_usec) / 1e6;
end_t = end.tv_sec + double(end.tv_usec) / 1e6;
t_diff = end_t - start_t;
ROS_DEBUG("Map update time: %.9f", t_diff);
if (publish_cycle.toSec() > 0 && layered_costmap_->isInitialized())
{
unsigned int x0, y0, xn, yn;
layered_costmap_->getBounds(&x0, &xn, &y0, &yn);
publisher_->updateBounds(x0, xn, y0, yn);
ros::Time now = ros::Time::now();
if (last_publish_ + publish_cycle < now)
{
publisher_->publishCostmap();
last_publish_ = now;
}
}
r.sleep();
// make sure to sleep for the remainder of our cycle time
if (r.cycleTime() > ros::Duration(1 / frequency))
ROS_WARN("Map update loop missed its desired rate of %.4fHz... the loop actually took %.4f seconds", frequency,
r.cycleTime().toSec());
}
}
核心功能在于调用updateMap();
void Costmap2DROS::updateMap()
{
if (!stop_updates_)
{
// get global pose
geometry_msgs::PoseStamped pose;
if (getRobotPose (pose))
{
double x = pose.pose.position.x,
y = pose.pose.position.y,
yaw = tf2::getYaw(pose.pose.orientation);
layered_costmap_->updateMap(x, y, yaw);
geometry_msgs::PolygonStamped footprint;
footprint.header.frame_id = global_frame_;
footprint.header.stamp = ros::Time::now();
transformFootprint(x, y, yaw, padded_footprint_, footprint);
footprint_pub_.publish(footprint);
initialized_ = true;
}
}
}
函数layered_costmap_->updateMap(x, y, yaw);
定义
void LayeredCostmap::updateMap(double robot_x, double robot_y, double robot_yaw)
{
// if we're using a rolling buffer costmap... we need to update the origin using the robot's position
if (rolling_window_)
{
double new_origin_x = robot_x - costmap_.getSizeInMetersX() / 2;
double new_origin_y = robot_y - costmap_.getSizeInMetersY() / 2;
costmap_.updateOrigin(new_origin_x, new_origin_y);
}
if (plugins_.size() == 0)
return;
minx_ = miny_ = 1e30;
maxx_ = maxy_ = -1e30;
for (vector<boost::shared_ptr<Layer> >::iterator plugin = plugins_.begin(); plugin != plugins_.end();
++plugin)
{
(*plugin)->updateBounds(robot_x, robot_y, robot_yaw, &minx_, &miny_, &maxx_, &maxy_);
}
int x0, xn, y0, yn;
costmap_.worldToMapEnforceBounds(minx_, miny_, x0, y0);
costmap_.worldToMapEnforceBounds(maxx_, maxy_, xn, yn);
x0 = std::max(0, x0);
xn = std::min(int(costmap_.getSizeInCellsX()), xn + 1);
y0 = std::max(0, y0);
yn = std::min(int(costmap_.getSizeInCellsY()), yn + 1);
ROS_DEBUG("Updating area x: [%d, %d] y: [%d, %d]", x0, xn, y0, yn);
if (xn < x0 || yn < y0)
return;
{
// Clear and update costmap under a single lock
boost::unique_lock<Costmap2D::mutex_t> lock(*(costmap_.getMutex()));
costmap_.resetMap(x0, y0, xn, yn);
for (vector<boost::shared_ptr<Layer> >::iterator plugin = plugins_.begin(); plugin != plugins_.end();
++plugin)
{
(*plugin)->updateCosts(costmap_, x0, y0, xn, yn);
}
}
bx0_ = x0;
bxn_ = xn;
by0_ = y0;
byn_ = yn;
initialized_ = true;
}
updateMap
分为两个阶段,第一个阶段是UpdateBounds
:这个阶段会更新每个Layer的更新区域,这样在每个运行周期内减少了数据拷贝的操作时间。
(*plugin)->updateBounds(robot_x, robot_y, robot_yaw, &minx_, &miny_, &maxx_, &maxy_);
第二个阶段是 ` UpdateCosts :
for (vector<boost::shared_ptr<Layer> >::iterator plugin = plugins_.begin(); plugin != plugins_.end();
++plugin)
{
