Semantic Mapping for Safe and Comfortable Navigation of a Brain-Controlled Wheelchair.

This paper presents a novel navigation system designed for a brain-controlled wheelchair, which interacts with human user by the low throughput interface. The navigation system proposes the semantic map, which is integrated with the navigation points, semantic targets and local 3D map, to a human user who can choose one of the navigation points as a goal for navigation. The semantic targets provide category, geometry and functionality information of the recognized objects, such as a table which can be docked. The local 3D map provides the navigation points in the traversable areas. The human-wheelchair interactive system shows the semantic map to user, and the user selects the goal via a brain-computer interfaces BCI. Therefore, this method can help the wheelchair implement accurate navigation e.g. docking with a low throughput interface and the safety and comfortability are improved. Our navigation system is successfully tested in real environment.