starlightlwj / rrtstar-motion-planner-with-minimum-jerk-for-quadrotor

Quadrotor path planning using various sampling strategies on top of Informed RRT*.

Quadrotor path planning

This repo contains implementation of Informed RRT* path planning using various sampling strategies. The sampling strategies tested are Uniform sampling, Obstacle-based sampling, Gaussian sampling and Max clearance based sampling. The waypoints generated by RRT* are then converted into minimum jerk trajectory which can be safely executed by quadrotor. Simulator used is RotorS simulator and a custom world in gazebo is used to test planning. Different gazebo worlds can also be used, but the octomap should be constructed and then initialized in the program. The gazebo file used in the project is quad_gazebo_world which is made available in the repo. Also, download and setup RotorS simulator by following the instructions in its repo before running this project.

At first, the RotorS simulator is setup and run using the following launch file.

roslaunch rotors_gazebo mav_hovering_example.launch mav_name:=firefly world_name:=quad_gazebo_world

The path planning nodes then can be run using the following launch file,

roslaunch path_planning path_planning.launch

In order to run benchmarking for different types of sampling strategies, run the following launch file:

roslaunch path_planning benchmark.launch

Path planning visualization

Visualization of path planning of quadrotor from one point to other is shown below. Gazebo and Rviz visualizations are shown below. In Rviz, green one is orginal trajectory and red one is the jerk minimal trajectory. The planning time is restricted to just 5 seconds. When more time is allowed, the path obtained gets shorter and more smooth as well.