|
|
|
|
|
VRROOM (Virtual Reality
Robotics and Optical Operations Machine)
James Patton, Robert Kenyon, Sandro
Mussa-Ivaldi, Alon Fishbach, Felix Huang, Assaf Dvorkin, Mark Kovic, Sarah Housman, Ross Bogey
A key
development for the lab is a new system that uses augmented reality, motion
analysis, and robotics technology.
|
|
|
Reorganizing Muscle Synergies
Devjani Saha
Objective: This study examines whether
muscle activation patterns associated with an isometric force contraction task
can be changed through training in able-bodied individuals.
Background: For most isometric force
contractions several muscle activation patterns can be used to perform the
task. This redundancy results when the
number of muscles acting across a joint exceeds the degrees of freedom
associated with the joint. The large
number of muscles allow for an infinite number of possible muscle activations
to accomplish the same joint torque.
Yet, studies have shown that subjects tend to exhibit the same activation
pattern for a given task.
Clinical
Relevance:
Studying the plasticity of muscle activation patterns in healthy individuals
may help us design rehabilitation protocols aimed at overcoming abnormal
muscle synergies exhibited by stroke patients.
Current
Work: The current work involves developing a
learning paradigm to determine whether subjects can be trained to exert a
specific endpoint force using a muscle activation pattern that is different
from the one they typically employ.
|
|
|
SERKA (Series Elastic Remote Knee
Actuator)
James Sulzer
A Robotic
Orthosis for Gait in Stroke
It is
commonly held that the cause of stiff-knee gait in stroke is due to
inappropriately timed or graded muscle activity in the knee extensors and
flexors. The object of this study is
to develop a lightweight orthosis that can assist
the knee at key points in the gait cycle, but feel transparent when
necessary. This orthosis,
called SERKA (Series Elastic Remote Knee Actuator) uses remote actuation and
series compliance to create a safe, lightweight and transparent knee
actuator. The results of this work
will lead to a better understanding of the mechanisms of stiff-knee gait, and
perhaps result in the design of a portable gait orthosis.
|
|
|
Error Augmentation
James Patton, Robert Kenyon, Sandro
Mussa-Ivaldi, Mark Kovic,
Sarah Houseman, Ross Bogey
Recent developments in human-robot interactions (haptics) that have revealed prospects in the areas of
motor teaching and rehabilitation that can speed up, enhance, or trigger the
motor relearning process. A promising alternative to simply having a robot
guide movements is to make movements more difficult by deflecting them from
the desired path. People develop, through practice, the ability to counteract
forces and visual distortions. If these distortions are properly designed and
applied, a desired movement pattern occurs when the distortions are
eventually switched off. The subject sees something unexpected that is
perceived as an error.
Our results point to a single unifying theory:
Errors induce learning, and judicious error augmentation (through forces or
visual distortions) can lead to lasting desired changes. Interestingly, this
process appears to bypass conventional learning mechanisms that require
intense concentration -- results are the same if the subjects have a
conversation or listen to music. They often consider it a game.
Until now very little of this research has been
functionally relevant because the devices’ ranges of motion were small, were
two dimensional, and were lacking an appropriate visual interface. Our goals
are to build on our promising body of
evidence and expand our error augmentation training work to a large workspace
in three dimensions using VRROOM, and
move towards clinical application by testing our approaches on stroke
survivors.
|
|
|
Force Control
Vikram S. Chib, Ph.D.
Objective: To understand how the brain learns to
coordinate a variety of simple modules of control to generate complex
behaviors. Specifically, how does the nervous system simultaneously control
hand movements and interaction forces during active haptic
discrimination?
Background: When manipulating objects, we must
control our hand motion as well as the interaction forces that arise from
contact with the environment. At the
level of musculoskeletal biomechanics, motions and forces are coupled by
intrinsic limb impedance. However, it has yet to be established whether at
the neural level the control of motion and force are coupled or independent.
