Multi-Grouped Grid Map for Semantic-Aware Motion Planning of Mobile Robots

In current navigation systems, path planning is performed on a grid map considering the presence of obstacles in the environment. Although this approach can produce collision-free paths, its performance cannot be guaranteed in dynamic, human-centered environments. The reason for this is that more factors need to be considered, including people’s perceptions, than simply ensuring that one gets from one place to another without colliding with an obstacle. Therefore, the ideal map should be able to manage various types of semantic information in a flexible manner, in addition to tracking obstacles. We propose a multi-group map based on a multi-layer map, which is characterized by allowing the layers to store multiple types of information and be flexibly combined to produce multiple sub-maps that can provide references for semantic navigation. The proposed map has been evaluated through simulation to determine its potential to contribute to the creation of semantic navigation behaviors.