Path planning and collision avoidance for UAVs using 3D A* algorithm and artificial potential field technique
Seungchan Shin, Sangho Ko · 2019
In this paper, we developed obstacle collision avoidance method using 3D A* algorithm and artificial potential field (APF). The 3D A* algorithm creates a global path that directs the final destination while avoiding unmanned aerial vehicle (UAV) collisions, and APF creates a local path that avoids surrounding obstacles in real time. The 3D A* algorithm verifies static obstacle information using the map generated from real-time appearance-based mapping (RTAB-Map). Then, a 3D evaluation function is obtained to the destination to generate a global path which is the shortest flight path. The RTAB-Map uses the depth camera to generate the Odometry and the surrounding terrain map of the UAV. The RTAB-Map maps the surrounding environment without recognizing objects that are not stationary. Therefore, it does not have information about moving objects in real time. APF was used to supplement this. The APF is used to avoid any dynamic or static obstacles. In this paper, we use Gaussian mixture model (GMM) based APF. All obstacles data observed on 2D-LiDAR create and update GMM in real time. The UAV is subjected to an artificial force that is pushed by the slope obtained by partial differentiation of the GMM generated on the 2D-LiDAR plane. Therefore, the UAV will fly on the local path out of the global path. The obstacle collision avoidance simulation was carried out using the control law of the open source PX4. The actual UAV test used Pixhawk to implement the PX4 firmware and the Robot Operating System (ROS) to implement the collision avoidance algorithm.