Current Work: We have found results suggesting the
existence of independent neural controllers for arm motion and interaction
forces. This evidence is offered by transcranial magnetic stimulation (TMS) of posterior
parietal cortex (PPC) resulting in the differential disruption of the control
of motion but not of force.
|
|
|
Delays and Time
Assaf
Pressman
Objective: Explore the effect of sensory and
environmental delays on perception of surfaces stiffness
Background: Advanced technology has recently
provided truly immersive virtual environments with teleoperated
robotic devices. In order to control movements from a distance, the human sensorimotor system has to overcome the effects of delay.
Currently, little is known about the mechanisms that underlie haptic estimation in delayed environments
Current Work: The results of the experiment
indicate a systematic dependence of the estimated stiffness upon the delay
between position and force
|
|
|
Controlling powered
wheelchairs using motion tracking technologies
Alon Fishbach
Objective:
We investigate and develop new methods for controlling assistive devices
(e.g. powered wheelchairs) using motion tracking technologies.
Background: Many
patients, especially tetraplegic, have difficulties
in maneuvering and steering their wheelchairs. The combination of a patient’s
limited mobility and an interface that requires a precise and
inflexible
manipulation is a source of problems for these patients. This highlights the
need for the development of control interfaces that are tailored to the
specific residual motor skills of the patient. Motion tracking technology
offers a convenient and flexible way of capturing the motions of a patient
with limited mobility, as sensors can be placed virtually anywhere on the
body and measure very small motions.
Current
Work: Patients and healthy controls
practice the control of a wheelchair in a virtual reality environment. Apart
from capturing much of the motion repertoire a patient can comfortably
produce, our interface allows for flexible and adjustable mapping between the
patient motions and the wheelchair controls. At the end of this process, we
plan to apply the evolved interface to the actual device (wheelchair or
robot) and test the efficacy of learning in real-life context.
|
|
|
Machine Learning
Danziger, Z.
Objective: The goal of this study is to create and
examine a machine learning algorithm that adapts in
a
controlled and cadenced way to foster a harmonious learning environment
between the user and the machine.
Background: Human-machine interfaces must
reconcile two concurrent learners in a high dimensional signal space, the
person learning to use the interface, and the machine learning algorithm.
Achieving a maximally beneficial dynamic between the two is essential to this
technology's viability. By using previously made errors, the new algorithm
may guess how subjects will err in the future, and provide corrections before
the mistakes are even made.
Clinical Relevance: Patients who no longer have
the use of their limbs now have access to various technologies which can harness
the volitional signals that their bodies can still produce, such as EEG
electrode hats, eye movement recognition, or position sensors worn over a
garment. Each one of these technologies are cumbersome to learn and awkward
to control, making everyday locomotion extremely difficult. With the aid of
an effective computer learning algorithm learning these control paradigms
will become easy, and control can become graceful, greatly improving quality
of life.
Current Work: Healthy subjects wear a sensor
glove which maps high dimensional hand configurations into a cursor on a
screen in a very non-intuitive way. They must then demonstrate mastery of the
transformation by placing the cursor in targets. The learning algorithm
assists subjects by updating the mapping from hand to cursor based on their
own errors, which speeds up their learning rates.
|
|
|
Training with Enhanced Interactive
Priming
FC Huang, FA Mussa-Ivaldi,
JL Patton
We are interested in investigating how motor learning
is influenced by robot-applied loads designed to exaggerate the dynamics of
the arm and or wielded object. In addition, we would like do determine how
such loading, coupled with free manual exploration, affects the ability to
generalize learned strategies to functional performance. We hypothesize that
heightened sensory-motor experiences that reflect the existing mechanical
behavior of the arm can accelerate the learning of manual skills.
We investigated how motor planning in an object
manipulation task is influenced by free manual interaction augmented with a
simple external load: negative damping. Subjects controlled a force-feedback
planar manipulandum that presented a primary
loading condition of simulated anisotropic inertial forces in five different
orientations. The motor task consisted of two phases: (1) free
"interactive priming" where no movement is prescribed, followed by
(2) a performance evaluation consisting of three circular movements about a
fixed track.
