Fuzzy Greedy RRT Path Planning Algorithm in a Complex Configuration Space

Ehsan Taheri,Mohammad Hossein Ferdowsi,Mohammad Danesh

DOI: https://doi.org/10.1007/s12555-018-0037-6

2018-10-30

Abstract:<h3 class="a-plus-plus">Abstract</h3> <p class="a-plus-plus">A randomized sampling-based path planning algorithm for holonomic mobile robots in complex configuration spaces is proposed in this article. A complex configuration space for path planning algorithms may cause different environmental constraints including the convex/concave obstacles, narrow passages, maze-like spaces and cluttered obstacles. The number of vertices and edges of a search tree for path planning in these configuration spaces would increase through the conventional randomized sampling-based algorithm leading to exacerbation of computational complexity and required runtime. The proposed path planning algorithm is named fuzzy greedy rapidly-exploring random tree (FG-RRT). The FG-RRT is equipped with a fuzzy inference system (FIS) consisting of two inputs, one output and nine rules. The first input is a Euclidean function applied in evaluating the quantity of selected parent vertex. The second input is a metaheuristic function applied in evaluating the quality of selected parent vertex. The output indicates the competency of the selected parent vertex for generating a random offspring vertex. This algorithm controls the tree edges growth direction and density in different places of the configuration space concurrently. The proposed method is implemented on a Single Board Computer (SBC) through the xPC Target to evaluate this algorithm. For this purpose four test-cases are designed with different complexity. The results of the Processor-in-the-Loop (PIL) tests indicate that FG-RRT algorithm reduces the required runtime and computational complexity in comparison with the conventional and greedy RRT through fewer number of vertices in planning an initial path in significant manner.</p>