Once an interaction is established between the robot and a target user, how does the robot figure out exactly what it is the user is trying to accomplish? 

We initially imagined the robot doing a simple probing of the user by asking a question (like “Can I help you?”) and listening for keywords (like room numbers, times, or event names). This was a little too open-ended, so we fine-tuned the idea to have the robot ask a similar probing question, but also give the user specific options. This way, the user knows exactly what the robot is for, the robot doesn’t have to listen for too many words and phrases, and the whole process is generally more efficient.