On the computation of Delaunay triangulations via genetic algorithms

In this work, we introduce a new approach for computing Delaunay triangulations. Delaunay triangulations have numerous applications in geolocation, communications and other ICT systems or practical applications. Having available various types of approaches for computing such structures is rather desired from an implementation and computational point of view. We adopt Genetic Algorithms for computing Delaunay triangulations and present the design and evaluation of our novel approach. We consider a set of points in the plane as vertices and connect them with edges, creating the point graph. We have developed in C++ an application framework based on genetic algorithms, called Delaunay_Genetic, which produces the Delaunay triangulation structure of a given set of points in the plane. Delaunay_Genetic considers a novel graph-based chromosome representation of desired solutions, creates an initial population of individuals (chromosomes), an initial generation, and produces from the original population (generation) new generations of individuals in each repetition of the genetic process of Reproduction. Each new generation emerges more robust than the previous one. Our evaluations have revealed that the Delaunay triangulation yielded by Delaunay_Genetic, achieves an accuracy of 98–100% of the optimal Delaunay triangulation, while maintaining good convergence speed. Despite its limitations in computational time and space, the proposed novel approach exhibits several complementary benefits to computational geometry based approaches, such as allowing the insertion of new points in the triangulation dynamically, leading to seamless adaptation to new conditions, parallelization of the computational process and tolerance to noise regarding the coordinates of the points. Therefore, this work provides a useful alternative approach for computing Delaunay triangulations.