This is my implementation of the Rapidly Exploring Random Trees (RRT) algorithm to plan collision-free paths in a 2D environment with a ROS wrapper.
Based loosely on the algorithm here
Place it in your catkin workspace src folder and build it using:
catkin buildYou can run the launch file given in the launch folder using:
roslaunch rrt_planner rrt_planner.launchThis launch file automatically launches three nodes:
- RViz For visualization
- Map Server to load a map from a .png file and publish it as a
nav_msgs::OccupancyGridon the/maptopic - RRT Planner to receive a map, initial pose, and goal pose, and calculate and publish a collision-free path as a
nav_msgs::Pathmsg
There are 5 example map images in the resources directory can be used to test the RRT implementation.
The map server node is responsible for loading a map file and publishing it as a nav_msgs::OccupancyGrid on the /map topic.
To select the map file that should be loaded and published, configure the parameters in cfg/map.yaml file.
When a map has been loaded successfully it should be visible in RViz. The user can then set the initial pose and the goal pose through RViz.
Press the 2D Pose Estimate button in RViz to set the initial pose. Press the 2D Nav Goal button in RViz to set the goal pose.
Or you can provide the same through the topics /initialpose and /move_base_simple/goal respectively.
Change the parameters in the cfg/config.yaml file to configure the algorithm:
- Variation: the variantion of RRT we use for path finding
- 1: Bi Directional RRT algorithm
- Other numbers : Vanilla RRT algorithm
- Step_size: the max Euclidean distance for a new connection
- Delta: the interval distance to check for collision between 2 points of the connections
- Max_iteration: the max number of sample to do before concluding that no path can be found
