Multimodal cognitive interface for robot navigation.

To build effective interactions between humans and robots, they should have common ground of understanding that creates realistic expectations and forms the basis communications. An emerging approach to doing this is to create cognitive models of human reasoning and behavior selection. We have developed a robot navigation system that uses both spatial language and graphical representation to describe route-based navigation tasks for a mobile robot. Our proposed route instruction language (RIL) is intended as a semi-formal language for instructing the robot to execute a route in an indoor environment. We implemented an instruction interpreter to process the route description and generate its equivalent symbolic and topological map representations. A topological map is generated to describe relationships among features of the environment in a more abstract form without any absolute reference system to treat the ambiguity which can occur when the robot cannot recognize the current landmark. The symbolic and topological map representations are supplied to other system components as an initial path estimation to guide the robot while it plans its navigation task. We conducted some experiments to evaluate the routes which are written by using the RIL instructions.