A Novel Map-Merging Technique for Occupancy Grid-Based Maps Using Multiple Robots: a Semantic Approach

Abstract

Map merging is a noteworthy phenomenon for cases such as search and rescue and disaster areas in which the duration is quite significant when gathering information about an environment. It is obvious that the total mapping time decreases if the number of agents (robots) increases. However, the use of multiple agents leads to problems such as task allocation schemes and the fusing of local maps. Examining the present methods, it is generally observed that the common features of local maps have been found and the global map is formed by obtaining related transformation between local maps. However, such implementations may be risky when local maps have symmetrical areas. Hence, a novel and semantic approach has been developed to solve this problem. The developed method counts on the reliability level of feature points. If relevant feature points are trusted, local maps are merged according to the best point or points. The simulation results from a robot operating system and a real-time experiment support the proposed method’s efficiency, and mapping can be performed even for environments that have symmetrical similar parts and the task time can thus be reduced.