Current robot hands emulatethe structure of human hands, but they are far from dexterous. This researchaimed to define the ways that tactile information can improve robot dexterity.One notable achievement in this area was the development of high-frequencytactile sensing. This work established that vibratory information can signalimportant events such as the first instant of contact and the onset of slip.These events convey information about the state of the hand-object systemthat is essential for robust control of manipulation. Vibrations also provideperceptual information about properties such as surface texture and friction.This research effort included the development of new tactile sensing devices and signal processing algorithms, correlation of tactile phenomena withtask attributes, and the use of this information in control of manipulation.
We have studied a number of additional areas in contact sensing inmanipulation. We have conducted arigorous experimental evaluation of an object stiffness controlalgorithm on a multifingered hand; these results and our analysis showhow tactile sensing can improve performance. Another project derived practical modelsof the frictional mechanics of a sliding finger, which is essentialfor planning many manipulation tasks. A newer effort combined computer vision and tactile sensing to permit gentlegrasping of arbitrary objects in unstructured environments.
References:
Smooth Bipedal
Walking
This project was directed at determining thesensing and
control strategies that would permit a legged robot to carry apayload over rough
ground as smoothly as a wheeled vehicle rolling over aflat road. Kinematic
analysis has provided a criteria for smooth transferof support at footfalls.
Other results include algorithms for controllingfoot placement and forward
velocity while maintaining smoothness. We have experimentally tested these
algorithms on the planar biped robot inour laboratory.
References