known_map_localization

ROS code related to my bachelor thesis "ROS-Modul zur Lokalisierung in a priori bekannten Karten".

Content

This repository consists of multiple packages

• known_map_localization: The main package containing the localization node developed in the thesis
• mapmerge: This package contains a slightly modified version of the mapmerge library by Stefano Carpin [1] available from [2]
• mapstitch: This package contains an also slightly modified version of mapstitch, a ROS package available from [3]
• cs_merge_methods: A modified package from the cs_merge stack available from [4]
• cs_merge_msgs: A package from the cs_merge stack available from [4]

Dependencies

Additionally to standard ROS packages and the packages from this repository the following ROS packages or system libraries are needed to build known_map_localization:

• map_server
• geodesy
• geographic_msgs
• orb_slam (a modified version of ORB-SLAM from https://github.com/cehberlin/ORB_SLAM)
• Boost
• OpenCV

See known_map_localization/package.xml for a full list of dependencies.

Usage

The known_map_localization node is launched with the known_map_localization/launch/kml_node.launch file. It takes 3 mandatory arguments:

• slam_map_topic: The topic where the SLAM occupancy grid map is published as a nav_msgs/OccupancyGrid message
• gps_topic: The topic where GPS fixes are published as sensor_msgs/NavSatFix messages
• known_map: The full path to a configuration file which determines the known map. The format should be as described in http://wiki.ros.org/map_server#YAML_format extended by an additional tag specifying the anchor coordinates. As an example:

# Geographic pose of the anchor, using the WGS 84 reference ellipsoid.
# Orientation uses the East-North-Up (ENU) frame of reference (yaw angle, ccw, in degrees). 
anchor: {position: {altitude: 0., latitude: 52.51273962, longitude: 13.32486415}, heading: 90}

free_thresh: 0.196
image: known_map.png
negate: 0
occupied_thresh: 0.65
origin: [-10.0, -10.0, 0.0]
resolution: 0.1

The optional argument algorithm specifies which alignment algorithm to use. Valid options are: mapmerge (default), mapstitch, cs_merge_icp_gradient and cs_merge_icp_svd.

A sample roslaunch call starting known_map_localization:

roslaunch known_map_localization kml_node slam_map_topic:=/orb_slam/projected_map gps_topic:=/robot/gps known_map:=PATH_TO_YAML.yaml

Output

The following topics are used to publish the results of the localization:

• known_map_localization/localization_node/geo_pose: The estimated robot pose
• known_map_localization/localization_node/kml_base_link: The transform from the anchor to the robot pose

Documentation

The known_map_localization code is documented using doxygen-style comments. Therefore documentation can be generated using doxygen.

Resources

[1] Stefano Carpin. Fast and accurate map merging for multi-robot systems. In: Autonomous Robots 25.3 (2008), S. 305-316

[2] mapmerge library: http://robotics.ucmerced.edu/map-merging-code

[3] mapstitch ROS package: http://wiki.ros.org/mapstitch

[4] cs_merge ROS stack: http://wiki.ros.org/cs_merge