MAGIC team members with robots

Anne Ju/Cornell Chronicle
Members of Cornell's Multi-Autonomous Ground-robotic International Challenge in the Autonomous Systems Lab with two of their robot prototypes. From left: graduate student Danelle Shah; Daniel Lee '10; lab manager Aaron Nathan, M.Eng. '08; Chuck Yang, M.Eng. '10; and graduate student Mark McClelland.


Anne Ju/Cornell Chronicle
The MAGIC team's latest prototype robot.

Scouring an earthquake site or other disaster area is treacherous for humans and often leads to more lives lost. In the quest to develop technologies to protect humans in such search-and-rescue operations, as well as other applications, a new team of Cornell researchers is creating a fleet of intelligent robots that will compete internationally.

In early November, Cornell was named one of 10 semifinalists in the first Multi-Autonomous Ground-robotic International Challenge (MAGIC), co-sponsored by the U.S. Department of Defense and the Australian Defense Science and Technology Organization. The competition requires teams to build various robots that can accomplish a series of yet undisclosed tasks that are likely to involve such skills as scanning for dangerous objects, protecting civilians and mapping for obstructions -- all autonomously, or with minimal human supervision.

As a semifinalist, Cornell received $50,000 from the organizing agencies to continue developing their robots and preparing for another round of cuts in June. Five finalists will compete in Australia in November 2010.

Assistant professor Hadas Kress-Gazit and associate professor Mark Campbell, both mechanical engineering, co-lead the project as part of the Autonomous Systems Lab, which also produces Cornell's DARPA Urban Challenge team that builds autonomously run vehicles.

"We took an interest in the research -- not just the competition," Kress-Gazit said of MAGIC. "Of course we want to win, but that's not the main drive."

The Cornell MAGIC team, consisting of about 20 people including faculty, graduate students, staff and some undergraduates, currently has four robot prototypes with more possible in the future. They are built on Segway platforms, and each will eventually be equipped with GPS units, laser range finders, cameras and inertial measurement units, which detect acceleration.

For the competition, the robots will operate in an "urban environment," both indoors and outdoors. In each environment, the robots must find, identify and correctly respond to stationary and moving "objects of interest." They will also be required to have mapping capabilities to create a record of the area.

Robotics has many potential applications ranging from military to humanitarian, the researchers say. A competition like MAGIC aims to find ways in which "one to two operators can control 50 robots to magnify their effectiveness," said graduate student Mark McClelland. "The swarm of robots can do their thing, while the humans do the high-level stuff that humans are good at."

Creating a fleet of autonomous robots means a host of technical challenges. For example: Robots use lasers to detect their environment, but sometimes the laser line is imperfect, sending false information. GPS systems can break. And if the robot's wheels have spun a certain number of times, that doesn't always translate to a certain distance, because such factors as terrain and environment can lead to faulty readings. The job of team members like graduate student Danelle Shah is to write algorithms to mitigate such errors, like the robot becoming confused about its location.

"It's better to know that you don't know where you are, than to think you do, and be wrong," Shah said.

The team aims to have its basic mapping software completed by the end of the semester and to eventually begin testing the robots in larger areas outside.