Communications of the ACM
Robots Explore Morphological Computation
Pager motors and a genetic algorithm induce locomotion on this tensegrity robot.
Credit: Union College
Before your brain fully understands the object to be grasped, your fingers can adjust the tension required to lift the object with their tendons. John Rieffel, assistant professor of computer science at Union College, is studying morphological computation, where fingers or another mechanism act as a mind, and is exploring the phenomenon with tensegrity robots.
Made with only springs and rods, a tensegrity's shape is maintained through the balance of pushing forces (the rods) and pulling forces (the springs).
In Rieffel's lab, a tensegrity becomes a robot with the addition of small, vibrational motors, which cause the structure, designed by William Keat, associate professor of mechanical engineering, to resonate chaotically.
Depending on the voltage used, this resonance can move the robot forward, sideways, or in circles. While it's difficult to predict which voltage will produce which result, artificial intelligence techniques are helping Rieffel discover effective motions.
"The significant result is that we've made this robot move at all," he says. "As far as we know, it's the smallest, fastest tensegrity robot out there, and the only one that moves by vibrating."
Typical, non-tensegrity robots move deliberately and are built rigidly to house the large, heavy computers that control them. As a result, their weight often limits versatility.
Rieffel's creation would not be encumbered by such things.
Using just small motors and specific voltages, he hopes to develop a robot that might navigate any landscape. Its light-weight body could respond to obstacles or objects much like your fingers. Rieffel's tensegrity, still in early research stages, theoretically wouldn't rely so heavily on computers to tell it when and how to move.
"By outsourcing aspects of control and locomotion to a robot's body, we can use a robot's computational resources to perform more high-level tasks, like tracking objects or detecting survivors trapped in rubble," he says.
Rieffel's team's paper was accepted for publication and presentation at the European Conference on Artificial Life.
No entries found