在OpenCV-3.4.9的环境中编译 Prometheus 和prometheus_px4（填坑实录）

从 opencv-3.2.0 到 opencv-3.4.9 的代码更改

0. 相关依赖安装

ROS（省略）、mavros包、建图模块、激光SLAM依赖项、NLopt、OMPL等

### mavros
sudo apt-get install ros-melodic-mavros ros-melodic-mavros-extras
wget https://raw.githubusercontent.com/mavlink/mavros/master/mavros/scripts/install_geographiclib_datasets.sh
sudo chmod 777 ./install_geographiclib_datasets.sh
sudo ./install_geographiclib_datasets.sh
### 建图模块 RTAB-map, OCTO-map
sudo apt-get install ros-melodic-rtabmap*
sudo apt-get install ros-melodic-octomap-*
### 激光SLAM依赖项 cartographer
sudo apt-get install ros-melodic-cartographer*
### 优化相关模块
git clone git://github.com/stevengj/nlopt  
cd nlopt  
mkdir build  
cd build  
cmake -D CMAKE_BUILD_TYPE=release \
      -D CMAKE_INSTALL_PREFIX=/usr/local/nlopt \
	  ..
make  
sudo make install
### 规划相关模块
sudo apt install ros-melodic-ompl*
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22

2. 编译 complie_detection.sh 中的问题

1.1 问题1

~/Prometheus/Modules/object_detection/src/ellipse_lib/spire_ellipsedetector.cpp:1182:25: error: conversion from ‘cv::Mat’ to non-scalar type ‘CvMat’ requested
  CvMat temp = src;
1
2

更改代码：

	CvMat c_src = src;
    CvMat c_dst = _edges.getMat();
    CvMat c_dx = _sobel_x.getMat();
    CvMat c_dy = _sobel_y.getMat();
1
2
3
4

为新的标准方式：

	CvMat c_src = cvMat(src);
	CvMat c_dst = cvMat(_edges.getMat());
	CvMat c_dx = cvMat(_sobel_x.getMat());
	CvMat c_dy = cvMat(_sobel_y.getMat());
1
2
3
4

相应的，如果有 IplImage Ipi_z = z 则需要更改为 IplImage Ipi_z = cvIplImage(z)。

1.2 问题2

~/Prometheus/Modules/object_detection/src/darknet_ros/src/YoloObjectDetector.cpp:579:49: error: no matching function for call to ‘_IplImage::_IplImage(cv::Mat&)’
   IplImage* ROS_img = new IplImage(camImageCopy_);       
1
2

更改代码：

  IplImage* ROS_img = new IplImage(camImageCopy_);
1

为新的标准方式：

  // IplImage* ROS_img = &cvIplImage(camImageCopy_);   	// ERROR: taking address of temporary [-fpermissive]
  IplImage tmp_img = cvIplImage(camImageCopy_);
  IplImage* ROS_img = &tmp_img;
1
2
3

1.3 问题3

~/Prometheus/Modules/planning/FastPlanner/plan_env/ThirdParty/sdf_tools/src/sdf_tools/sdf_builder.cpp: In member function ‘bool sdf_tools::SDF_Builder::BuildInternalPlanningScene()’:
~/Prometheus/Modules/planning/FastPlanner/plan_env/ThirdParty/sdf_tools/src/sdf_tools/sdf_builder.cpp:87:134: error: no matching function for call to ‘planning_scene::PlanningScene::PlanningScene(boost::shared_ptr<urdf::Model>&, boost::shared_ptr<srdf::Model>&)’
         planning_scene_ptr_ = std::shared_ptr<planning_scene::PlanningScene>(new planning_scene::PlanningScene(urdf_model, srdf_model));
                                                                                                                                      ^
1
2
3
4

解决办法，更改代码：

        boost::shared_ptr<urdf::Model> urdf_model(new urdf::Model());
        urdf_model->initString(GenerateSDFComputeBotURDFString());
        // Make the SRDF model
        boost::shared_ptr<srdf::Model> srdf_model(new srdf::Model());
        srdf_model->initString(*urdf_model, GenerateSDFComputeBotSRDFString());
        // Make the planning scene
        planning_scene_ptr_.reset();
        planning_scene_ptr_ = std::shared_ptr<planning_scene::PlanningScene>(new planning_scene::PlanningScene(urdf_model, srdf_model));
1
2
3
4
5
6
7
8

为新的标准方式：

    	auto urdf_model = std::make_shared<urdf::Model>();
    	urdf_model->initString(GenerateSDFComputeBotURDFString());
    	// Make the SRDF model
    	auto srdf_model = std::make_shared<srdf::Model>();
    	srdf_model->initString(*urdf_model, GenerateSDFComputeBotSRDFString());
		// Make the planning scene
		planning_scene_ptr_ = std::make_shared<planning_scene::PlanningScene>(urdf_model, srdf_model);
1
2
3
4
5
6
7

