These instructions are intended for use on Linux (including WSL) only.

You'll probably need to install these packages:

build-essential cmake libboost-all-dev protobuf-compiler ninja-build mold

With these packages installed, configure using CMake. I'm using Ninja, but any generator will work. And mold is just insanely fast

cmake -B build -GNinja -DCMAKE_EXPORT_COMPILE_COMMANDS=ON -DCMAKE_BUILD_TYPE=RelWithDebInfo -DCMAKE_EXE_LINKER_FLAGS="-fuse-ld=mold" -DCMAKE_SHARED_LINKER_FLAGS="-fuse-ld=mold"

For Windows, use this instead: cmake -B build/ -S . -DCMAKE_TOOLCHAIN_FILE="vcpkg/scripts/buildsystems/vcpkg.cmake" -DCMAKE_BUILD_TYPE=RelWithDebInfo

Depending on your RAM, you may need to use -j <number> to control the number of jobs when building.

And build and run the actual executable! This will take a while

clear && cmake --build build --target gtsam-node && ./build/bin/gtsam-node

This sim example in the gtsam-testing branch is set up to provide simulated data to the current build of gtsam-playground. Just run it as a robot simulation project! The NT server URI is set by changing the JSON below. If you're running this in WSL2, the URI must be set to the IP address of your computer in Windows.

I visualize output data usually using advantagescope. The 3d visualizer is great.

In order for the data transfer to work properly, start advantagescope, then sim, then run gtsam-node. The /SmartDashboard/VisionSystemSim-main/Sim Field/Gtsam Robot topic will show the fused pose of the robot.

As of right now, these are our publishers/subscribers. The NT4-provided set timestamp is used for latency compensation. The list of camera names is configured by changing the config JSON. If no arguemnts are provided, the config is pulled from the file test/resources/simulator.json. If one argument is provided to gtsam-node, then that arguemnt is used as the path instead. The JSON example below shows two cameras with given tag corner pixel standard deviations, plus global odometry standard deviations on rotation and translation.

A reference implementation from robot code is also available.

{
    "rootTableName": "gtsam_meme_localizer1",
    "ntServerURI": "10.TE.AM.2",
    "cameras": [
        {
            "subtableName": "sim_camera1",
            "pixelNoise": 10
        },
        {
            "subtableName": "sim_camera2",
            "pixelNoise": 12
        }
    ],
    "rotNoise": [ 0.087263889, 0.087263889, 0.087263889 ],
    "transNoise": [ 0.001, 0.001, 0.001 ]
}

Subscribers

Publishers

WPILib uses a version of Eigen from https://github.com/wpilibsuite/allwpilib/blob/main/upstream_utils/update_eigen.py#L100 SHA is 96880810295b65d77057f4a7fb83a99a590122ad