The gunman is barricaded in a small room with hostages. As he calculates his next move, he fails to notice that he is not alone--not by a long shot. A handful of robots the size of Palmetto bugs move in on him, navigating floors and furniture, scaling walls and ceilings, tunneling through the ventilation system. Before he realizes what is happening, the SWAT team storms into the room...
The above sounds like a scenario dreamed up by Steven Spielberg, but is actually created by a multidisciplinary team out of Michigan State University's College of Engineering. The team of six is collaborating on a three-year, $1.6 million grant from the Defense Advanced Research Project Agency, DARPA for short, to design and build adaptable, reconfigurable micro-robots for use in law enforcement, intelligence gathering, and search and rescue.
Lal Tummala, professor of electrical engineering and manufacturing, and project coordinator, says that in a scenario such as the one described above, law enforcement officials have only a few seconds between the moment they open the door and the time in which they act. Obviously, he says, the more information they have at the time of entry, the better off they are.
The U.S. Department of Defense wants to develop a means for safe and efficient fact-finding when the environment is dangerous or inaccessible to humans. The idea is that if very small robots--no bigger than five centimeters in diameter--are equipped with cameras, thermal and infrared sensors, and microphones, they can obtain and transmit useful information about a situation before a person is required to enter the scene.
"Possibly," says Tummala, "the robots could be dropped by helicopter or shot like bullets into a building. From there, they could go about their business, gathering information without notice."
In putting together their proposal, the MSU team brainstormed numerous design ideas based upon several open-ended criteria: that the robots be manufactured inexpensively so that a large number could be dispersed and left on site, that they have excellent maneuverability and that they have the ability to communicate robot-to-robot and thus coordinate their actions. Considerations like keeping power usage to a minimum and keeping the components small enough to fit inside the five-centimeter framework provided more concrete restrictions to the group's eventual design.
The group arrived at a bipedal caterpillar-like structure that could both slink along floors and rugged terrain as well as climb vertically on stairways and walls using its suction-cup feet. According to Tummala, the use of suction to climb walls, ceilings and even glass windows was unique to Michigan State's proposal.
Dean Aslam, associate professor in the Department of Electrical and Computer Engineering, is designing the suction cup feet--what he refers to as SRF's, short for "smart robotic foot." The SRF consists of a suction cup, a pressure sensor and a vacuum pump. The sensor, mounted inside the suction cup, will signal whether the pressure is at or below atmospheric pressure, if it is below, the motor in the vacuum pump will switch on, creating a vacuum within the cup.
Other distinguishing features of the micro-robots include the use of diamond coatings and sensors. Diamond coatings result in reduced friction between components, thereby lengthening battery life, and can be applied to even the most difficult-to-reach places.
Ranjan Mukherjee, associate professor of mechanical engineering, designs and builds the robot, while Ning Xi, assistant professor in the Department of Electrical and Computer Engineering, creates the task-driven controller, which maintains stability of the robot, and commands the direction of movement.
Sridhar Mahadevan, assistant professor in the computer science and engineering department, and John Weng, associate professor in the same department, will be assisting in the cognitive development of the robots.
Mahadevan will be responsible for designing a hybrid task planner, which will provide the main software interface to control the robot. The planner comprises a high-level strategic module that can be given specific tasks, such as finding a window on the second floor of the building and taking a picture, and a lower-level tactical system, which will provide basic reflex behaviors such as avoiding obstacles. Mahadevan also studies the use of group behaviors to coordinate the actions of multiple robots and examines how the robots' performance can be improved through reinforcement learning.
Weng applies the SHOSLIF technology that he and his students have developed to help the micro-robots "see" (through the use of micro-cameras) and to learn from those visual inputs. Weng also will investigate a new learning direction for robots, which he calls "developmental learning." Weng will be employing a developmental algorithm that allows the robot to learn as it experiences throughout its "life"--from "birth" through "adulthood"--in much the same way that humans learn.
MSU is one of 11 schools in the country to receive a grant under this DARPA program. A finished product will be delivered to the Department of Defense in May 2001.
Materials provided by Michigan State University. Note: Content may be edited for style and length.
Cite This Page: