Technical Paper - EDGE - Rochester Institute of Technology

advertisement
Multidisciplinary Senior Design Conference
Kate Gleason College of Engineering
Rochester Institute of Technology
Rochester, New York 14623
Project Number: P14253
UNDERWATER MCKIBBEN MUSCLE MANIPULATOR
Will Fickenscher
Mechanical Engineering
Joseph Taddeo
Mechanical Engineering
Jared Warren
Mechanical Engineering
Erika Mason
Electrical Engineering
Chris Jasinski
Mechanical Engineering
ABSTRACT
This project is part of a partnership between the Rochester Institute of Technology (RIT) and Boeing to research
and develop underwater technologies. This project's primary focus is on the development of a robotic arm, or
manipulator that can offer more dexterity and environmental safety than current underwater manipulators. The arm
is actuated by pumping water into McKibben Muscles, which act similarly to human muscles. The use of water to
actuate the muscles not only is safe for the environment, but also ensures that the arm will be able to function at
great depths in the sea.
BACKGROUND
Boeing, most commonly known for their industry leading aircraft, has a separate division specializing in subsea
exploration. It currently employs an unmanned underwater vehicle known as the Echo Ranger. The Echo Ranger
was designed mainly for scientific use and data gathering ranging from acoustic mapping to water sampling. The
Echo Ranger competes with a wide array of other versatile unmanned underwater vehicles. Boeing has teamed up
with the Rochester Institute of Technology to research and develop new technologies that could be outfitted to their
subsea vessel to improve it’s functionality and increase Boeing’s foothold in the Unmanned Underwater Vehicle
(UUV) market. Beginning in the 2013-2014 academic
year, Boeing has sponsored three Multidisciplinary
Senior Design projects. Among these was a project to
develop a robotic manipulator that could be used on an
underwater vehicle.
Current robotic manipulators feature a pincher-style
gripper powered by hydraulics, electronic servos, or a
combination of both (Fig. 1). These are important for
moving objects on the ocean floor, but they suffer from
some drawbacks. Mainly, the pincher-style grippers can
be very difficult to control and grab objects. The
hydraulic arms also pose an environmental risk with the
Figure 1: Hydraulic Cylinder Underwater Manipulator [1]
possibility of the hydraulic fluid leaking out into the
ocean. An alternative to hydraulic power for actuation of
the manipulator is McKibben Muscles. McKibben muscles create a tensile force, similar to human muscles. They
consist of a latex tube and an outer mesh. When the latex bladder is pressurized, it inflates and pushes the mesh out
radially causing the muscle to contract (Fig. 2). McKibben muscles have been around since the 1950’s and one of
Copyright © 2014 Rochester Institute of Technology
the most successful applications of this technology has been in the Shadow Dexterous Hand [2], which allows for an
unprecedented amount of control and dexterity in a robotic manipulator. Previous RIT MSD projects have also used
McKibben muscles to demonstrate the feasibility and advantages that this technology offers.
To develop a McKibben
muscle actuated arm for use on a
UUV, improvements need to be
made to the current design and use
of this technology. First, air would
not be a suitable medium for
actuating the muscles. Since this
arm will be used at great depth and
Figure 2: McKibben Muscles
pressure, air would compress under
these circumstances and would not
properly inflate the muscles. As a
result this project would need to incorporate water as the actuating fluid instead of the air. Previous versions of
RIT’s air muscle hands offered a limited number of muscles in the hand resulting in a low amount of dexterity. To
improve current pincher-style grippers, more muscles would need to be added in order to allow for the level of
control that would be needed to successfully use this style of gripper. Additionally, the hand alone offers little
functionality without an arm to position the gripper into its proper location so a functional arm would need to be part
of the system. Finally, a more intuitive user interface would need to be created to allow for easy use and control of
the entire system of the more dexterous hand.
Taking into account all the needed improvements to the current McKibben muscle hand designs, the goal of this
MSD project was to create an articulating arm and hand that could ultimately pick up a golf ball, a dive stick, and a
small brick. The design would also need to take into account that it would be working in a fully submerged
environment so special attention needs to be given to selecting materials that have anti-corrosive properties.
MUSCLE ACTUATION
The design of the Underwater McKibben Muscle Manipulator was influenced by the Festo Humanoid, the
Shadow Dexterous Hand, and the Beijing Institute of Technology Dexterous Hand; all of which focused on
developing a functional hand through the use of McKibben Muscles [2-4]. These successful designs however do not
have to operate at the extreme pressures that an underwater manipulator faces while operating at great depths in the
sea. This allowed Festo, Shadow, and the Beijing Institute of Technology to use compressed air to actuate the
McKibben muscles, whereas this would be difficult to utilize on a UUV because the air would have to be stored
somewhere on the vessel and compressed additionally as the vehicle descended. As a result, the Underwater
McKibben Muscle Manipulator utilizes water to actuate the McKibben Muscles.
