The obstacle_detector package

The obstacle_detector package provides utilities to detect obstacles from a 2D laser scan or an ordered point cloud. Detected obstacles come in a form of segments and circles. The package requires Armadillo C++ library for compilation and runtime .

1. The nodes

The package contains several nodes, which are:

The obstacle_detector node

The main node which converts messages of type sensor_msgs/LaserScan from topic scan or messages of type sensor_msgs/PointCloud from topic pcl into obstacles, which are published as messages of custom type obstacles_detector/Obstacles under topic obstacles. The point cloud message must be ordered in the angular fashion, because the algorithm exploits the poolar nature of laser scanners. The node is configurable with the following set of local parameters:

• world_frame (string, default: world) - name of the global coordinate frame,
• scanner_frame (string, default: scanner) - name of the local coordinate frame (in which the points are expressed),
• use_scan (bool, default: true) - use laser scan messages,
• use_pcl (bool, default: false) - use point cloud messages (if both scan and pcl is chosen, scan will have priority),
• transform_to_world (bool, default: true) - choose whether the obstacles should be published described in the local or global coordinate frame,
• max_scanner_range (double, default: 5.0) - limitation on laser scanner range (in meters),
• max_x_range (double, default: 2.0) - limitation on global coordinates (obstacles detected behind these limitations will not be published),
• min_x_range (double, default: -2.0) - as above,
• max_y_range (double, default: 2.0) - as above,
• min_y_range (double, default: -2.0) - as above.

The following set of local parameters is dedicated to the algorithm itself:

• use_split_and_merge (bool, default: false) - choose wether to use Iterative End Point Fit or Split And Merge algorithm to detect segments,
• min_group_points (int, default: 3) - minimum number of points comprising a group to be further processed,
• max_group_distance (double, default: 0.100) - if the distance between two points is greater than this value, start a new group,
• distance_proportion (double, default: 0.006136) - enlarge the allowable distance between point proportionally to the range of point,
• max_split_distance (double, default: 0.100) - if a point in group lays further from a leading line than this value, split the group,
• max_merge_separation (double, default: 0.200) - if distance between obstacles is smaller than this value, consider merging them,
• max_merge_spread (double, default: 0.070) - merge two segments if all of their extreme points lay closer to the leading line than this value,
• max_circle_radius (double, default: 0.300) - if a circle obtained from a group would have greater radius than this value, skip it,
• radius_enlargement (double, default: 0.050) - enlarge the circles radius by this value.

The obstacle_visualizer node

The auxiliary node which converts messages of type obstacles_detector/Obstacles from topic obstacles into Rviz markers of type visualization_msgs/MarkerArray, published under topic obstacles_markers.

The scans_merger node

The auxiliary node which converts two laser scans of type sensor_msgs/LaserScan from topics front_scan and rear_scan into a single point cloud of type sensor_msgs/PointCloud, published under topic pcl. The scanners are assumed to be mounted in the same plane, back-to-back (rotated 180 deg) with some separation betweend them. The node uses following local parameters:

• pcl_frame (string, default: base) - name of the coordinate frame used for the origin of produced point cloud,
• max_unreceived_scans (int, default: 1) - if one of the scanners stopped providing scans, after this number of missing scans the node will switch from merging to copying points directly,
• omit_overlapping_scans (bool, default: true) - if some of the points provided by both scans exist on the same angular area, omit them,
• scanners_separation (double, default: 0.45) - the distance between centres of both scanners.

The static_scan_publisher node

The auxiliary node which imitates a laser scanner and publishes a static, 360 deg laser scan of type sensor_msgs/LaserScan under topic scan.

2. The messages

The package provide three custom messages:

• CircleObstacle
  • geometry_msgs/Point center
  • float64 radius
• SegmentObstacle
  • geometry_msgs/Point first_point
  • geometry_msgs/Point last_point
• Obstacles
  • Header header
  • obstacle_detector/SegmentObstacle[] segments
  • obstacle_detector/CircleObstacle[] circles

3. The launch files

Provided launch files are good examples of how to use obstacle_detector package. They give a full list of parameters used by each of provided nodes.

• single_scanner.launch - Starts a single hokuyo_node to obtain laser scans from Hokuyo device, a laser_scan_matcher_node to provide appropriate transformation from global to local coordinate frame, obstacle_detector and obstacle_visualizator nodes, as well as rviz with provided configuration file.
• two_scanners.launch - Starts two hokuyo_nodes, assuming that the udev configuration provides links to both devices (if not, change the devices names to /dev/ttyACM0 and /dev/ttyACM1 appropriately), provides appropriate transformations with static_transform_publishers, uses scans_merger to convert both scans into pcl, and runs obstacle_detector, obstacle_visualizator as well as rviz.
• virtual_scanner - Used for debug and tests purposes. Starts a static_scan_publisher, provides global to local transformation and runs obstacle_detector, obstacle_visualizator and rviz.