A Non-Uniform Quadtree Map Building Method Including Dead-End Semantics Extraction

Xiuzhong Hu · Guangming Xiong · Junyi Ma · Gege Cui · Quanfu Yu · Shihao Li · Zijie Zhou

Beijing Institute of Technology

Our work has been accepted by Green Energy and Intelligent Transportation. [paper link](https://www.sciencedirect.com/science/article/pii/S2773153723000075)

0. Dependency & Install

Dependency

Our program was tested on ubuntu 18.04, ros-melodic. Related packages are: pcl-1.8, eigen3, opencv, Ceres, etc.

Install

cd xxx/Non-uniform_quadtree_map_master
catkin_make # or catkin_make -jn，n is the number of cores in the computer's cpu.

1. Non-uniform quadtree map building including dead-end_semantic_information_extraction

To improve the application transferability, we adapt the application input to the common ROS topic nav_msgs/OccupancyGrid i.e. grid map. A sample grid map is shown below.

1.1 Non-uniform quadtree map building

Publish 2D raster maps, and convert 2D maps to non-uniform quadtree maps.

cd xxx/Non-uniform_quadtree_map_master
source ./devel/setup.bash
roslaunch ./launch/non-uniform_quadtree_map.launch

1.2 Non-uniform quadtree map building including dead-end_semantic_information_extraction

Publish 2D grid map; Extract dead end semantic information from 2D grid map, and convert grid map to 2D dead end semantic map on this basis; Convert 2D map to non-uniform quadtree map.

cd xxx/Non-uniform_quadtree_map_master
source ./devel/setup.bash
roslaunch ./launch/non-uniform_quadtree_map_with_dead-end_semantics.launch

2. map matching on non-uniform quadtree map

cd xxx/Non-uniform_quadtree_map_master
source ./devel/setup.bash
roslaunch ./launch/map_match_quadtree_kitti.launch

The map matching effect is shown below, where white is the point cloud corresponding to the offline map and the other colors are the point clouds corresponding to the online map.

3. Global and local path planning

3.1 Publish map

cd xxx/Non-uniform_quadtree_map_master
source ./devel/setup.bash
roslaunch ./launch/non-uniform_quadtree_map_with_dead-end_semantics.launch

3.2 Receive map topics and run planning programs

Planning related code involves other work, so no details will be released for now. After receiving the map topic and setting the start and end points, global planning based on A-star algorithm and local path planning based on DWA are initiated. The planning result is as follows.

The zoomed-in effect on the left side is shown below.