Dear author, thank you very much for your open source work.
I can't understand why you use static data to test GNSS-RTK. Why not use dynamic data testing?
For GNSS-PPP, the positioning accuracy is poor, and using static data to test the positioning accuracy seems to be a correct choice.
In addition, I would like to ask, if I want to test the RTK data set I collected, where the code needs to be modified, I only see the need to modify rtklib.h file, I hope you can answer my doubts, thank you!

Can you test your code just by modifying the coordinates of the reference station?

Professor Wen, I cloned your code today and collected static gnss data under open and good environment.

Finally, the localization results of fgo and rtklib were compared, and the conclusion was drawn that the positioning accuracy of fgo is much worse than that of rtklib.
rtklib's data can be well stabilized around a value (with an error of centimeters), while fgo's data often experiences sudden drift (sometimes with an error of 2-3 meters).
May I ask this conclusion is correct? Is it possible to achieve centimeter level accuracy in RTK localization using GraphgnssLib.

Wen博士， 对于psr_doppler_car_rtk_dyna.h中的addDopplerFactors函数, 每个历元都掉用了 dopplerConstraint::create 工厂函数，在这个函数里一直在重复参数块的清除和循环重新绑定，直接添加最新的状态不可以吗，有什么特殊的逻辑？感谢回答

Dear Welson
I have a question for you, when i run the RTK-FGO Demo, as fellow:

source ~/GraphGNSSLib/devel/setup.bash
roslaunch global_fusion dataublox_TST20200603.launch
roslaunch global_fusion psr_doppler_car_rtk.launch

I find that the positioning error is larger than the results mentioned in your paper.
The maximum positioning error in ENU direction is 3.2 meters。
So, i want to know why? It is related to the computer? or other reasons?
Thank you very much !

If possible, I would like to add your WeChat. My WeChat account is j15212774116.

Hi, Professor @weisongwen there is a bug in the WeightedLeastSquare function in the gnss_tools.h file.
That is, you use 0 and 1 to represent whether the satellite is Beidou or GPS in the third and fourth columns of the eH_Matrix

However, if all satellites are Beidou or GPS in this case, then during this step of calculation
eDeltaPos = (eH_Matrix.transpose() * weight_matrix * eH_Matrix).inverse() * eH_Matrix.transpose() * weight_matrix * eDeltaPr;

eH_Matrix.transpose() * weight_matrix * eH_Matrix is a singular matrix and cannot be inverted, resulting in a nan result finally.

@weisongwen Thanks for your reply, Professor Wen. It seems that I have fallen into confusion. I will list them one by one:

1、As you wrote in the MD, the results of RTK position (solution 2) will be saved by default in ~/GraphGNSSLib/trajectory_psr_dop_fusion.csv. However, I cannot find the function to log the results.

But I did find the log function factor_graph.logResults(logPath); in psr_doppler_fusion.launch. As you said, the default name is ~/GraphGNSSLib/trajectory_psr_dop_fusion.csv, but I see that the name is <param name="logPath" type="string" value="$(findglobal_fusion)/FGO_trajectoryllh_psr_dop_fusion.csv" />. It seems different.

2、I was confused about the meaning of ref_lat, ref_lon, and ref_alt. Initially, I thought they were transformed from station_x, station_y, and station_z to represent llh coordinate, but after testing, I found that they represent different positions. Then I looked up the previous issue and found that it seems to be related to the enu coordinate system?, What I want to know is whether this value will affect the final positioning result?

#define ref_lon 114.179000972 /* Weisong: reference longitude */
#define ref_lat 22.3011535667 /* Weisong: reference latitude */
#define ref_alt 6.42821512092 /* Weisong: reference altitude */

#define station_x -2414266.9200 /* Weisong: pose x of base station */
#define station_y 5386768.9870  /* Weisong: pose y of base station */
#define station_z 2407460.0310  /* Weisong: pose z of station */

3、No matter how I change the opt in the code, I still cannot get the same result as in your paper. Is the data (solution 2 static) the same as in your paper? Or did I make some mistakes?
Here are the results of my run, with orange representing rtklib, blue representing fgo, and red representing groundtruth.

