LEGO-SLAM

LEGO-SLAM is a light weight stereo visual SLAM system which contains multi hand-made modules, such as a frontend with the pyramid KLT optical flow method based on the Gauss-Newton algorithm & OpenCV ParallelLoopBody and a backend with the graph-based Levenberg-Marquardt optimization algorithm (LEGO or g2o (optional)). Generally speaking, LEGO-SLAM is a silding-window stereo visual SLAM framework for learning and practicing visual SLAM theories. Now, the optimization algorithm in the backend (LEGO) only depends on the Levenberg-Marquardt algorithm and it will include more algorithms in the future, such as Dog-leg. Moreover, the loop-closure module will be implemented in the near future. This version is only tested on Ubuntu 18.04.

Installation

Prerequisites

• GCC and CMake

• Eigen3: Linear algebra

  On Ubuntu 18.04, we can install CMake and Eigen3 following:

  # gcc and cmake 
  sudo apt-get install gcc cmake 
  # Eigen3 
  sudo apt-get install libeigen3-dev 

• OpenCV 3: Only for image visualization and I/O (Version >= 3.2)

• Sophus: Lie groups and Lie algebra based on Eigen

• Pangolin: Display and interactive visualization based on OpenGL

• glog, gtest, gflags

• CSparse: Sparse matrix computing

  # suitesparse 
  sudo apt-get install libsuitesparse-dev

• g2o (optional): A general framework for graph optimization

• LEGO: A light weight graph-based optimization library

Build and Run

• Build

  We can build and run LEGO-SLAM from this repository and follow:

  git clone https://github.com/zha0ming1e/LEGO-SLAM.git 
  cd LEGO-SLAM/ 
  mkdir build/ 
  cd build/ 
  cmake .. && make -j6 

  Now we have already built the LEGO-SLAM and then we can test and run it.

• Test and Run: There will be 2 executable files under the bin/ folder

  • Test: Test triangulation in 3D

    # test triangulation 
    ./bin/legoslam_test_triangulation 
    
    [==========] Running 1 test from 1 test case.
    [----------] Global test environment set-up.
    [----------] 1 test from LEGOSLAMTest
    [ RUN      ] LEGOSLAMTest.Triangulation
    [       OK ] LEGOSLAMTest.Triangulation (0 ms)
    [----------] 1 test from LEGOSLAMTest (0 ms total)
    
    [----------] Global test environment tear-down
    [==========] 1 test from 1 test case ran. (1 ms total)
    [  PASSED  ] 1 test.

  • Run: Modify the config files under the config/ folder, such as the path of KITTI datasets and then run LEGO-SLAM

    # KITTI 00 
    ./bin/legoslam_kitti_feature ./config/kitti_00.yaml 
    # KITTI 05 
    ./bin/legoslam_kitti_feature ./config/kitti_05.yaml

Results

This version only supports KITTI odometry datasets.

• KITTI 00

  

• KITTI 05

  

• Demo

  

References

• ORB-SLAM2: Real-Time SLAM for Monocular, Stereo and RGB-D Cameras, with Loop Detection and Relocalization Capabilities
• ORB-SLAM3: An Accurate Open-Source Library for Visual, Visual-Inertial and Multi-Map SLAM
• DSO: Direct Sparse Odometry
• LDSO: DSO with SIM(3) pose graph optimization and loop closure
• VINS-Mono: A Robust and Versatile Monocular Visual-Inertial State Estimator
• VINS-Course: VINS-Mono code without Ceres or ROS
• slambook2: Edition 2 of the slambook