Improved Frontier Exploration Strategy for Active Mapping with Mobile Robot

摘要

Map learning is an important ability for mobile robots. There are two approaches to map learning: passive and active. Frontier exploration is an active approach to map learning. This approach builds frontiers based on the boundary between free and unknown spaces. It then selects the nearest frontier as the next exploration goal. When using this technique, often the path taken is less optimal. It is because sometimes there are more than one frontiers with relatively similar distances or the frontier is just not significant enough to be visited. To solve this problem, an additional metric is needed to evaluate the frontiers. This paper presents an improved strategy for frontier exploration that additionally accounts for the potential information gain of each frontier which is calculated heuristically based on a Kullback-Leibler divergence when selecting the next exploration goal.