Map server

The map server builds a 2D occupancy grid of the world from the drone's sensors, for the path planner to plan on. Its plugin turns a laser scan into the grid, and two launch presets add a converter in front of it for depth cameras and point clouds.

Every option publishes the same two grids in the drone's namespace:

The map server needs the transform from earth to the sensor's frame, so the state estimator has to be running. Scans whose transform isn't available in time are dropped.

Running it

ros2 launch as2_map_server map_server.launch.py \
  namespace:=drone0 \
  plugin_name:=scan2occ_grid

Each option has its default configuration in the package. A file given as plugin_config_file is merged over it, so it only needs the values you change. Each value can also be passed as a launch argument, such as map_width:=600. Start the node through these launch files, since it has no usable defaults of its own for the size, the resolution and the range.

Scan to occupancy grid

scan2occ_grid traces every ray of the scan in the earth frame. Values go from 0 for free to 100 for occupied, with -1 for cells never seen:

• The first ray through an unknown cell marks it free, at 0.
• Each later ray through a cell removes miss_confidence.
• Each hit adds hit_confidence. With the defaults, one hit takes a cell to about 40 and three hits take it to 100.
• A cell that goes above 80 stays occupied for good.

Ranges beyond scan_range_max, and readings with no return, are cut to that length and count as free space.

The grid has a fixed size, centered on the origin of earth, and doesn't follow the drone. Hits outside it are dropped, and the planner can't plan to a goal outside it, so make it large enough for the whole flight area. Heights are ignored, since the grid is 2D.

The defaults give a 30 by 30 m map, 15 m to each side of the origin.

Depth camera and point cloud

depth2occ_grid and point_cloud2occ_grid are launch presets, not separate plugins. Each starts a ROS converter that turns its input into a laser scan on sensor_measurements/lidar/scan. It then starts the map server with scan2occ_grid.

Set the input topics on the launch line, since they aren't read from the configuration file:

ros2 launch as2_map_server map_server.launch.py \
  namespace:=drone0 \
  plugin_name:=point_cloud2occ_grid \
  cloud_topic_in:=sensor_measurements/lidar_3d/points

The configuration file holds the converter's parameters next to the scan2occ_grid ones. Both converters cut ranges at 10 m by default. The point cloud preset keeps points between min_height 0.0 and max_height 1.0, measured in the cloud's own frame rather than above the ground.

The default depth camera configuration sets the scan's frame to drone0/front_camera/rgbd_camera. For another drone or camera, set output_frame in your own file.