MPC-MF: Multi-Point Cloud Map Fusion Based on Offline Global Optimization for Mobile Robots

This paper proposes a Multi-Point Cloud Map Fusion (MPC-MF) approach based on offline global optimization for mobile robots. To realize the effective map fusion of multiple point clouds an offline global optimization approach is firstly proposed based on self constraint generation and mutual constraint generation. Different from the simple addition of fusing multiple maps to generate final complete map this work makes use of mutual constraint factors between two adjacent sessions to eliminate inconsistencies in the fused map. Note that the factor generation of mutual constraint and the map fusion of global optimization are main contributions of this work. To evaluate the effectiveness of the proposed map fusion approach three independent sessions have been collected in a real environment. Experimental results on multi-trajectory fusion and multi-map fusion show superior performance of our proposed approach on eliminating local inconsistencies and realizing effective fusion.