This project is the SLAM project of the Udacity Robotics Software Engineer Nanodegree. For the project, I applied the Real-Time Appearance Based Mapping (RTAB-Map) in ROS to perform SLAM in a simulated environment. See the writeup for an extended discussion of the theoretical content on SLAM algorithms and specifics of RTAB-Map.
The project was developed on Ubuntu 16.04 LTS with ROS Kinetic, Gazebo and catkin installed.
The robot relies on the rtabmap_ros ROS package, which should be installed through apt-get.
Use catkin to build the packages from source. From the catkin workspace where you cloned the repo, run:
catkin_make; source devel/setup.bash
to build the workspace packages and add them to the paths of ROS.
After the above steps, you should be able to run the commands below in separate terminals:
Launch the world in Gazebo:
roslaunch slam_project slam_world.launch
Launch the teleop node for keyboard control:
roslaunch slam_project teleop.launch
Launch the RTAB-Map mapping node
roslaunch slam_project mapping.launch
Launch the RViz GUI:
roslaunch slam_project rviz.launch
The project repository contains source code of a catkin workspace, with some supporting code and shared object files provided by Udacity. All of them are wrapped in the slam_project package. The package includes a custom made robot model, a few world files specifying an environment in Gazebo, and a few shell scripts interfacing with the RTAB-Map library.
There are two tasks involved in this project: robot model configuration and rtabmap for SLAM.
The robot model was extended from the previous project RoboND-Where-Am-I and instead has an RGB-D camera to input depth information to RTAB-Map.
The robot uses the information from the odometer, the RGB-D camera and the laser rangefinder to perform SLAM in the designated environment through rtabmap. The rtabmap library produces map database files that can be read by its own visualization tools. Details of the algorithm can be found in the writeup. Here is an image of the reconstructed kitchen dining environment:
The files were too large to host on Github, so I uploaded them to Google Drive and the sharing link is [here]https://drive.google.com/file/d/1kWlC5CdOrbdKblN0IiVoP8pLaOwIMDZy/view?usp=sharing).