Home

Awesome

Build Status

TL;DR: Sparser Relative Bundle Adjustment (SRBA) is a header-only C++ library for solving SLAM/BA in relative coordinates with flexibility for different submapping strategies and aimed at constant time local graph update. BSD 3-Clause License.

Related papers:

1. Compile

Requisites:

In Ubuntu, install requisites with:

sudo apt-get install build-essential cmake libmrpt-dev

Clone, configure and build as usual with CMake:

git clone https://github.com/MRPT/srba.git 
cd srba
mkdir build && cd build
cmake .. 
make 
make test

2. Theoretical bases

Bundle adjustment is the name given to one solution to visual SLAM based on maximum-likelihood estimation (MLE) over the space of map features and camera poses. However, it is by no way limited to visual maps, since the same technique is also applicable to maps of pose constraints (graph-SLAM) or any other kind of feature maps not relying on visual information.

The framework of Relative Bundle Adjustment (RBA) was introduced in a series of works by G. Sibley and colleagues:

Sparser RBA (SRBA) is the name of the generic and extensible framework for RBA implemented in this C++ library, and introduced in the ICRA 2013 paper (PDF, see full citation above).

3. Programming guide and documentation

4. Run sample datasets

4.1. Monocular visual SLAM with synthetic dataset

https://www.youtube.com/watch?v=ZXti4GxqkUg

4.2. Relative 2D graph-SLAM

Now you can run RBA on the dataset with:

srba-slam --se2 --graph-slam -d dataset_30k_rel_graph_slam_SENSOR.txt \\
--submap-size 10 --max-spanning-tree-depth 3 --max-optimize-depth 3 \\
--verbose 1 --noise 0.001 --noise-ang 0.2 --add-noise \\
--gt-map dataset_30k_rel_graph_slam_GT_MAP.txt \\
--gt-path dataset_30k_rel_graph_slam_GT_PATH.txt # --step-by-step