A Study on Real-time Biased Tree Generation Based on RH-RRT*

This study proposes a biased tree generation method based on receding horizon-based rapidly-exploring random tree* (RH-RRT*) for an autonomous ground vehicle (AGV). The RH-RRT* improves memory efficiency and the degree of optimality of the path of the RRT* by employing tree removal and biased random node generation methods. However, the biased random tree of the RH-RRT* is based on the states of the AGV, and the collision check with obstacles does not consider the size of the AGV. This may degrade the overall performance of the AGV. The study proposes a biased tree generation method utilizing the global path, tree, and size of the vehicle to overcome these limitations. The obtained biased tree is employed for the local path generation of the AGV. Additionally, the local path is utilized to create the biased tree at the next step. This procedure makes the proposed method more efficient and optimal than the existing method. The results obtained from the robot operating system (ROS) and Gazebo-based simulations were analyzed to validate the performance of the proposed method.