Heterogeneous Map Fusion from Occupancy Grid Histograms for Mobile Robots

Abstract

Abstract With the increase in the capabilities of robotic devices, there is a growing need for accurate and relevant environment maps. Current robotic devices can map their surrounding environment using a multitude of sensors as mapping sources. The challenge lies in combining these heterogeneous maps into a single, informative map to enhance the robustness of subsequent robot control algorithms. In this paper, we propose to perform map fusion as a post-processing step based on the alignment of the window of interest (WOI) from occupancy grid histograms. Initially, histograms are obtained from map pixels to determine the relevant WOI. Subsequently, they are transformed to align with a selected base image using the Manhattan distance of histogram values and the rotation angle from WOI line regression. We demonstrate that this method enables the combination of maps from multiple sources without the need for sensor calibration.