Map Fusion for Precise Yet Efficient Collaborative SLAM

Resumo

Collaborative Simultaneous Localization and Mapping (C-SLAM) is an active research area in robotics that aims to enable the collaboration of multiple robots in constructing a shared map and simultaneously estimating their positions. However, Map fusion poses a significant challenge, especially when involving a large group of robots. It aims at obtaining an accurate global representation of the environment. This paper proposes an novel approach using the Fourier Transform, the Pearson correlation coefficient and Particle Swam Optimization to address the map fusion problem. Efficiently merging maps into a global representation requires careful consideration of spatial relationships and alignment of these maps. The Fourier Transform analyzes spectral features in each robot’s measurements, extracting insights about spatial distribution. The Pearson correlation coefficient evaluates
spectral similarity between different map sections, facilitating region pairing for successful fusion. The search for
optimal fusion parameters is performed using the Particle Swarm Optimization Algorithm. These distinct regions
guide the fusion process, optimizing global map creation. Instead of a complete map fusion, selective fusion of sections increases the likelihood of success. Experiments involving five robots in a simulated environment validate the proposed approach, demonstrating the capability of optimized map fusion to provide a more accurate and comprehensive representation of the environment. This enhancement should contribute to refine further the
C-SLAM.