Our research group focuses on the role of sensing and mechanical design in motor control, in both robots and humans. The principal scientific goal is to find the fundamental ways that contact sensing contributes to dexterity. This work draws upon diverse disciplines, including controls, solid mechanics, systems analysis, biomechanics, and neurophysiology. The main approach is experimental, although analysis and simulation play important parts. In conjunction with industrial partners, we are developing applications of this research in biomedical instrumentation, teleoperated robots, and intelligent sensors.
Note: See publications page and individual students' pages for on-line access to many of the papers cited.
Robotic manipulation and tactile sensing. Current robot hands emulate the structure of human hands, but they are far from dexterous. This research aims to define the ways that tactile information can improve robot dexterity. One notable achievement in this area is the development of high-frequency tactile sensing. This work establishes that vibratory information can signal important events such as the first instant of contact and the onset of slip. These events convey information about the state of the hand-object system that is essential for robust control of manipulation. Vibrations also provide perceptual information about properties such as surface texture and friction. This research effort includes the development of new tactile sensing devices and signal processing algorithms, correlation of tactile phenomena with task attributes, and the use of this information in control of manipulation.
We are studying a number of additional areas in contact sensing in manipulation. We have conducted a rigorous experimental evaluation of an object stiffness control algorithm on a multifingered hand; these results and our analysis show how tactile sensing can improve performance. Another project derived practical models of the frictional mechanics of a sliding finger, which is essential for planning many manipulation tasks. A new effort combines computer vision and tactile sensing to permit gentle grasping of arbitrary objects in unstructured environments.
References:
Students:
Support:
Tactile feedback systems for teleoperation. In teleoperation, a human
operator remotely controls a robot in a hazardous or inaccessible environment.
These systems have been used for years in the nuclear industry and for undersea
exploration, but present systems are slow and clumsy, primarily due to the
lack of appropriate sensory feedback to the operator. Our research aims to
ameliorate this sensory deficit by creating new systems that allow the operator
to literally feel the objects that the remote robot is handling. One system
reproduces the small-scale shape of grasped objects on the operator's finger
tip. Another conveys vibrations from the remote robot's end effector. Our
studies have demonstrated that tactile feedback permits remote execution
of tasks that are otherwise impossible with present technology. Beyond the
system development aspects, the work is directed at understanding how tactile
information relates to task properties.
Our research is addressing several other aspects of haptic (force and tactile)
feedback. We have demonstrated a method for conveying slip information from
the remote robot to the operator's fingers that triggers a physiological
slip reflex, providing automatic and unconscious regulation of grasp force.
We have also made the first measurements of force reflection bandwidth
requirements for telemanipulation tasks with a multifingered
hand. Another effort is aimed at formulating the first algorithms for
synthetic tactile feedback
in virtual environments for training and simulation. A new project is
developing a system for
automatically identifying the properties of objects in remote environments.
We are working with two companies to commercialize the results of this research.
Schilling Robotic Systems, Inc. (Davis, CA) is the leading producer of
high-performance teleoperated robots for the offshore oil, marine salvage,
and nuclear power industries. Immersion Corp. (San Jose, CA) is one of the
first companies to produce haptic interfaces for virtual environments.
Applications of these projects include surgical training and critical-procedure
preparation for astronauts.
References:
Students:
Support: Office of Naval Research
Smooth bipedal walking. This project is directed at determining the
sensing and control strategies that will permit a legged robot to carry a
payload over rough ground as smoothly as a wheeled vehicle rolling over a
flat road. Kinematic analysis has provided a criteria for smooth transfer
of support at footfalls. Other results include algorithms for controlling
foot placement and forward velocity while maintaining smoothness. We are
now experimentally testing these algorithms on the planar biped robot in
our laboratory.
References
Student: Eric Dunn
Support: National Science Foundation (Graduate Fellowship)
Mechanical impedance of the human hand. This project will determine
how humans modulate the mechanical impedance of their hands in response to
the task requirements. The results help explain sensing and motor control
strategies in dextrous manipulation. Our approach involves experimental
measurement of force-motion relationships of the hand and fingers during
task execution. These studies have measured the impedance of the index finger
in extension and abduction, and the impedance of the precision pinch grasp
during lifting.
These measurements form the basis for the biomechanical analysis of drumming.
This task is of particular interest because skilled drummers can play drum
rolls at frequencies well in excess of the usual human motor control bandwidth.