总结1～3问题：

（1）由于OpenCV的版本从 opencv2 到 opencv3 再到 opencv4 的升级过程中，对于 cv::Mat, CvMat 和 IplImage 之间的相互转换关系操作一直在变化中。源码编译的时候，需要根据实际的编译环境 opencv 的版本来调整相应的代码。
（2）由于C++语言的标准一直变化中，一些与之相关的代码也可能需要调整。例如关于指针尤其是指针模板的定义和使用。

1.4 问题4

又再次遇到奇葩的问题。

~/Prometheus/Modules/swarm_control/src/ego_traj_to_cmd.cpp:21:10: fatal error: quadrotor_msgs/PositionCommand.h: No such file or directory
 #include "quadrotor_msgs/PositionCommand.h"
          ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

1
2
3
4

网络上的方式是重新编译相关的功能包。但是，对于我来讲还是非常的迷惑！
https://www.freesion.com/article/76371450240/

cd ~/fast-planer_ws  //fast-planer_ws为我存放该工程的工作空间
catkin_make
source devel/setup.bash 
1
2
3

重新在另外一个网页上找到了看起来比较靠谱的办法。
https://www.freesion.com/article/16841452429/

step 1：在CMakeList.txt文件中，find_package() 里添加quadrotor_msgs如下

find_package( #...
  quadrotor_msgs # 添加的
)
1
2
3

step 2：同样在此CMakeList.txt中，include_directories里添加${catkin_INCLUDE_DIRS}如下：

include_directories(include ${catkin_INCLUDE_DIRS})
1

在更改这个 CMakeList.txt 的时候发现，他的一些依赖与ego_planner 有关，于是事先编译一下 ego_planner.

~/Prometheus$ ./complie_ego.sh
1

然后再次编译 swarm 模块。

~/Prometheus$ ./complie_swarm.sh
1

结果得到一个新的错误提示。

CMake Error at  ~/Prometheus/devel/share/quadrotor_msgs/cmake/quadrotor_msgsConfig.cmake:173 (message):
  Project 'prometheus_swarm_control' tried to find library 'encode_msgs'.
  The library is neither a target nor built/installed properly.  Did you
  compile project 'quadrotor_msgs'? Did you find_package() it before the
  subdirectory containing its code is included?
1
2
3
4
5

很迷的一个错误提示。然后再次编译 ego 模块。
也可能是期间，同时在安装的 ompl 功能模块起作用了。

sudo apt install ros-melodic-ompl*
1

不管他，继续了。

~/Prometheus$ ./complie_ego.sh
1

出现了一个错误提示如下。

[ 86%] Building CXX object ego-planner-swarm/src/uav_simulator/Utils/rviz_plugins/CMakeFiles/rviz_plugins.dir/src/probmap_display.cpp.o
[ 87%] Building CXX object ego-planner-swarm/src/uav_simulator/Utils/rviz_plugins/CMakeFiles/rviz_plugins.dir/src/aerialmap_display.cpp.o
ego-planner-swarm/src/planner/traj_utils/CMakeFiles/traj_utils.dir/build.make:94: recipe for target 'ego-planner-swarm/src/planner/traj_utils/CMakeFiles/traj_utils.dir/src/polynomial_traj.cpp.o' failed
make[2]: *** [ego-planner-swarm/src/planner/traj_utils/CMakeFiles/traj_utils.dir/src/polynomial_traj.cpp.o] Error 1
CMakeFiles/Makefile2:5100: recipe for target 'ego-planner-swarm/src/planner/traj_utils/CMakeFiles/traj_utils.dir/all' failed
make[1]: *** [ego-planner-swarm/src/planner/traj_utils/CMakeFiles/traj_utils.dir/all] Error 2
make[1]: *** Waiting for unfinished jobs....
1
2
3
4
5
6
7

暂时忽略它，继续编译 swarm 模块。

~/Prometheus$ ./complie_swarm.sh
1

完美通过。

总结4问题：

此过程中由于 prometheus 编译过程中有相互的关联需求，而且其中还夹杂着因为缺少相应的功能包而出现的一些问题。总体思路是（1）调整编译的顺序；（2）安装相关依赖。

1.5 问题5

使用阿木实验室的Prometheus项目专用的PX4仓库来安装 prometheus_px4，安装方法如下：

git clone https://gitee.com/amovlab/prometheus_px4.git
cd prometheus_px4
git submodule update --init --recursive
pip3 install --user toml empy jinja2 packaging
make amovlab_sitl_default gazebo

cd prometheus_px4/Tools/setup
source ./ubuntu.sh
1
2
3
4
5
6
7
8

其中，在进行到 make amovlab_sitl_default gazebo 的时候可能有报错（实际没有碰到），解决办法：

make clean
make px4_sitl gazebo

sudo apt-get install libprotobuf-dev libprotoc-dev protobuf-compiler libeigen3-dev libxml2-utils python-rospkg python-jinja2
sudo apt-get install libgstreamer-plugins-base1.0-dev gstreamer1.0-plugins-bad gstreamer1.0-plugins-base gstreamer1.0-plugins-good gstreamer1.0-plugins-ugly -y
1
2
3
4
5

