Page 15 - ME-06-E
P. 15
completed her master's in mechanical engineering at give robots the ability to change their hardness based Prof.
MIT, worked in the oil industry in Houston for a while, on the task they needed to perform, using controls in- Zacksenhouse
then returned to academia and completed her Ph.D. stead of hardware. and her
in electrical engineering at Rice University in Hous- team at the
ton, where she also did her post-doc. Her Ph.D. and "Another project we worked on at the lab was Sensory-Motor
post-doctoral research were in the field of neuronal developing bipedal robots that utilize their natural Integration
signal processing. When she returned to Israel, she dynamics with central pattern generators like the Lab. "We train
got her second post-doc position – this time in neu- ones humans and animals have. Central pattern robots to teach
roscience – at the Weizmann Institute. generators are biological mechanisms that drive the themselves
synchronized movement of our joints by producing the parameters
In 1995, she started her first lab at the Technion: rhythmic outputs with phase differences depending for a specific
the Sensory-Motor Integration Lab. "Back then," she on the required task; they are what enables us to do task, generalize
says, "we focused on developing control methods for things like spontaneously picking up our pace from what they'd
bio-inspired robots. We based our work on insights a walk to a run and vice versa." "When we start walk- learned, apply it
into humans and animal motor control. For example, ing," she explains, "we put one foot forward, and the to other tasks,
hardness control methods inspired by how humans dynamics of that movement pull us forward. The body and handle
control the hardness of their arm muscles. When has something like preset patterns or templates for uncertainties."
the arm is horizontal, you can change its hardness – walking. It selects one and activates it while automat-
meaning relax or harden the muscle – without chang- ically adjusting the parameters based on the exter-
ing the way it's positioned. This way, you can change nal conditions, like the slope, and requirements, like
how your arm responds to outside forces – whether the desired speed. "At our lab, we created two-step
it moves or resists. Humans change the hardness of central pattern generators that allowed for effective
their arms by simultaneously activating the muscles optimization, and developed machine learning tech-
that bend and straighten the arms. The idea was to niques for learning the required parameters, includ-
Faculty of Mechanical Engineering | MEgazine | 15