Dear Dr Wen：
Thank you for sharing the repository, I'm trying to run the code, but It is stuck in the "---> psr_doppler_car_rtk_node Started", and nothing happended next.
The raw data read and publish launch "dataublox_TST20190428.launch" is running perfectly. After launching this, I typed Ctrl+C to stop, and then roslaunch the "psr_doppler_fusion.launch" but it is stuck, I don't konw whether I missed something important? Can you help me with this peoblem?

I need your help!
I want to use google map to display GPS location in real time. How?
Thank you for your guidance

Hi, thanks for the nice work. In gnss_tools.h, there is a missing include file at "#include <nlosExclusion/GNSS_Raw_Array.h>". I didn't find the file in that location and other locations. Would you please tell me how to access the file?

Hi, thanks for your work!
Could you tell pls what's the logic under checkCarrierPhaseConsistency?

Any ideas how to extend your solution to work with other satellite types
like glonass, etc. (all satellites which rtklib currently supports?)
for rtk (psr_doppler_car_rtk)?

您好，当我运行sudo apt-get install ros-kinetic-novatel-msgs这行安装程序的时候，因为我是ubuntu20.04.6所以，我改成了ros-noetic-novatel-msgs，显示无法定位软件，我去了ROS-noetic，搜索novetal ,只是提供个gps-driver等包，我想了解一下在ubuntu20.04系统如何进行正确的编译

Dr Wen：
I haven't used ROS before. So I want to know the debug method.

Hi,

When I am running catkin_make for the first time, it is failing with the following error:

WARNING: Catkin package name "nlosExclusion" does not follow the naming conventions. It should start with a lower case letter and only contain lower case letters, digits, and underscores.
-- ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
-- ~~  traversing 3 packages in topological order:
-- ~~  - nlosExclusion
-- ~~  - global_fusion
-- ~~  - rviz_satellite
-- ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
-- +++ processing catkin package: 'nlosExclusion'
-- ==> add_subdirectory(GraphGNSSLib/nlosExclusion)
-- Could NOT find nodelet (missing: nodelet_DIR)
-- Could not find the required component 'nodelet'. The following CMake error indicates that you either need to install the package with the same name or change your environment so that it can be found.
CMake Error at /usr/share/catkin/cmake/catkinConfig.cmake:82 (find_package):
  Could not find a package configuration file provided by "nodelet" with any
  of the following names:

    nodeletConfig.cmake
    nodelet-config.cmake

  Add the installation prefix of "nodelet" to CMAKE_PREFIX_PATH or set
  "nodelet_DIR" to a directory containing one of the above files.  If
  "nodelet" provides a separate development package or SDK, be sure it has
  been installed.
Call Stack (most recent call first):
  GraphGNSSLib/nlosExclusion/CMakeLists.txt:4 (find_package)

In my CMakeLists.txt, I get this same error for all of the contents of this

find_package(catkin REQUIRED COMPONENTS
  message_filters
  nodelet
  roscpp
  pcl_conversions
  sensor_msgs
  tf2
  tf2_ros
  pcl_ros
  rospy
  std_msgs
  message_generation
)

When using data collected by ourself, the command " roslaunch global_fusion dataublox_TST20190428.launch " cannot be run.
So why? what do we need to pay attention to ？

Thank you !

Since the remaining error can still reach more than 7 meters, do you have any plan to integrate more sensors, such as LiDAR and Vision? In structured environments, the LiDAR and Vision usually works well.

