An Integrated RRT*SMART-A* Algorithm for solving the Global Path Planning Problem in a Static Environment

• Main Authors: Suwoyo, Heru, Adriansyah, Andi, Andika, Julpri, Ubaidillah, Abu
• Format: Article
• Language: English
• Published: IIUM 2023
• Subjects: T Technology (General)
• Online Access: http://eprints.uthm.edu.my/8829/1/J15663_95d4320a330786e807526f22c3ad04c4.pdf; http://eprints.uthm.edu.my/8829/; https://doi.org/10.31436/iiumej.v24i1.2529
• Institution: Universiti Tun Hussein Onn Malaysia

The use of sampling-based algorithms such as Rapidly-Exploring Random Tree Star (RRT*) has been widely applied in robot path planning. Although this variant of RRT offers asymptotic optimality, its use is increasingly limited because it suffers from convergence rates, mainly when applied to an environment with a poor level of obstacle neatness and a narrow area to the target. Thus, RRT*-Smart, a further development of RRT*, is considered ideal for solving RRT* problems. Unlike RRT*, RRT*-Smart applies a path optimization by removing the redundant nodes from the initial path when it is gained. Moreover, the path is also improved by identifying the beacon nodes used to steer the bias of intelligent sampling. Nevertheless, this initial path is found with termination criteria in terms of a region around the goal node. Consequently, it risks failing to generate a path on a narrow channel. Therefore, a novel algorithm achieved by combining RRT*-Smart and A* is proposed. This combination is intended to switch method-by-method for the exploration process when the new node reaches the region around the goal node. However, before RRT*-Smart is combined with A*, it is improved by replacing the random sampling method with Fast Sampling. In short, by involving A*, the exploration process for generating the Smart-RRT*’s initial path can be supported. It gives the optimal and feasible raw solution for any complex environment. It is logically realistic because A* searches and evaluates all neighbors of a current node when finding the node with low cost to the start and goal node for each iteration. Therefore, the risk of collision with an obstacle in the goal region is covered, and generating an initial path in the narrow channel can be handled. Furthermore, this proposed method's optimality and fast convergence rate are satisfied.