An information theoretic approach to path planning for frontier exploration

Efficient path planning for autonomous robots to explore unknown areas is a critical area of research due to the requirements on mission times and problems that dictate as quick a solution as possible e.g. search and rescue scenarios. It also proves to be a difficult problem to solve due to its inherent NP-hard nature, which requires that an optimal (albeit not necessarily perfect) path is defined based on a set of defined principles. This gives rise to a wide variety of logical solutions. This paper proposes an information theoretic addition to the well-established Frontier Exploration in order to build a 2D spatial map of an area of interest as intelligently as possible. The informative method is then compared with a greedy approach toward information gain, as well as the traditional ‘nearest frontier’ approach to frontier exploration. The proposed method is shown to outperform other methods in terms of the total number of actions required to resolve the map, as well as being consistently the quickest method of reducing map entropy throughout the mapping procedure. We also discuss how, by exploiting an information theoretic framework, other quantities of interest can be mapped efficiently alongside a spatial map.