Computer simulations of path generation and path form modification with local rules working on a parallel cell-based architecture

Many mathematical and engineering algorithms are dealing with path finding and path planning problems which arise for a mobile navigating robot or the endeffector of a manipulator. Natural mobile systems, like animals navigating in their everyday surroundings, are faced with the same problems and normally generate adequate solutions. For this reason we performed psychometric experiments with human probands [1,2] to investigate both decision-making and path generation in the case of full information. In this paper we present several methods which can be used to generate path forms roughly comparable to those produced by the human probands. The computer simulations proposed here concentrate on algorithms which use local rules and work upon a parallel, cell-based architecture. On the basis of the 'dynamic-system metaphor,' we define a framework for the investigation of several algorithms. A diffusion-algorithm proposed in [3,4] is modified and combined with a so-called relative potential field. This diffusion in potential fields (DIP) provides a method to influence the smoothness of the forms of the path during the development of the gradient field. Once a path has been found, the relative potentials can also help to gradually modulate the first found path with a time-saving mechanism called Tube-Diffusion. In addition to the DIP-algorithm, we investigated two wave-spreading algorithms which generate cost-optimal paths. These are a further development of the path planning approach used in [5] and a new approach which we called Inclusive-Or-algorithm (IOR).