For the test subject group enhanced priming,
we introduced negative damping during the interactive priming phase prior to
task performance. As a control, we presented a second subject group normal
priming with interactive priming with only the primary loading. Our
results showed significantly greater reduction in maximum curvature error for
the subject group that received priming augmented with negative damping
compared to the control group.
|
|
|
T-WREX
Sarah J. Housman, Vu Le
and David J. Reinkensmeyer,
The goal of this project is to test a device
which allows stroke survivors to practice arm movement therapy with
indirect therapist supervision. The Therapy-Wilmington Robotic
Exoskeleton (T-WREX) was developed at the University of California-Irvine.
This device was designed for adults with significant arm weakness resulting
from stroke, and provides intense movement training without continuous
supervision from a therapist.
T-WREX is a five degree-of-freedom passive
antigravity orthosis and computer workstation.
The orthosis relieves the weight of the arm using
elastic bands attached around its frame. It is instrumented and
contains mechanical joints which correspond to joints of the human arm,
allowing naturalistic arm movements. Stroke survivors are able to
practice repetitive arm movements in the T-WREX by playing functionally
oriented computer games such as reaching for objects on a shelf, eating, and
cooking. This study will assist researchers in determining whether
these types of exercises improve arm movement without direct supervision from
a therapist
|
|
|
Enhancing Learning in Laparoscopy
FC Huang, FA Mussa-Ivaldi,
JL Patton
We are interested in investigating how individuals
learn how to perform skillful manipulation of a laparoscopic tool. Medical students report particular
difficulty in learning complex maneuvers in laparoscopic surgery, especially
those involved in intra-corporeal knot-tying. We speculate that the
difficulty involved can actually be divided in several component skills,
including learning the kinematic transformation due
to tool pivoting, interpreting visual cues in the context of a spatial
mapping, and movement planning that integrates both tool tip position and
hand posture. Furthermore, we hypothesize that training can be accelerated by
presenting feedback conditions that enhance the learner’s perception of the
appropriate sensory-motor transformations.
In a human subject study involving medical
students, we intend to investigate learning and generalization of manipulation
skills in a laparoscopic tool simulation that present visual feedback while
tracking hand position and posture.
Subjects will be asked to control a virtual laparoscopic tool with
their hand in order to perform targeted reaching movements with the tool tip
along with a specified final hand posture. We will present experiment
sessions that each includes different levels of scaling that alters the
demands of hand versus tool-tip range of motion, in terms of both position
and orientation. We will analyze hand
and tool movement trajectories and characterize learning in terms of
systematic distortion to path, final targeting error, variability.
|
|
|
Brain
Machine Interface
Alessandro Vato, PhD,
Marianna Semprini
A typical BMI system is a device that a) records
neural activity from the brain, b) infers the intended action and, c)
depending on this, controls the movement of an external device (e.g. limb
prostheses, a robotic arm or the position of a cursor displayed on the
screen). Feedback (i.e. somatosensory information)
can come from intact senses such as vision and hearing or can be generated
artificially by translating information collected by the external device into
sensory stimulations.
Somatosensory real-time
feedback is fundamental for motor planning and for executing “on-line” errors
correction during the movement. In people with sensory motor disabilities,
the sensory information that cannot reach the brain can be “substituted”
through an intact sensory channel (i.e. eye or ear) different from the
damaged one. Alternatively, the damaged sensory pathway can be “replaced” trying to achieve
the same sensation in an artificial way.
The main goal of this project is to produce
artificial proprioception by stimulating directly
the somatosensory cortex of rats while performing
complex motor tasks. We are developing
a new experimental paradigm to train rats in motor tasks with visual feedback
information coming from an animated scene.
