Recent Advances of Hybrid Intelligent Systems Based on Soft Computing. Studies in Computational Intelligence

Abstract

Path planning is an essential task in autonomous mobile robotics that demands to navigate following a minimum-cost path, which involves partitioning the landscape in nodes and the use of combinatorial optimization methods to find the optimal sequence of nodes to follow. Traditional algorithms such as the A* and Dijkstra are computationally efficient in landscapes with a reduced number of nodes. Most of the practical applications require to use a significantly large number of nodes up to the point that the problem might be computationally explosive. This work contributes to state-of-the-art with two heuristics for the A* algorithm that allows finding the optimal path in landscapes with a large number of nodes. The heuristics used the Euclidean and Manhattan distance in the estimation function. We present a comparative analysis of our proposal against the Dijkstra’s and A* algorithms. All the experiments were achieved using a simulation-platform specially designed for testing important algorithm features, such as the grid size, benchmark problems, the design of custom-made test sceneries, and others. Relevant results are drawn to continue working in this line.