CHAD TSDF

CHAD TSDF data structure for SLAM with 3D mesh reconstruction.

Platform	Build Status
Linux
MacOS	-
Windows	-

Description

TSDF mapping backend with real-time capabilities and high compression. It is tailored towards large-scale maps, while still being able to handle smaller voxel sizes (~0.05).

Getting Started

LVR2

The 3D mesh reconstruction currently relies on LVR2, so you will likely need to build and install it yourself. If not present, chad_tsdf will attempt to build it dynamically via CMake's FetchContent mechanism, but this method may fail on some systems and will still require the LVR2 dependencies.

CMake

include(FetchContent)
FetchContent_Declare(chad_tsdf
    GIT_REPOSITORY "https://github.com/M2-TE/chad_tsdf.git"
    GIT_TAG "main"
    GIT_SHALLOW ON)
FetchContent_MakeAvailable(chad_tsdf)
target_link_libraries(${PROJECT_NAME} PUBLIC chad::tsdf)

C++

#include <chad/tsdf.hpp>

int main() {
    float sdf_size = 0.05f;
    float sdf_trunc = 0.10f;
    chad::TSDFMap map(sdf_size, sdf_trunc);

    // insert glm points
    std::vector<glm::vec3> points_glm;
    glm::vec3 position_glm;
    map.insert(points_glm, position_glm);

    // insert eigen points
    std::vector<Eigen::Vector3f> points_eigen;
    Eigen::Vector3f position_eigen;
    map.insert(points_eigen, position_eigen);

    // reconstruct 3D mesh
    map.save("mesh.ply");
    return 0;
}

Roadmap

• Compressed sparse TSDF map
• Submapping
• Loop closure
• Space carving

Acknowledgments

• gtl: Hashmap implementations.
• lvr2: 3D mesh reconstruction via sdf marching cubes.
• libmorton: Morton code en-/decoding with BMI2 acceleration.