Simulation of the process of building and merging terrain maps by a group of autonomous robot agents

Abstract

The article discusses one of the problems of multi-agent systems, namely the integration of obstacle grid maps created by various robots exploring different parts of the same environment. Simultaneous localization and Mapping (SLAM) is a well-known field of research that attracts considerable attention because of its enormous practical importance. However, most research focuses on the problem of building one common map either based on data coming from one robot or from several autonomous robot agents. In many situations, this approach is impractical. In addition to the problems arising from the dimension of the compiled map increase, when robots explore large spaces, communication between agents can be available only at rare moments of their convergence in space.

The paper presents a fast and accurate algorithm for combining several maps built by agent robots forming a self-organizing episodic network, and a variant of its software implementation. The algorithm is used to combine maps presented as an obstacle map. It is deterministic, so its results are repeatable, and the calculation time is predictable. The algorithm calculates several functions describing the characteristic features of the maps to be combined, and builds a combined map based on them. Despite the fact that the algorithm is deterministic, it can be used in a probabilistic structure.