Our research on a mobile robotic assistant for the factory floor and prediction of human motion has been highlighted on "Meet Baxter, your new robotic co-worker"

Read the full story at http://www.csmonitor.com/USA/2014/0927/Meet-Baxter-your-new-robotic-co-worker

"... Automotive factories have relied on automated machines to do much of the heavy lifting since the 1960s, but again, those robots tend to weigh hundreds, even thousands, of pounds and move so fast that they could easily run a person over.

Foremen at a BMW factory contacted Shah’s group inquiring if the scientists could develop a robot that could scoot around the factory floor delivering heavy parts to people working on a moving assembly line.

Vaibhav Unhelkar, a PhD student working in Shah’s lab, took on the challenge and is currently working to adapt a stout robot, aka Robbie, into a strong but sensitive workhorse that can both assist and look out for its human co-workers.

Mr. Unhelkar equipped Robbie with laser sensors that can detect an upcoming obstacle, and he programmed the robot to stop short of crashing, even if it receives a remote-control command to continue driving forward.

For people and robots to truly collaborate – in either the factory or the home – robots need to be able to adapt to the presence of people and carry on working. For that to happen, robots need to not only sense a person’s presence, but also formulate some prediction about what the person might do next.

People are very good at picking up on subtle cues in others’ body language that indicate where they might go next. That intuition is formed by years of experience watching other people move. A robot’s experience is limited to what people have written into its program.

Claudia Pérez D’Arpino, another PhD candidate in Shah’s lab, is working to compile a digital library of people’s movements using a motion-capture system, in which cameras register and catalog the movements of volunteers placing toy blocks into one of four boxes. Ms. Pérez D’Arpino wrote a program that can predict which box a person is moving toward just 4/10ths of a second after the person begins to move – with 80 to 90 percent certainty.

Pérez D’Arpino’s and Unhelkar’s projects each represent one small piece of a greater robotics puzzle. Thanks to the collaborative nature of the open-source ROS (robot operating system) developed by robotics pioneer Willow Garage, their contributions can be combined with those of roboticists around the world ..."