A Robot to mimic horizontal Eye Movement - Biomedical ...

Dr. John D. Enderle
Professor of Biomedical
Engineering
University of Connecticut
Phone:(860) 486 5521
Email: jenderle@bme.uconn.edu
TEAM
No.9

Main Goals
• Replicate the Linearized Horizontal Fast Eye Movement System
• Quickly target a stimulus in a +/-20°range
• Camera inside the eye will capture an image
• Image will be processed and appropriate movement executed
• Movement should mimic that of the human Saccade
• Target Saccade profiles for Human Eye
• Speed
• Acceleration
• Viscoelasticity
• Precision

Anatomy
• Eye
• Optical Nerve
• Six Muscles
Types of Movement
• Smooth Pursuit
• Vestibular ocular
• Optokinetic
• Vergence
• Saccades

This linear muscle model exhibits accurate nonlinear force-velocity and length-tension
relationships. No other linear muscle model to date is able to accurately reproduce
these nonlinear relationships.
• Each branch models a specific muscle
• Medial Rectus muscle modeled by the
Total Agonist force.
• Lateral Rectus muscle modeled by the
Total Antagonist force.
Superior and Inferior Rectus’ muscles
modeled by a single spring and dashpot
in parallel.
F ag
and F ant
represent the agonist and
antagonist active-state tensions. These
are our controls over the system.

• F constants represent
forces based on the
ocularmotor elements.
Tau and cap T constants
intrinsic to humans and are
set values based on
experimental data.
• Fag and Fant functions will
output to the motors.

• Neural and Agonist/Antagonist
inputs shown to the right.
• These inputs produce a 15
degree simulated saccade with a
max velocity of 400 deg/s.
• The Agonist/Antagonist inputs
are relayed to the motors as
force/time inputs for movement.
• Simulink block diagram simulates the
function of the eye robot.
• Constants are defined in a MATLAB
workspace.
• Workspace is then transferred into
Simulink.

• 40 degree range ( +/- 20 )
• Peak Velocity of ~400 degrees per second
• Duration
• ~30 ms for < 5 degree saccades
• 50+ ms for > 10 degree saccades
• 1.6 Newton Force

MigeOne10 Shape Memory Alloy (SMA)
• Force of 8.82 N
• Time - 0.034 seconds (32 Volts)
• Speed 243.6 mm/s
• Stroke length 8.19 mm
• Forced Cooling Methods

Miga MOSFET Drivers
• External GATE (CNTL) Signals
• Range (5 V– 30V)
Alps Position Sensors
• 10k Linear Potentiometer
Arduino Uno
• Analog Output (0V – 5V)
• Receive Position Information
(Analog-to-Digital Converter)
• Generate Control Signals

It’s not enough to simply have a motor move the same length
as muscle contraction at the same speed as muscle contraction
• The Actuator motion doesn’t
account for the natural
elasticity or viscosity of the
muscle
• Each branch models a specific
muscle
• Antagonist/Agonist – Lateral
and Medial Rectus
• Superior/Inferior Rectus- Model
the top and bottom eye muscles,
which don’t have muscle input
horizontally, but do affect the
eye rotation

DAMPER AND SPRING SELECTION
Spring Selection
• Component only exerts force
relative to POSITION
• Select for resting length and K
Damper Selection
• Component only exerts resistive
force relative to VELOCITY
• Represents viscosity of muscle
• Controlling the flow rates of
air out of a pneumatic
cylinder, we can resist
velocity of stroke

MOUNTING VISCOELASTICS
Rail System
• A system of friction-reducing
linear bearings on a guided rail
will transfer energy step-by-step
from actuation to Saccade
• Viscoelastic components will be
attached to the bearing blocks
• Keeps the motion linear with no
torque

• The Human eye has a diameter of 24 mm
• Mass of roughly 7.5 grams
• Why is scale important here?
• Globes to be fabricated by QCI Engineering
• Designed to mimic physiological dimensions
• Needs to house Camera
• Needs to provide USB cable outlet.

• 5 MegaPixels
• USB Compatible
• Roughly 15mm in Diameter
• 320x240 at 30 Frames/s
• SnoopyPro to capture driver communication with
camera hardware (GNU General Public License)
• Flatten Camera components
• Redesign in Multisim
• Order PCB via Ultiboard

• Color detection via HSV Thresholding
• Easier to discern target objects
• Flattens an RGB image into 2 colors
• Advantages?
• Smaller file size for transfer
• Calculate position by finding
centroid of white pixel mass
• Disadvantages?
• Note top right image

Dr. John D. Enderle | Client ● Uconn BME Professor
Marek Wartenberg |BME senior design Advising
Alex Korentis | UConn PhD student ● QCI Engineering
Mark Gummin | Miga Motor Company