Semantic enhanced navigation among movable obstacles in the home environment

An autonomous household robot has to be able to navigate through a variable environment and perform commonhousehold tasks. Additionally, in cluttered and narrow homes movement can become impossible unless obstacles are moved out of the way. Both challenges involve the manipulation of objects and a planning algorithm that can integrate the function of regions and objects to avoid the creation of new safety hazards during robot movement. We present a semantic detection method during path planning using a gridded semantic map to improve navigation among movable obstacles (NAMO) and support for simple household sub-tasks like cleaning a table and moving obstructing objects to another location. In our tests, the spatial planning was completed well within human reaction time, which is important for a natural interaction between a human and a robot.