CGR (Laser Scanner, Kinect - FSPF) Localization
with KLD Resampling

1. Summary

This package provides the code for CGR localization to localize a robot in 2D using either laser rangefinder readings or depth images obtained from Kinect-style sensors.

Author: Joydeep Biswas (joydeepb AT ri DOT cmu DOT edu), Brian Coltin (bcoltin AT ri DOT cmu DOT edu)
Benjamin Mende (bmende AT cs DOT umass DOT edu)
License: LGPL
Original source: hg http://hg.cobotrobots.com/cgr_localization
Original Documentation: http://www.cs.cmu.edu/~coral/projects/localization/source.html

With KLD (this!) on github at: https://github.com/bmende/cgr_localization

2. Dependencies

To install all dependencies on Ubuntu, run "./InstallPackages", or copy & run the following command:

sudo apt-get install g++ libqt4-dev cmake libpopt-dev libusb-1.0-0-dev liblua5.1-dev libglew1.5-dev libeigen3-dev
On other platforms, you will have to manually install the following packages:
- A C++ compiler like GCC
- cmake
- QT
- popt
- libusb
- LUA
- GLEW
- eigen3

3. Compiling

Run "make" in the cgr_localization folder OR run "rosmake cgr_localization" after adding the full path of the cgr_localization package to the ROS_PACKAGE_PATH environemt variable.

4. Testing with demo data

The following demo data sets are available:

CoBot2 in GHC7 (LIDAR)
Download the ROS bag file from http://cobotrobots.com/data/cgr_localization/cobot2_ghc7_lidar.bag
Modify the file config/localization_parameters.cfg, line 9 to mapName = "GHC7";

CoBot2 in GHC7 (Kinect)
Download the ROS bag file from http://cobotrobots.com/data/cgr_localization/cobot2_ghc7_kinect.bag
Modify the file config/localization_parameters.cfg, line 9 to mapName = "GHC7";

CoBot2 in NSH4 (LIDAR)
Download the ROS bag file from http://cobotrobots.com/data/cgr_localization/cobot2_nsh4_lidar.bag
Modify the file config/localization_parameters.cfg, line 9 to mapName = "NSH4";

To run the code with the demo data:
Use "roslaunch cgr_localization cgr_demo.launch" to run the localization code and display that on the bundled GUI),
Play back the data file with "rosbag play <bagfile.bag> --clock"

5. Running on your own robot

To run cgr localization on your own robot, you need to generate a vector map of your environment. For an example vector map, see maps/GHC7/GHC7_vector.txt . Each entry in the vector map represents a line in the world, and is of the form:
x1, y1, x2, y2
where x1, y1 is the start location and x2, y2 the end location of the line. The vector maps reside in the maps folder. For example, the map for "GHC7" resides in the file maps/GHC7/GHC7_vector.txt . Once the map is created, it needs to be added to maps/atlas.txt so that cgr_localization loads it on startup.

CGR localization uses analytic raycasts on vector maps for the observation functions. To speed up the analytic renders at runtime, visibility lists are pre-computed for the vector maps. To generate the visibility lists for your map, run the following command:
./bin/pre_render -m "YourMapName" [-n NumberOfThreads]
e.g. ./bin/pre_render -m"GHC7" -n8

The required ROS topics are "odom" of type nav_msgs/Odometry , "scan" of type sensor_msgs/LaserScan and/or "kinect_depth" of type sensor_msgs/Image (Raw 16-bit depth data gathered via libfreenect from the kinect)



Publications

"Corrective Gradient Refinement for Mobile Robot Localization", Joydeep Biswas, Brian Coltin, and Manuela Veloso, Proceedings of IEEE/RSJ International Conference on Intelligent Robots and Systems, September, 2011, pp. 73 - 78.

"Depth Camera Based Indoor Mobile Robot Localization and Navigation", Joydeep Biswas and Manuela Veloso, Proceedings of IEEE International Conference on Robotics and Automation, May, 2012.