N-Sliced Informed RRT*: Intelligent Sampling-Based Path Planning In High Eccentricity Informed Ellipsis
Giray Uzun, Aykut Özdemir, Seta O. Bogosyan · 2022 IEEE 31st International Symposium on Industrial Electronics (ISIE) · 2022
Many algorithms have been developed based on the Rapidly-exploring Random Tree (RRT). Among them, the Informed-RRT* stands out, because it aims to sample only a particular elliptical sub-region of the map to facilitate the optimal path convergence process. This way, it can find an optimal path in a shorter time by reducing the search area. However, this algorithm loses its benefits and acts similar to the the RRT* algorithm on multi-curve paths. In this study, we propose the n-Sliced Informed RRT*(nSIRRT*) to address this issue by dividing the reference path into a certain number of path segments and optimizing these segments one by one. Using this approximation, we addressed the convergence problem of Informed-RRT* in multi curved paths. Results showed that our method is more sample-efficient and converges the optimal trajectory faster than Informed-RRT* method. This method optimizes the path by focusing the sampling area around small segments on the path. Therefore generated trajectory converges optimal results faster.