A Genetic Algorithm Using Changing Environments for Pathfinding in Continuous Domains

This paper describes a technique for finding paths through a two dimensional continuous space using a genetic algorithm. While A* currently dominates the field of pathfinding, it requires that the space be discretized. For spaces which are not easily discretized, or in which a fixed grid size is not permissible, a pathfinding algorithm which does not require discretization is necessary. In addition to not requiring the discretization of the search space, a genetic algorithm employs a depth-first strategy which may yield suitable paths early in the search. This algorithm dynamically morphs the search space from an easy space into a difficult space to reduce the number of generations required to find a solution to the original space. The algorithm is demonstrated on a small set of sample environments consisting of polygonal obstacles.