caomw/graphslam-1

GraphSLAM

A GraphSLAM implementation using g2o framework

Requirements

• CMake (https://cmake.org/)
• Eigen3 (http://eigen.tuxfamily.org/index.php?title=Main_Page)
• suitesparse (http://faculty.cse.tamu.edu/davis/suitesparse.html)
• g2o (https://github.com/RainerKuemmerle/g2o)

This project was developed in linux platform, with C++11 and Python 2.7

Directories:

• src:
  • graphSLAM: C++ scripts that performs SLAM magic
  • python-helpers: Python scripts that make easy to run GraphSLAM tests
• doc:
  • Documentation: Code documentation
  • Memoria: Memoria (tesis) document explaining the work
  • Presentation for thesis defense

Compilation

To compile the C++ scripts simply go to the src/graphSLAM folder and do:

• mkdir build
• cd build
• cmake ..
• make

Notice that you must have g2o installed in your machine.

Execution

• For simulated data:

  1. Run the simulator binary my-simulator to generate the simulated data (use -h for options).
  2. Run the GraphSLAM binary my-slam to optimize the data of the simulator. Use the "guess" file for the optimization (use -h for options).
  • Use -anonymize flag from my-simulator to anonymize landmarks (for unknown data association).
  • Alternatively run g2o -i 0 -guessOdometry -o <guessInFile> <guessOutFile> to generate an initial guess file (for plotting).
  • See src/python-helpers/commons/ for functions to plot the results and change its format.

• For real data: just make sure that your data is in g2o format and proceed as before.

Seesrc/python-helpers subdirectories for test srcipts ready to run.