// the find
PRBonn/kiss-slam
A LiDAR SLAM system that just works
KISS-SLAM is a 3D LiDAR SLAM system from the Photogrammetry & Robotics lab at Uni Bonn, the same group behind KISS-ICP. It does full SLAM — odometry, local mapping, loop closure, pose graph optimization — as a single pip install. Aimed at robotics researchers and practitioners who need a working LiDAR mapping baseline without spending a week on build systems.
The pip-installable packaging is genuinely rare for a SLAM system — most require ROS and a CMake adventure; here the C++ core (Eigen, g2o, SuiteSparse) compiles via pybind11 during install and stays out of your way. The academic lineage is solid: builds on KISS-ICP for odometry and MapClosures for loop detection, both of which have published benchmarks. The IROS25 git tag means you can reproduce paper results exactly while the main branch keeps moving — a thoughtful separation that most research code ignores. Config is a single YAML file with sensible defaults and explicit indoor tuning guidance.
The C++ dependency chain (libeigen3-dev, libsuitesparse-dev) means Linux is the real target; macOS is listed but SuiteSparse on Apple Silicon has historically been painful. No ROS integration is mentioned, which is a problem for anyone running this on real hardware where sensor drivers live in a ROS node — you'll be writing your own bridge. At 524 stars and less than a year old it's still early, so the config surface is thin and there's no documentation beyond the README and the paper. Real-time performance figures on common hardware (Ouster, Velodyne, Livox) aren't in the README — you have to read the paper to know what to expect.