SLAM algorithm implementation in C++ that's compatible with windows?

Question

I need a SLAM algorithm for a robot that will move around a track while avoiding obstacles (only one lap so loop will be closed at the end). The robot uses GPS, compass and lidar for navigation.

I was about to implement a version of online graph slam based on Probabilistic Robotics but then read another answer on stackoverflow that said current algorithms have moved beyond it.

What are some of the state-of-the-art algorithms being used today for online slam? I wanted to use iSAM but it's not compatible with windows.

Answer 1

most open source slam solution is developed on ubuntu.

up to now, for indirect method, you can refer to ORB-SLAM

for direct method, you can refer to LSD slam

both two is available on github.

Answer 2

A new version of the open source InfiniTAM system developed at the University of Oxford has recently been released. It comes with a CMake build, so building is as simple as

git clone https://github.com/victorprad/InfiniTAM.git
cd InfiniTAM
mkdir build
cd build
cmake-gui ../InfiniTAM #SELECT ADDITIONAL BUILD OPTIONS HERE
MSBuild InfiniTAM.sln /property:Configuration=Release /property:Platform=x64

If you have a CUDA enabled GPU and the CUDA development kit installed then this will automatically be detected. If not, then I strongly recommend enabling OpenMP in the CMake GUI.