(*plugin)->updateCosts(costmap_, x0, y0, xn, yn);
}
这个阶段将逐一拷贝数据到Master Map,关于Master Map是如何得到的,见下图,图来源于David Lu的Paper《Layered Costmaps for Context-Sensitive Navigation》:
void Costmap2DROS::start(),
这里通过成员变量layered_costmap_
拿到类LayeredCostmap
的数据成员std::vector<boost::shared_ptr<Layer> > plugins_;
,然后调用没个Layer
的子类的方法(*plugin)->activate();
void Costmap2DROS::start()
{
std::vector < boost::shared_ptr<Layer> > *plugins = layered_costmap_->getPlugins();
// check if we're stopped or just paused
if (stopped_)
{
// if we're stopped we need to re-subscribe to topics
for (vector<boost::shared_ptr<Layer> >::iterator plugin = plugins->begin(); plugin != plugins->end();
++plugin)
{
(*plugin)->activate();
}
stopped_ = false;
}
stop_updates_ = false;
// block until the costmap is re-initialized.. meaning one update cycle has run
ros::Rate r(100.0);
while (ros::ok() && !initialized_)
r.sleep();
}
函数
void Costmap2DROS::stop()
void Costmap2DROS::stop()
{
stop_updates_ = true;
std::vector < boost::shared_ptr<Layer> > *plugins = layered_costmap_->getPlugins();
// unsubscribe from topics
for (vector<boost::shared_ptr<Layer> >::iterator plugin = plugins->begin(); plugin != plugins->end();
++plugin)
{
(*plugin)->deactivate();
}
initialized_ = false;
stopped_ = true;
}
函数 void Costmap2DROS::resetLayers()
void Costmap2DROS::resetLayers()
{
Costmap2D* top = layered_costmap_->getCostmap();
top->resetMap(0, 0, top->getSizeInCellsX(), top->getSizeInCellsY());
std::vector < boost::shared_ptr<Layer> > *plugins = layered_costmap_->getPlugins();
for (vector<boost::shared_ptr<Layer> >::iterator plugin = plugins->begin(); plugin != plugins->end();
++plugin)
{
(*plugin)->reset();
}
}
函数 bool Costmap2DROS::getRobotPose, 这里只需要指定global_pose
和 robot_pose
各自的frame_id_
就可以通过tf_.transformPose(global_frame_, robot_pose, global_pose);
获得机器人的 global_pose
。
bool Costmap2DROS::getRobotPose(geometry_msgs::PoseStamped& global_pose) const
{
tf2::toMsg(tf2::Transform::getIdentity(), global_pose.pose);
geometry_msgs::PoseStamped robot_pose;
tf2::toMsg(tf2::Transform::getIdentity(), robot_pose.pose);
robot_pose.header.frame_id = robot_base_frame_;
robot_pose.header.stamp = ros::Time();
ros::Time current_time = ros::Time::now(); // save time for checking tf delay later
// get the global pose of the robot
try
{
tf_.transform(robot_pose, global_pose, global_frame_);
}
catch (tf2::LookupException& ex)
{
ROS_ERROR_THROTTLE(1.0, "No Transform available Error looking up robot pose: %s\n", ex.what());
return false;
}
catch (tf2::ConnectivityException& ex)
{
ROS_ERROR_THROTTLE(1.0, "Connectivity Error looking up robot pose: %s\n", ex.what());
return false;
}
catch (tf2::ExtrapolationException& ex)
{
ROS_ERROR_THROTTLE(1.0, "Extrapolation Error looking up robot pose: %s\n", ex.what());
return false;
}
// check global_pose timeout
if (current_time.toSec() - global_pose.header.stamp.toSec() > transform_tolerance_)
{
ROS_WARN_THROTTLE(1.0,
"Costmap2DROS transform timeout. Current time: %.4f, global_pose stamp: %.4f, tolerance: %.4f",
current_time.toSec(), global_pose.header.stamp.toSec(), transform_tolerance_);
