Vincent Cross is a Computer Science PhD candidate at Auburn University. He is currently conducting research in Human Robot Interaction (HRI). HRI is a multidisciplinary area of research that explores how we can improve the way robots meet our needs. Specifically, Cross’ research focused on how a single operator can control a team of robots and how they can do it efficiently without being overloaded with information, making the task more difficult.
Cross speaks on various robot interfaces, comparing multiple interfaces with a single combined interface. He mentions a couple of problems associated with multiple interfaces. One problem is situational awareness, this is the operator’s awareness of the current situation that is being focused on. With multiple interfaces, there are greater chances that the operator may become confused. Another issue is that multiple interfaces puts the operator in a reactive state vs a proactive state, meaning, the operator would be reacting to whatever is seen, instead of planning the next move.
Cross is researching a few issues regarding these interfaces. Interface design, if the design of the interface is poor, the operator will have trouble completing the task at hand. He is also researching the lack of situational awareness associated with various interfaces and, the operator interaction with robots will take part in his research, as well. Cross took elements from various sources, some including a job he had working security for Best Buy with their video cameras, also, from split-screen gaming.
While doing his research, Cross designed three interfaces. They differed in the actual layout of the interface, as well as, speech recognition, varying video feeds, and controls.
To better his research, he designed an experiment. This experiment compared a “focus” view, a “monitor” view, and a “single” view interface. Within these views, he compared the time taken to complete the task, the task completion rate, and the work load. The task of the experiment consisted of a user using one of the interfaces with four, five, then six robots. They had to find three bombs, and they had to diffuse all three bombs. His independent variables were the interface and the number of robots. The dependent variables were the increments in response time and task completion rate.
The results of his experiment showed that users that used the monitor view interface completed the task the most, with the focus and single view interfaces following respectively. Users that used the monitor view interface performed the best, with very small differences in time, varying between the number of robots involved.
In future experiments, Cross would like to create more complex environments for the robots to explore. He would also like to include a virtual bread crumb trail system, where the robot knows where it has been so it does not re-explore an area. Also, Cross explained that he would like to involve heterogeneous robot teams instead of the robots being the same.
Link to presentation: SmartViewer: Efficient Supervision of Multi-robot Task Coordination

