caomw / Robust-View-Graph-SLAM

C++ implementation of view-graph SLAM using nonlinear least-squares

Release Notes (The current source is out-of-date, and is not maintained. Contact me if you need any details.).

• MATLAB and C++ Implementations of View-Graph SLAM.
• This is a robust mixture between Nonlinear Least-Squares Estimation and Multiple-Views Pose-Graph SLAM. This implementation is Applicable for both, stereo and monocular settings.
• If you are planning on using this implementation, please cite our paper: T. Abuhashim and L. Natale, "Robustness in view-graph SLAM," 2016 19th International Conference on Information Fusion (FUSION), Heidelberg, 2016, pp. 942-949. URL: http://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=7527987&isnumber=7527857
• Copyright (C) 2016 iCub Facility - Istituto Italiano di Tecnologia
• Authors: Tariq Abuhashim, Nicolo' Genesio
• Emails: [e-mail], [e-mail]
• Last Updated: Nov 2016
• Acknowledgement: This research has received funding from the European Union’s Seventh Framework Programme for research, technological development and demonstration under grant agreement No. 611909 (KoroiBot).
• License: Released under the terms of the LGPLv2.1 or later, see LGPL.TXT

Installation of the C++ implementation

This implementation was tested on Ubuntu Trusty Tahr (14.04.5 LTS) and Kylin (16.04.1 LTS). Install all required tools:

sudo apt-get install git g++ cmake cmake-curses-gui

Cmake 3.2.2 or higher is required, so:

sudo apt-get install software-properties-common
sudo add-apt-repository ppa:george-edison55/cmake-3.x
sudo apt-get update
sudo apt-get upgrade

Then, check

cmake --version

OpenCV and OpenCV_Contrib

git clone https://github.com/opencv/opencv.git
git clone https://github.com/opencv/opencv_contrib.git
mkdir opencv_build
cd opencv_build
cmake -DOPENCV_EXTRA_MODULES_PATH=../opencv_contrib/modules -DCMAKE_INSTALL_PREFIX=/install/path ../opencv
make -j5

To install:

sudo make install

If not installing, then update the environmental variable:

OPENCV_DIR=/path/to/your/opencv_build

To check the currently installed version of opencv:

pkg-config --modversion opencv

SuiteSparse

#####To download the most recent version:

#####git clone https://github.com/jluttine/suitesparse.git

#####However, our implementation of Takahashi's inverse (spinv) needs UFconfig, this is found in SuiteSparse-3.7.1.tar.gz. Also, need to install lapack, blas, openblas, metis, and parmetis (not really needed at the moment).

sudo apt-get install liblapack-dev libblas-dev libopenblas-dev 
#####sudo apt-get install libmetis-dev libparmetis-dev
cd SuiteSparse

#####Download metis-4.0.3, and rename the folder to matis-4.0. In the Makefile.in, update the following:

#####CC = gcc
#####OPTFLAGS = -O3 
#####COPTIONS = -fPIC

Download SuiteSparse. To compile without metis, edit like 293 in SuiteSparse_config/SuiteSparse_config.mk

CHOLMOD_CONFIG ?= $(GPU_CONFIG) -DNPARTITION

Then;

make
sudo make install INSTALL=yourprefix

Then update the environmental variable:

SUITESPARSE_DIR=path/to/your/SuiteSparse/install

Eigen

Download and compile Eigen.

hg clone https://bitbucket.org/eigen/eigen/
cd eigen
mkdir build
cd build
cmake ../ -DCMAKE_INSTALL_PREFIX=yourprefix
sudo make install

Then update the environmental variable:

EIGEN_INCLUDE_DIR=path/to/your/eigen/build/include

libboost

Install using precompiled binaries,

sudo apt-get install libboost-all-dev

If you don't like to install boost system-wise, download source from Boost, then follow instructions to compile.

cd path/to/boost
mkdir build
./bootstrap.sh PREFIX=path/to/install
./b2
./b2 install --prefix=path/to/install

Then update the environmental variable:

EIGEN_INCLUDE_DIR=path/to/your/eigen/build/include

YARP Network and iCub

Our implementation uses YARP to replay data, implement multi-threading, and locate different resources. I personally install it from source. Also, if you are a fan of iCub, you will find very useful tools and simulations. To run iCubSIM, you would need to install SDL, GLUT, ODE, IPOPT, and gfortran:

sudo apt-get install libsdl1.2-dev freeglut3 freeglut3-dev libode-dev coinor-libipopt-dev libgsl2 libgsl-dev
sudo apt-get update && sudo apt-get install gfortran -y

Additionally, for Ubuntu 16.04.01

sudo apt-get install libace-dev libghc-glut-dev 

In addition to the installation instruction, if you like to install ICUB_SIM, then set:

ICUB_SHARED_LIBRARY = ON

VLFeat

Download VLFeat 0.9.20 binary package

run <VLFEATROOT>/toolbox/vl_setup

If compiling without MATLAB, in the makefile, comment out the line

#include make/matlab.mak. Then, "make".

To compile our code with VLFeat, update the environmental variable:

VLFEAT_ROOT=/path/to/vlfeat/directory

Matlab

After installing Matlab, update the environmental variable:

MATLAB_ROOT=path/to/your/MATLAB/RXXXXx

Compiling MEX functions in MATLAB

If you are planning on using our mex wrappers, you need to download and install MATLAB. Update all the related paths in compile_PwgOptimiser.m and compile_GraphOptimiser.m, then;

run compile_PwgOptimiser;
run compile_GraphOptimiser;

This will compile and test against the MATLAB code (if this isn't needed, then comment out this comparison part).

Installing the MATLAB code dependencies

The MATLAB code isn't yet available, but we will upload the code soon. In order to run our MATLAB implementation, you would need to install GP-stuff (we use their sparse inverse, if you have an alternative solution, then you may skip this step):

git clone https://github.com/gpstuff-dev/gpstuff

In MATLAB, navigate to the folder, then

run matlab_install('SuiteSparseOn')

There were two fixes. First, in matlab_install.m, replace cd SuiteSparse with cd /your/path/to/suitesparse. Second, in SuiteSparse_install.m, replace function SuiteSparse_install(input) with function paths = SuiteSparse_install(input).

You would also need to install mexopencv

git clone https://github.com/kyamagu/mexopencv.git
cd mexopencv
DIR_MATLAB=/usr/local/MATLAB/R2016b
make all MATLABDIR=$DIR_MATLAB

C++ Development tools

One example that runs on Ubuntu is Eclipse. Notice that this requires to have JRE, and optionally, JDK downloaded.

sudo apt-get update
sudo apt-get install eclipse
sudo apt-get install eclipse eclipse-cdt

Doxygen

sudo apt-get install doxygen doxygen-gui

run doxywizard or create a configuration file using doxygen -g /doxygen/doxygen.cfg. To generate an index.html file in the main code folder, the easiest solution is probably to create a symbolic link or shortcut to the index.html file generated by doxygen, rather than trying to get doxygen to change the layout of it's output files. This symlink/shortcut can then be placed in the root directory of your project (or elsewhere), pointing to doxygen/html/index.html, and named anything you like to make it obvious to your users what it is.

ln -s doxygen/html/index.html index.html