The following step will be to introduce external visual disturbances
and to train rats to compensate them; subsequently the disturbances will be
associated with patterns of cortical microstimulation
while reducing progressively the visual information. Our goal is to simulate
a situation in which the state of the controlled external device (i.e. the
“state” of the cursor) is encoded via cortical microstimulation
that is the type of information (e.g. the “state” of the limb) coming from
afferent neurons.
|
|
|
Proprioceptive Feedback
Mark Shapiro, Cynthia Poon,
Fabian David, Minos Niu,
Daniel Corcos
Objective: To
investigate the descending control of proprioceptive
feedback during movement.
Background: Proprioceptive input from muscle spindles can be used by the
CNS for feedback control during movement. The gains of the spinal and
supra-spinal feedback loops are centrally modulated depending on the desired
movement speed, expected load, and other parameters of the task.
Clinical Relevance:
Abnormalities in the descending control of the proprioceptive
feedback pathways are believed to underlie motor impairments in spasticity, Parkinsonian
rigidity.
Current work:
Subject makes a series of arm movements to a target. The robot unexpectedly
changes the movement trajectory, and the responses in muscle electromyogram (EMG) are analyzed.
|
|
|
Patient Mediated Rehabilitation
through Telerobotics
Kari Danek
We are investigating a new approach to robotic
therapies wherein the patient would be given manual
control over
the exercise movements of the therapy.
Through the use of a teleoperated lectromechanical robot, the patient could manipulate and
assist their impaired limbs with their unaffected limbs. This project is specifically aimed at lower
limb rehabilitation where the impaired leg or legs can be assisted by the
unimpaired arm or arms. As a first
step, we have chosen to examine one degree of freedom manipulation of virtual
environments through the primary ankle joint assisted by wrist actions.
We have found that that subjects develop
anticipatory adjustments in the lower limbs based on upper limb
dynamics. Given that lower limb
neuromuscular commands incorporate predictions of interaction forces from the
upper limbs, we are building support for upper limb guidance of lower limb
motion as a productive means of neurological rehabilitation. In contrast, it
appears that interaction forces from external agents (i.e. robots or
therapists) can not be anticipated or incorporated into lower limb
neuromuscular control during therapy in the same manner.
Current Work: The current study in development
addresses the impact of upper limb contributions on a lower limb dynamic
task. Specifically, we are exploring the
ability of subjects to maintain excitations in a resonant system using ankle
and/or wrist movements. This task
requires both temporal and spatial precision of ankle movements, as does locomotion.
Possible mechanisms that could enable improved performance in the
cases where both upper and lower limbs are involved include additional
afferent information, increased dexterity, and the potential for
specialization in task demands.
|
|
|
MARIONET
James Sulzer
Chronic stroke survivors lack sufficient outpatient
therapy, despite indications that more therapy at the chronic stage can
restore some function. Both insurance and physical constraints on therapists
prevent training in the home, most likely where this activity would take
place. Nevertheless, this gap reveals
a promising application for robots, low-cost home care.
A robot designed for home use needs to be
inexpensive, portable and safe. Earlier, we have explored a type of compliant
variable transmission known as the MARIONET (Moment arm Adjustment for Remote
Induction Of Net Effective Torque). The proof-of-concept, behaving similar to
a rotary Series Elastic Actuator, has been found suitable for low-cost, light
weight applications. This paper discusses further analysis of the singlejoint MARIONET and proposes the design for the new
planar, upper extremity two-joint manipulandum for
clinical and home use.
|
|
|
VR-based assessment tool for spatial neglect
Assaf Dvorkin, Jim Patton
The neglect syndrome, which is characterized by a failure
to respond to stimuli that appear on the side
of space
opposite the lesion, is a complex disorder of spatial representation and
attention. The current
methods of
assessing neglect are poor and insensitive. Most of them take place on 2
dimensional surfaces which do not reflect the reality of a 3 dimensional
world. We are exploring novel possibilities of robotics and virtual reality
technology for assessment of neglect in multiple spatial dimensions. Our
initial results support the hypothesis that neglect patients exhibit spatial
bias in more than one spatial dimension simultaneously.
|
|
|
|
