One area that is often neglected, and could probably use more study, in the area of "robot/human interaction" is that of "face to face" issues. Note that this could tread into the "uncanny valley" arena. There are many questions that could be asked and pursued in this; you could, for instance build a robot head that had a repertoire of facial expressions - perhaps you could figure out a way to have such a device mimic your own expressions (or the expressions of others?). Maybe you could study how humans perceive machines with such expressions, and how to make it easier for humans to interact with such machines? Why and how can a machine cause an emotional response in a human (we know this is possible), and can this make interaction better, or more intuitive (or easier) in some manner?
Think about how people perceive robots - note how both R2D2 and C3PO from Star Wars fame elicit emotional responses from the audience, in such different manners. Also - to an extent, iRobot has embodied similar traits into its Roomba (use of sound alone to elicit emotional responses?). What about Johnny 5 from the movie "Short Circuit", or Wall-E?
This is all fairly "virgin" territory; there has been some exploration in it, but not nearly enough...