return false;
}
return true;
}
函数 void Costmap2DROS::getOrientedFootprint 完成将机器人坐标系下的机器人轮廓点的坐标转化为机器人在当前全局坐标系下的轮廓点的值。具体定义如下:
void Costmap2DROS::getOrientedFootprint(std::vector<geometry_msgs::Point>& oriented_footprint) const
{
geometry_msgs::PoseStamped global_pose;
if (!getRobotPose(global_pose))
return;
double yaw = tf2::getYaw(global_pose.pose.orientation);
transformFootprint(global_pose.pose.position.x, global_pose.pose.position.y, yaw,
padded_footprint_, oriented_footprint);
}
到此,基本上 Costmap2DROS
的定义就这么多了。不过其中类和类之间的调用关系依然还是很复杂,因此需要需要分析plugin原理,才能真正知道这些类的关系是如何实现的。
文章浏览阅读1.6k次。安装配置gi、安装数据库软件、dbca建库见下:http://blog.csdn.net/kadwf123/article/details/784299611、检查集群节点及状态:[root@rac2 ~]# olsnodes -srac1 Activerac2 Activerac3 Activerac4 Active[root@rac2 ~]_12c查看crs状态
文章浏览阅读1.3w次,点赞45次,收藏99次。我个人用的是anaconda3的一个python集成环境,自带jupyter notebook,但在我打开jupyter notebook界面后,却找不到对应的虚拟环境,原来是jupyter notebook只是通用于下载anaconda时自带的环境,其他环境要想使用必须手动下载一些库:1.首先进入到自己创建的虚拟环境(pytorch是虚拟环境的名字)activate pytorch2.在该环境下下载这个库conda install ipykernelconda install nb__jupyter没有pytorch环境
文章浏览阅读5.2k次,点赞19次,收藏28次。选择scoop纯属意外,也是无奈,因为电脑用户被锁了管理员权限,所有exe安装程序都无法安装,只可以用绿色软件,最后被我发现scoop,省去了到处下载XXX绿色版的烦恼,当然scoop里需要管理员权限的软件也跟我无缘了(譬如everything)。推荐添加dorado这个bucket镜像,里面很多中文软件,但是部分国外的软件下载地址在github,可能无法下载。以上两个是官方bucket的国内镜像,所有软件建议优先从这里下载。上面可以看到很多bucket以及软件数。如果官网登陆不了可以试一下以下方式。_scoop-cn
文章浏览阅读4.5k次,点赞2次,收藏3次。首先要有一个color-picker组件 <el-color-picker v-model="headcolor"></el-color-picker>在data里面data() { return {headcolor: ’ #278add ’ //这里可以选择一个默认的颜色} }然后在你想要改变颜色的地方用v-bind绑定就好了,例如:这里的:sty..._vue el-color-picker
文章浏览阅读640次。基于芯片日益增长的问题,所以内核开发者们引入了新的方法,就是在内核中只保留函数,而数据则不包含,由用户(应用程序员)自己把数据按照规定的格式编写,并放在约定的地方,为了不占用过多的内存,还要求数据以根精简的方式编写。boot启动时,传参给内核,告诉内核设备树文件和kernel的位置,内核启动时根据地址去找到设备树文件,再利用专用的编译器去反编译dtb文件,将dtb还原成数据结构,以供驱动的函数去调用。firmware是三星的一个固件的设备信息,因为找不到固件,所以内核启动不成功。_exynos 4412 刷机
文章浏览阅读2w次,点赞24次,收藏42次。Linux系统配置jdkLinux学习教程,Linux入门教程(超详细)_linux配置jdk
文章浏览阅读3.3k次,点赞5次,收藏19次。xlabel('\delta');ylabel('AUC');具体符号的对照表参照下图:_matlab微米怎么输入
文章浏览阅读119次。顺序读写指的是按照文件中数据的顺序进行读取或写入。对于文本文件,可以使用fgets、fputs、fscanf、fprintf等函数进行顺序读写。在C语言中,对文件的操作通常涉及文件的打开、读写以及关闭。文件的打开使用fopen函数,而关闭则使用fclose函数。在C语言中,可以使用fread和fwrite函数进行二进制读写。 Biaoge 于2024-03-09 23:51发布 阅读量:7 ️文章类型:【 C语言程序设计 】在C语言中,用于打开文件的函数是____,用于关闭文件的函数是____。
文章浏览阅读3.4k次,点赞2次,收藏13次。跟随鼠标移动的粒子以grid(SOP)为partical(SOP)的资源模板,调整后连接【Geo组合+point spirit(MAT)】,在连接【feedback组合】适当调整。影响粒子动态的节点【metaball(SOP)+force(SOP)】添加mouse in(CHOP)鼠标位置到metaball的坐标,实现鼠标影响。..._touchdesigner怎么让一个模型跟着鼠标移动
文章浏览阅读178次。项目运行环境配置:Jdk1.8 + Tomcat7.0 + Mysql + HBuilderX(Webstorm也行)+ Eclispe(IntelliJ IDEA,Eclispe,MyEclispe,Sts都支持)。项目技术:Springboot + mybatis + Maven +mysql5.7或8.0+html+css+js等等组成,B/S模式 + Maven管理等等。环境需要1.运行环境:最好是java jdk 1.8,我们在这个平台上运行的。其他版本理论上也可以。_基于java技术的停车场管理系统实现与设计
文章浏览阅读3.5k次。前言对于MediaPlayer播放器的源码分析内容相对来说比较多,会从Java-&amp;gt;Jni-&amp;gt;C/C++慢慢分析,后面会慢慢更新。另外,博客只作为自己学习记录的一种方式,对于其他的不过多的评论。MediaPlayerDemopublic class MainActivity extends AppCompatActivity implements SurfaceHolder.Cal..._android多媒体播放源码分析 时序图
文章浏览阅读2.4k次,点赞41次,收藏13次。java 数据结构与算法 ——快速排序法_快速排序法