Intelligent Robotics Laboratory Vanderbilt School of Engineering Artificial Muscle based on Flexinol motor wire Scott Renkes Advisor: David Noelle

Intelligent Robotics Laboratory Vanderbilt School of Engineering Purpose  Design a new actuator  Robotics  Humanoid Robots  Human interaction  Prosthetics  Realistic limbs  Better for user  More controllable  Teleoperation  Decrease stress on user  Increase accuracy

Intelligent Robotics Laboratory Vanderbilt School of Engineering Goals  Design actuator similar to muscle  Control it with biologically inspired controller  Allow for easy interfacing  Make system portable and user-friendly

Intelligent Robotics Laboratory Vanderbilt School of Engineering Current Technology ActuatorProsCons Electric Motors Light weight Low power consumption Limited torque Stress on axel Hydraulics High force Heavy Pump required Pneumatics Medium force Elastic Non linear Pump required

Intelligent Robotics Laboratory Vanderbilt School of Engineering Why a new actuator?  Replicate human movement  Refined force and velocity control of device  Human like movement allows for better man machine integration  Humanoid robots can more easily mimic humans

Intelligent Robotics Laboratory Vanderbilt School of Engineering Fibrous Bundled Structure  Flexinol motor wire to replace muscle fibers  Package wires similar to muscle  Flexinol/muscle fiber proportional elasticity  Similar force/length curves

Intelligent Robotics Laboratory Vanderbilt School of Engineering Recruitment  One bundle, one neuron  Weak fast, slow strong  Properties of motor wire allow for variety of activation  Neural Network Controller  Force feedback training  Inverse Dynamics

Intelligent Robotics Laboratory Vanderbilt School of Engineering Feedback  Force feedback  Golgi tendon organ  Strain gauge  Length feedback  Muscle spindle  Approximation of Flexinol properties nawrot.psych.ndsu.nodak.edu/.../Movement/Reflex.html

Intelligent Robotics Laboratory Vanderbilt School of Engineering Flexinol Properties Diameter(mm) Linear Resistance  /m) Typical current (mA) Deformation Weight (g) Recovery Weight (g) Typical Rate (cyc/min)

Intelligent Robotics Laboratory Vanderbilt School of Engineering Flexinol vs Muscle Density(g/cc) Energy conversion Efficiency (%) Max Deformation Ratio (%) Breaking Strength (MPa) Muscle1.04??? Flexinol ,000

Intelligent Robotics Laboratory Vanderbilt School of Engineering Force Length Comparison  Muscle Force Length  See figure  Flexinol Force Length  Steeper slopes  Nature of Force Length Relationship  Tension  stretch

Intelligent Robotics Laboratory Vanderbilt School of Engineering Neural Net Recruiting  Neural Net Basics  Learning  Memory  Back Propagation  Neural Net Implementation  Computer  Microcontroller  Amplifier  Force Controlled  Force feedback  Desired Force

Intelligent Robotics Laboratory Vanderbilt School of Engineering EMG Controller  Muscle Voltage vs Muscle Force  EMG signals represent muscle force  Neural Network for EMG pattern recognition  Scaled EMG of biscep as input  Desired force as ouput  Relative to min and max The green line is bicep voltage The blue line is tricep voltage 3 and 4 are unused channels

Intelligent Robotics Laboratory Vanderbilt School of Engineering Sneak Preview Mk1

Intelligent Robotics Laboratory Vanderbilt School of Engineering Sneak Preview Mk2

Intelligent Robotics Laboratory Vanderbilt School of Engineering Cost  1 meter each  Flexinol HT 375  Flexinol HT 200  Flexinol HT 100  Flexinol HT 037  Microprocessor X 2  Multiplexer  Voltage Amplifier  Wiring  Epoxy  Cable  Total  $22.95  $19.95  $17.95  $16.95  $50.00  $8.00  $10.00  $8.00  $171.80

Intelligent Robotics Laboratory Vanderbilt School of Engineering Project Status  Completed work  Study properties of motor wire  Calculate efficiency of passive cooling  Develop Structure and Control for the device  Design test bed  Examine neural network solution  Current Work  Finish building actuator

Intelligent Robotics Laboratory Vanderbilt School of Engineering Future Work  Code user friendly interface for training and controlling neural network  Implement portable microcontroller neural networks  Design portable batter pack to drive system

Intelligent Robotics Laboratory Vanderbilt School of Engineering Questions?