Dancing With Robots -- Robotic Couriers Can Save Hospitals Money

The original study was conducted in 1998 at the University of Virginia Health Sciences Center, a 591-bed facility with an average daily census of 454. In FY1997 the center cared for 29,189 inpatients and 514,307 outpatients. Usually, five human couriers were used to deliver specimens and pharmaceuticals on each of three 8-hour shifts. In the study, a fleet of six HelpMate mobile robots were used to make the deliveries.

"These are fully autonomous robots that can carry out deliveries between hospital departments and nurses stations," explained Rossetti, assistant professor of industrial engineering. "They have a map of the hospital stored in memory and can communicate with humans by voice output."

Because the robots must use elevators to move between floors, they do pose some additional load on the elevator system. This is compounded by the robot programming, which prevents a robot from being on an elevator at the same time as a human.

Robots follow a specific assigned route and must be spaced out properly, since they compete with each other for the elevators. They can navigate independently and have sensors that allow them to avoid obstacles and people.

Early in the development process for the robots, Helpmate engineers noticed an interesting phenomenon. The robots are programmed to navigate around an obstacle. In interacting with people, the robot would recognize a person, stop, pause and navigate to the side. The person would see the robot, stop, pause and shift to the side at exactly the same time.

"This created an interesting spectacle, of people and robots dancing side to side in the hospital halls," laughed Rossetti. "Of course, this was easily resolved by delaying the robots reaction time."

Rossetti used the data from this experiment to develop a simulation that compared the robotic delivery system with a human courier system. These models are very complex and include large amounts of very inhomogeneous data derived from a variety of sources ranging from hospital blueprints to OTIS elevator repair statistics.

In addition, researchers gained a lot of data from actual observation. They recorded data such as the actual numbers of people using the elevators each hour and the time it took for personnel to load and unload specimens and pharmacy supplies.

The 15 human couriers averaged $21,344 per year, including benefits, resulting in an annual cost of $448,228 for the human delivery system. Although the initial cost and installation of the robotic system was $612,500, annual operating cost was $154,500.

Rossetti also simulated two introduction scenarios. If the robot system was completely installed at the beginning, it would pay for itself in 2 years and 5 months; if it were installed gradually over 4 years, payback time increased to 3 years and 6 months.

On the basis of these results, Rossetti concluded that the mobile robot fleet was a valid alternative to the human couriers. He will present his findings at the Industrial Engineering Research Conference on May 21-23 in Cleveland.