Natural Language and Robot Planning, Part 2: An Empirical Approach This project uses a robot which has incomplete knowledge of its environment and must acquire additional knowledge through sensors to carry out human commands, allowing the exploration of issues in robotic planning and control and natural language interaction with an active agent. The speaker giving the directions is provided with a three-dimensional graphic display that provides knowledge of how the robot is positioned within the building and how it is responding to the commands. The approach to developing the human- robot system will be to have humans work in pairs, with each acting as the robot in turn. In this way, the syntactic and semantic aspects of the way humans react in this situation and their interaction with the situation pragmatics can be studied. Naturally, the directions given depend upon and will be interpreted within the context of the robot at a given time, so the task is designed to provide a concrete laboratory for the study of the problems of context which are difficult enough to raise important issues but easy enough to expect insights.