在运行 cd prometheus_px4/Tools/setup && source ./ubuntu.sh 的时候可能会遇到找不到 cc1 等库文件的问题。解决办法：
首先，通过gcc -v查看 gcc 的路径。

Using built-in specs.
COLLECT_GCC=gcc
COLLECT_LTO_WRAPPER=/usr/lib/gcc/x86_64-linux-gnu/8/lto-wrapper
OFFLOAD_TARGET_NAMES=nvptx-none
OFFLOAD_TARGET_DEFAULT=1
Target: x86_64-linux-gnu
Configured with: ../src/configure -v --with-pkgversion='Ubuntu 8.4.0-1ubuntu1~18.04' --enable-languages=c,ada,c++,go,brig,d,fortran,objc,obj-c++ --prefix=/usr --with-gcc-major-version-only --program-suffix=-8 --program-prefix=x86_64-linux-gnu- --enable-shared 
Thread model: posix
gcc version 8.4.0 (Ubuntu 8.4.0-1ubuntu1~18.04) 
1
2
3
4
5
6
7
8
9

在 ~/.bash.bashrc 中添加

export LIBRARY_PATH=/usr/lib/gcc/x86_64-linux-gnu/8:$LIBRARY_PATH
1

然后再次运行 source ./ubuntu.sh 即可。
另外，
还需要在 ~/.bash.bashrc 中添加如下指令（以下若存在已添加过的命令，请勿重复添加），其中${your prometheus path}为Prometheus项目路径，${your px4 path}为安装PX4固件即prometheus_px4的路径。

source ${your prometheus path}/Prometheus/devel/setup.bash
export GAZEBO_PLUGIN_PATH=$GAZEBO_PLUGIN_PATH:${your prometheus path}/Prometheus/devel/lib
export GAZEBO_MODEL_PATH=$GAZEBO_MODEL_PATH:${your prometheus path}/Prometheus/Simulator/gazebo_simulator/models
export GAZEBO_MODEL_PATH=$GAZEBO_MODEL_PATH:${your prometheus path}/Prometheus/Simulator/gazebo_simulator/amov_models
source ${your px4 path}/prometheus_px4/Tools/setup_gazebo.bash ${your px4 path}/prometheus_px4 ${your px4 path}/prometheus_px4/build/amovlab_sitl_default
export ROS_PACKAGE_PATH=$ROS_PACKAGE_PATH:${your px4 path}/prometheus_px4
export ROS_PACKAGE_PATH=$ROS_PACKAGE_PATH:${your px4 path}/prometheus_px4/Tools/sitl_gazebo
1
2
3
4
5
6
7

然后可以继续安装 gazebo 中仿真所需的 3Dlidar 和 octomap 以及 Turtlebot3 等插件。

sudo apt-get install ros-melodic-velodyne-gazebo-plugins
sudo apt-get install ros-melodic-octomap-rviz-plugins
sudo apt-get install ros-melodic-turtlebot3-*
1
2
3

Gazebo仿真运行测试：

roslaunch prometheus_gazebo sitl.launch
1

一共启动了三个终端。
第一个终端同时运行了gazebo、Mavros、px4_pos_estimator、px4_pos_controller四个节点，第二个终端则运行了ground_station节点，第三个终端则运行了pos_estimator 节点。因此，若第一个终端无报错，第二个终端显示[Connected]并能够查看到飞机状态，且Gazebo成功运行，代表成功运行。

NODES
  /
    gazebo (gazebo_ros/gzserver)
    gazebo_gui (gazebo_ros/gzclient)
    ground_station (prometheus_station/ground_station)
    ground_station_msg (prometheus_station/ground_station_msg)
    mavros (mavros/mavros_node)
    px4_pos_estimator (prometheus_control/px4_pos_estimator)
    px4_sender (prometheus_control/px4_sender)
    sitl (px4/px4)
    tf_2Dlidar (tf/static_transform_publisher)
    tf_3Dlidar (tf/static_transform_publisher)
    tf_realsense_camera (tf/static_transform_publisher)
    tf_slam2map (tf/static_transform_publisher)
    tf_world_map (tf/static_transform_publisher)
    vehicle_spawn_UX303L_3799_8995328815101938340 (gazebo_ros/spawn_model)

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17

运行如下指令启动终端控制节点，并根据终端提示输入指令：

rosrun prometheus_control terminal_control
1

到这就成功了。也可以在第一个终端中，使用 help 以及 commander help 来学习使用命令操控无人机。
不用重启系统而关闭 gazebo 的命令。

killall gzserver
killall gzclient
1
2