Following this guide, you will run the loop exploration approach based on the Loop-aware Exploration Graph (see the article). This implementation uses a simulated robot by MobileSim and a modified version of GMapping SLAM.

If you use the LEAG, the Loop Exploration, or any map form the maps folder in an academic work, please cite the original article.

Note: This implementation was tested in Ubuntu 20.04.4 LTS

The project is compiled using catkin. From the project root directory:

catkin_make

The project pipeline is composed of four elements that must be executed in sequence, each one in its own terminal.

The robot is a Pionner Pioneer 3-DX equipped with a SICK LMS 200 laser range-finder simulated by the MobileSim. After installing the MobileSim, open it:

MobileSim

Note: You need to have a map to import in MobileSim; some samples are provided in the maps folder.

You can define the map to be loaded by MobileSim via command line:

MobileSim -m maps/castle.map

Rosaria is the bridge between the robot and ROS. From the root directory, run these commands:

source devel/setup.bash
roslaunch loop_exploration rosaria.launch

The GMapping SLAM provides the map and robot's pose estimations. The LAEG needs a map with the visited region (see the article for more details). The version of GMapping modified to provide the visited region is available here. After compiling it, run these commands from its root directory:

source devel/setup.bash
roslaunch gmapping gmapping.launch

Finally, to execute the exploration, run from the LAEG root directory:

source devel/setup.bash
./build/loop_exploration/exploration config_file.ini

Note: You need to have a configuration file, a sample is provided in src/loop_exploration/configs/config_sample_file.ini

The parameters of the configuration file are:

• slam_system: Defines the SLAM system to be used. So far, only GMapping is supported. To use it, set slam_system = GMAPPING.
• save_log: Defines if a log must be saved (save_log = true) or not (save_log = false)
• output_address: Defines the output folder to store the outputs.
• planning_iterations_per_second: Defines the maximum planning interations per second. We suggest to use planning_iterations_per_second = 4.
• map_resolution: Defines the grid map resolution in meters. We suggest to use map_resolution = 0.1.
• map_width_height: Defines the fixed maximum width and height of gird map in cells. We suggest to use map_width_height = 2000.
• free_threshold: Defines the threshold to consider a cell free. We suggest to use free_threshold = 40.
• obstacle_threshold: Defines the threshold to consider a cell occupied. We suggest to use obstacle_threshold = 60.
• external_border: Defines the range, in cells, of unknown cells surrounding the map that will be considered as a traversable area. We suggest to use external_border = 40.
• inflate_obstacles_pad_size: Defines the range of cells surrounding the obstacles that will be considered as obstacles, inflating them. We suggest to use * inflate_obstacles_pad_size = 2*.
• min_dist_to_obstacles_from_unknown_cells: Defines the minimum Euclidean distance (in cells) between outer frontiers and obstacles. We suggest to use min_dist_to_obstacles_from_unknown_cells = 4.
• min_dist_to_obstacles_from_unknown_cells: Defines the minimum Euclidean distance (in cells) between inner frontiers and obstacles. We suggest to use min_dist_to_obstacles_from_unknown_cells = 2.
• max_euclidean_dist_to_merge_frontiers: Defines the maximum Euclidean distance (in cells) between two outer frontiers to merge them. We suggest to use max_euclidean_dist_to_merge_frontiers = 10.
• max_nodes_dist_to_merge_frontiers: Defines the maximum distance (in graph nodes) between two outer frontiers to merge them. We suggest to use max_nodes_dist_to_merge_frontiers = 2.
• alpha: Defines the alpha parameter that set the importance of outer frontiers relative to the inner ones (see Eq. 5 in the article for more details). We suggest to use alpha = 0.5.
• potential_window_size: Defines the size (in cells) of the local potential window used to navigate. We suggest to use potential_window_size = 60.
• max_goal_radius: Defines the desired distance (in cells) between the robot and the local goal. We suggest to use max_goal_radius = 20.
• screen_x_init: Defines the x axis start position for the visualization.
• screen_y_init: Defines the y axis start position for the visualization.
• screen_scale_init: Defines the start scale for the visualization.
• screen_width_init: Defines the width of window for the visualization.
• screen_height_init: Defines the height of window for the visualization.

Some keyboard commands could be used to control the visualization interface and the robot.

• esq: Finish the exploration prematurely, without saving a map.
• f: Enable/Disable the exhibition of the local potential field.
• c: Locks and centers the camera in the robot (if the camera is already locked, it will disables the lock)
• l: Changes the LAEG visualization mode.
• p: Enable/Disable the exhibition of the path.
• m: Changes the map view (forward).
• n: Changes the map view (backward).
• w, a, s, d: Moves the cames in the four directions.
• +, -: Changes the zoom.

• 1: Mades the robot start to follow the potential provided by the exploration pipeline.
• space: Changes to manual control and stop the robot.
• ↑, ↓: Mades the robot go forward and backward in the manual control mode.
• ←, →: Mades the robot turn around its own center in the manual control mode.