General description

CRSM SLAM stands for Critical Rays Scan Match SLAM (Simultaneous Localization And Mapping). Its main characteristics follow:

• Uses scan-to-map matching, instead for the usual scan-to-scan matching, aiming at noise accumulation reduction.
• Scan matching is performed via a Random Restart Hill Climbing algorithm (RRHC).
• Only the critical rays take part in the RRHC. As critical are denoted the rays which contain more spatial information than the others.
• The map update is performed with dynamic intensity, depending on the current environmental structure.

The full algorithmic description can be found in the following paper:

http://link.springer.com/article/10.1007/s10846-012-9811-5#!

CRSM SLAM does not have a close loop behaviour, but it gives very good results in featured spaces.

Finally, it was used in the PANDORA autonomous vehicle that takes part in the world-wide robotic competition RoboCup-RoboRescue.

Installation

CRSM SLAM is provided through the official ROS repositories. To install execute the following command :

sudo apt-get install ros-$ROS_DISTRO-crsm-slam

You can also manually set it up:

cd <your_catkin_ws>/src
git clone https://github.com/etsardou/crsm-slam-ros-pkg.git
cd ../
catkin_make

Screenshots/Multimedia

The following environment was created in the Gazebo simulator :

One video that demonstrates CRSM SLAM in the above environment is the following :

The final map produced by CRSM SLAM is the following :

Package Input / Output

There is only one node spawn whose name is crsm_slam_node.

CRSM needs as input :

• sensor_msgs::LaserScan publisher that contains the Laser rays.

On the other hand, CRSM outputs :

• nav_msgs::OccupancyGrid publisher which contains the occupancy grid map.

• A tf transformation with the robot's pose.

• A nav_msgs::Path that contains the robot trajectory.

Parameters

CRSM is fully parameterizable by changing the parameter file "crsm_slam/config/crsm_slam/crsm_slam_parameters.yaml". The meaning of each parameter, as well as the expected effect on the algorithm follow:

• occupancy_grid_publish_topic

  • Description : The occupancy grid publishing topic

  • Default : /crsm_slam/map

• robot_trajectory_publish_topic

  • Description : The trajectory publishing topic

  • Default : /crsm_slam/trajectory

• trajectory_publisher_frame_id

  • Description : The trajectory frame ID

  • Default : map

• laser_subscriber_topic

  • Description : The laser subscriber topic

  • Default : /crsm_slam/laser_scan

• world_frame

  • Description : Holds the world frame

  • Default : world

• base_footprint_frame

  • Description : Holds the base footprint frame - (x,y,yaw)

  • Default : base_footprint_link

• base_frame

  • Description : Holds the base frame

  • Default : base_link

• map_frame

  • Description : Holds the map frame

  • Default : map

• laser_frame

  • Description : Holds the laser frame

  • Default : laser_link

• hill_climbing_disparity

  • Description : Disparity of mutation in pixels at hill climbing
  • Default : 40
  • Effect : If the disparity is increased, hill climbing searches for solution in a wider area. If the robot has large velocities, one measure to take is to increase the specific parameter.

• slam_container_size

  • Description : Map size of initial allocated map
  • Default : 500
  • Effect : This variable controls the initial size of the container of occupancy grid map. If needed during the experiment, the map is resized.

• slam_occupancy_grid_dimentionality

  • Description : [OC]cupancy [G]rid [D]imentionality - the width and height in meters of a pixel
  • Default : 0.02
  • Effect : If the current variable increases, the algorithm will run faster but there will be a loss of quality.

• map_update_density

  • Description : Map update density (0-127)
  • Default : 30
  • Effect : If the current variable increases, the map probabilities will change with a higher rate toward the probability extremes (0 for occupancy, 1 for free space).

• map_update_obstacle_density

  • Description : Coefficient for obstacle update density (0+)
  • Default : 3.0
  • Effect : In CRSM SLAM the obstacles probabilities update with a higher rate than the free space's ones. This aims at creating a better reference for the RRHC algorithm.

• scan_density_lower_boundary

  • Description : Scan density lower boundary for a scan-part identification
  • Default : 0.3
  • Effect : This variable holds the minimum distance two consecutive rays must have in order to be considered the extremes of two scan parts.

• max_hill_climbing_iterations

  • Description : Maximum RRHC iterations
  • Default : 40000
  • Effect : If more search power is needed, you can increase this variable.

• occupancy_grid_map_freq

  • Description : The occupancy grid map publishing frequency
  • Default : 1.0 Hz

• robot_pose_tf_freq

  • Description : The robot pose publishing frequency
  • Default : 5.0 Hz

• trajectory_freq

  • Description : The trajectory publishing frequency
  • Default : 1.0 Hz

• desired_number_of_picked_rays

  • Description : The desired number of picked rays [algorithm specific]
  • Default : 40
  • Effect : CRSM SLAM tries to keep the number of critical rays picked almost constant. If the specific variable increases, more rays in average will be picked.

• robot_width

  • Description : The robot width in meters
  • Default : 0.5

• robot_length

  • Description : The robot length in meters
  • Default : 0.5

How to run CRSM SLAM

CRSM SLAM can be executed either with rosrun or by roslaunch.

The package includes two launchers, one for a simulated laser (crsm_slam_simulation.launch) and one for a real one (crsm_slam_real.launch), which "listens" to a hokuyo_node.