Hybrid A Star

1. Introduce

Hybrid A Star achieved very good results in the DARPA Urban Challenge. And after years of development, the algorithm has been verified to perform very well in parking and some autonomous driving environments without driving rules.

Since the author did not open source the algorithm, it took me weeks to reproduce the algorithm. The final result is close to the speed and effect in the paper.

Video 0: bilibili | YouTube

Video 1: bilibili | YouTube

Case 1

img0

Case 2

img3

Case 3

img1

Case 3

img2

2. Prerequisites

(1). C++11 or C++0x Compiler

(2). ROS Melodic: I’m developing on the melodic version. Using other versions of ROS, it should also work normally

sudo apt-get install ros-melodic-costmap-*
sudo apt-get install ros-melodic-tf

(3). Eigen 3

sudo apt-get install libeigen3-dev

(4). glog

sudo apt-get install libgoogle-glog-dev

3. Build Hybrid A Star

# new folder
mkdir hybrid_a_star_ws/src -p
cd hybrid_a_star/src

# Clone the repository
git clone https://github.com/zm0612/Hybrid_A_Star.git

# Build
cd hybrid_a_star_ws
catkin_make

4. Run Hybrid A Star

source devel/setup.bash
roslaunch hybrid_a_star run_hybrid_a_star.launch

The starting point is selected by the 2D Pose Estimate in rviz, and the end point is selected by the 2D Nav Goal.

I provide multiple maps, you just need to modify the variable image in hybrid_a_star/maps/map.yaml file.

5. References

(1). Practical Search Techniques in Path Planning for Autonomous Driving

(2). The Bresenham Line-Drawing Algorithm

(3). A simple car

GitHub

View Github