They do this by allowing the drumstick to bounce passively against the drum
head at least twice for each hand stroke. We have obtained experimental evidence
that drummers control the bounce frequency by modulating grasp force, which
in turn controls the effective stiffness of the drumstick in its interaction
with the drum head. The results demonstrate that modulation of passive impedance
can permit a low bandwidth manipulator, human or robot, to execute certain
types of fast manipulation tasks.
References:
Students:
Support:
Finger pad soft tissue mechanics. We have characterized the mechanics
of the soft pad at the tip of the human finger through both measurements
and modeling. The results show that the force-displacement-velocity relationship
for indentation can be described by a nonlinear viscoelastic model. The pressure
profile across the finger pad is surprisingly well-modeled by the Hertzian
distribution derived for elastic spherical contact. These results are useful
for explaining the role of finger pad mechanics in neural response (see following
project), as well as dynamic tasks such as typing.
Reference:
Student: Dianne Pawluk
Support: Office of Naval Research
Mechanoreceptor neural modeling. We are developing the first holistic
models of the tactile sensing process in the skin of the human hand. These
models incorporate all of the essential functional components: the mechanics
of the skin, the mechanics of the mechanoreceptive nerve end organ, the creation
of the generator potential, and the initiation of the action potential
transmitted to the central nervous system. The initial models are simple
linear, lumped-parameter mechanical components coupled with the Hodgkin-Huxley
equations. We have shown that these models capture the essential properties
of the nonlinear frequency response of mechanoreceptor units to sinusoidal
displacement inputs applied to the skin surface. Based on these results,
we are developing more complex models that facilitate further examination
of mechanoreception, including models of branching afferent fibers and of
population responses. An ongoing collaboration with neurophysiologists
provides experimental data for testing these models.
Reference:
Student: Dianne Pawluk
Surgical instrumentation. Minimally invasive surgical techniques require
surgeons to work with long tools through small incisions, using only video
images for guidance. This deprives surgeons of one of their most valuable
assets: the sense of touch. The goal of this project is the development of
remote palpation instruments to convey tactile information from inaccessible
locations within the patient's body to the surgeon's finger tips. A first
target application is localization of pulmonary tumors, which are readily
detected as hard lumps in soft lung tissue. To optimize the sensing process,
we are modeling the interactions between the instrument and tissue using
finite element analysis (FEA) techniques. We are also characterizing the
forces and motions surgeons use in palpation. Our prototype instruments,
incorporating both tactile sensors and tactile display devices, are
now undergoing testing, and we are discussing commercialization of this
research with surgical instrument manufacturers. Remote palpation technology
will increase safety and reliability in present minimally invasive procedures,
and bring the advantages of minimally invasive techniques to more complex
procedures that are not feasible today.
On a fundamental level, benefits of this project include improved understanding
of soft tissue mechanics and the links between tactile sensing and motor
control. For clinical practice, benefits include standardized palpation
procedures for reliable detection of diseases such as breast and prostate
cancer, noninvasive diagnostic tactile probes, and improved surgical techniques.
Surgical robotics. Current medical robots are designed for simple
tasks like holding endoscopes and machining bone. The next challenge is the
automation of procedures that involve grasping and moving tissue. We are
contributing to the development of these systems through modeling the deformation
of soft tissues, based on our previous research in the mechanics of palpation
and the finger pad. We are also applying the tactile sensing and
signal processing
technology we developed for conventional robotic applications. In consultation
with surgical robotics companies and researchers in this field, we are
addressing a number of new applications.
Medical training systems. Virtual environment training systems allow
the user to interact with a computer model of a task through both visual
and haptic feedback. These systems could allow medical personnel to learn
new skills in a low-cost, risk-free setting. Current implementations, however,
lack the realism required for effective training. We are attacking two
of the major barriers to realism in these systems. First, we are characterizing
the mechanical interactions in surgical procedures, and determining
the pertinent
(large-deformation) properties of soft tissues. Second, based on our research
on tactile sensing and display, we are developing devices that recreate the
appropriate tactile stimulus for the trainee. For this application we
are targeting trauma procedure training for military and paramedical personnel.
Another project involves development of trainers for
needle insertion procedures such as biopsies and epidural anesthesia. These
surgical trainers promise to cut costs, decrease risks, and increase
effectiveness for a variety of medical training needs.
References:
Students:
Support: Whitaker Foundation
Return to Rob Howe's home page
Biomechanics Projects
Dianne T.V. Pawluk and Robert D. Howe.
Dynamic
Contact Mechanics of the Human Fingerpad, Part II: Distributed Response.
Harvard Robotics Lab Technical Report 96-004, December 1996.
Biomedical Projects