I am working on implementing the Geman-McClure loss function with a dynamic update of the theta parameter in each iteration to achieve Graduated Non-Convexity (GNC) results in the Ceres Solver. Despite following the recommended approach, my current implementation does not seem to influence the Ceres solver's behavior as expected. I would greatly appreciate your guidance on this matter. Below is a rough outline of my code:

// Geman-McClure
else if (loss == "Geman"){
double a =2;
loss_function = new GemanMcClureLoss(a);
}

// Geman-McClure loss function class
class GemanMcClureLoss : public ceres::LossFunction
{
public:
explicit GemanMcClureLoss(double delta) : delta_(delta), delta_square(delta * delta) {}

virtual void Evaluate(double s, double rho[3]) const override
{
    const double aux = 1.0 / (delta_ + s);

    rho[0] = delta_ * s * aux;
    rho[1] = delta_square * aux * aux;
    rho[2] = -2. * rho[1] * aux;
}

private:
const double delta_;
const double delta_square;
};

// Main loop for iterating and updating theta
int max_iter = 5;

for (int iter = 0; iter < max_iter; ++iter) {
ceres::Problem problem;

// Add residual blocks with Geman-McClure loss
ceres::CostFunction* cost_function = new ceres::AutoDiffCostFunction<ResidualFunction, 1, 1>(new ResidualFunction(res));
problem.AddResidualBlock(cost_function, loss_function, &initial_x);

// Solver configuration and solving
ceres::Solver::Options options;
ceres::Solver::Summary summary;
ceres::Solve(options, &problem, &summary);

// Update theta value and reset the loss function's theta
theta /= 1.4;
loss_function->UpdateTheta(theta);
}

您好，我注意到在psr_doppler_car_rtk.launch中并未有logPath的设置，而在psr_doppler_fusion.launch中有logPath的设置，这是不是就意味着FGO-RTK时是不会输出.csv的结果文件呢？感谢！

Hi,can you provide docker file or image to us?So we can not install the environment ,at the same time run the code as soon as possible.Thank you.

Q1: if I use docker, can I skip previous steps and start from step 5 ?
Q2: I clone the repo and use make build, then came the following problems

Cloning into 'ceres-solver'...
unable to access 'https://ceres-solver.googlesource.com/ceres-solver/': Failed to connect to ceres-solver.googlesource.com port 443: Connection refused

so I replace the url with https://github.com/ceres-solver/ceres-solver.git, but it still can not run success.

hi,thanks for your contribution!
We want to use Tight coupling algorithm to fuse raw imu data and gnss raw data.But the datasets does not provide extrinsic Parameters between imu and antenna, timestamp.At the same time,FGO is not fuse imu. Can help?

Thank you for sharing such a great package! Since the factor graph optimization is applied, why the error can still reach more than 7 meters?

Hello Dr. Weisong,

I had the same issue as seen in #6

However, when I ran the code as suggested in #6 :
First run roslaunch global_fusion psr_doppler_fusion.launch first, and then
run roslaunch global_fusion dataublox_TST20190428.launch shortly after.

It seems to work fine but it still got stucked again as seen in the below image.

I first ran this on this terminal: roslaunch global_fusion psr_doppler_fusion.launch

then I ran this next: roslaunch global_fusion dataublox_TST20190428.launch

Kindly let me know if am missing any step(s).

Should I run these steps on the same terminal? and not on separate terminals

source ~/GraphGNSSLib/devel/setup.bash
roslaunch global_fusion dataublox_TST20190428.launch
roslaunch global_fusion psr_doppler_fusion.launch

Thank you for your help.

I am sorry for disturb you in this panel:
I have a problem:
我安装了ceres-slover在/usr/local/ceres-solver2.0.0的文件夹，同时在cmakelist里面也通过set设置了相关的Ceres-DIR目录，C++版本设置为14


同时在相应的程序实现中，我也将include<ceres/ceres.h>类似的版本改成了下面的版本

但是程序依然报错：



我发现ceres调用出错的地方，指向的还是默认的ceres库，并没有使用ceres2.0.0的库中的函数，请问这个如何调整才能正确运行？