Water is the ideal medium to actuate the muscles for multiple reasons. First, water is incompressible, so there
would be no need to compensate for compression effects. Instead, water can be pumped from the surrounding water
or a tank in the vessel to the muscles. Water pumps are popular and create the pressure differential needed to expand
the McKibben muscles, thus making them act similarly to their air-actuated counterparts. Using water is also
beneficial because it reduces the footprint of the system inside the vessel. Instead of a tank of air and high-pressure
compressor, a water system only requires a water pump and valves for directing flow.
The primary drawback of using water instead of air is that it doesn’t flow as quickly. As a result, muscle and in
turn, manipulator motion will be slower. Depending on the task, this could either be beneficial or detrimental to the
objective at hand. However, using higher flow pumps and proportional control valves would reduce this effect.
FINGER DEVELOPMENT
In order to turn the axial force created by McKibben muscles into rotary force needed to rotate finger
appendages, fishing line was used similarly to how tendons are utilized by the human body. As a result, the finger
design was driven primarily on the need to allow ‘tendons’, or the fishing braid to travel along the finger, while also
being able to create a torque around the joints. The fishing braid passes through the center of each finger appendage
in order to keep the fishing braid safe from abrasion. Through concept selection, spring pins were chosen to be both
an anchoring point for the fishing braid and the rotating link between the separate finger appendages. The spring
pins are cost effective and easy to repair and replace.
Keeping the fishing braid and spring pin constraints in mind, the fingers were developed to resemble that of a
human hand as closely as possible. The fingers are comprised of 4 appendages: the distal, middle, proximal, and
knuckle (Fig. 3). Each appendage was designed with dimensions of the human hand in mind. The profile of the
finger was constrained by the water-jet manufacturing process, which was available and cost effective resulting in a
sharp cornered representation of the human finger. The appendage profile was designed to allow a 90° bend at each
Proceedings of the Multi-Disciplinary Senior Design Conference
Page 3
joint, which is roughly what a human finger is capable of at full constriction.
Each finger has 4 degrees of freedom, one at each joint with another at the
knuckle to allow side-to-side motion of the fingers (spreading fingers wide).
The index and pinky fingers each are allowed to spread 15° from pure vertical,
while the middle and ring fingers are given 7.5° to spread.
In order to rotate a finger appendage, the fishing braid must be able to
create a torque around the spring pin joints. To do this, the braid is knotted to
a separate spring pin placed above the joint and then traverses around the joint
and into the next appendage and continues through the finger and palm until
Figure 3: Finger Design
being connected to a McKibben Muscle. In order to create the most torque
possible around the spring pin, a Teflon bushing is placed on the spring pin
with a max diameter to allow the fishing braid to pass over it without causing abrasion against the appendages.
To reduce the cost of expensive valves and the complexity of the control circuitry and algorithms, the fingers
were designed to utilize McKibben Muscles only for contraction. Retraction of the fingers is done using springs
instead of muscles. As a result, plates are screwed onto the backs of the fingers that serve as mechanical stops to
prevent the finger from retracting past straight or
‘hyperextending’. An adverse effect of using springs
for retraction is a reduction of net force created by the
muscles since they have to counteract the linearly
increasing force created by the springs. Upon testing on
a plastic prototype finger, this effect was realized
quickly as the muscle struggled to overcome the
spring’s force while contracting the finger. To reduce
this effect, the contraction tendon is anchored farther
Figure 4: Finger Cross-Section and Fishing Line Routing
past the retraction tendon in order to create a greater
lever arm for torque transmission.
THUMB DEVELOPMENT
Similarly, the thumb was created using a distal, middle, proximal, and knuckle
section. The distal, middle, and proximal sections are identical to that of the finger with
the exception of length. The lengths of the thumb sections were altered to better
represent that of an actual human thumb. The knuckle joint on the thumb was a
complicated feature to design. A human thumb consists of a ball joint, which is difficult
to control using McKibben muscles. As a result, the side-to-side and open-close actions
were separated and designed using individual components. The first component, which
mounts to the palm of the hand, allows the entire thumb assembly to move from side-toside. This piece is connected to the second component of the thumb knuckle, which
covers the open-close motion. This is then in turned connected to the remaining
proximal, middle, and distal components of the thumb. The connections between all
components in the thumb are the same as in the fingers using spring pins and Teflon Figure 5: Thumb Knuckle
bushings.
MUSCLE REQUIREMENTS
In order to actuate each finger, two muscle-spring pairs are needed. Contracting the muscle closes the finger,
while the return action of the spring provides the force needed to open the finger. Similarly, the contraction of the
second muscle pair spreads the finger while the spring returns the finger to the straight and upright position. This
equates to 8 muscles and 8 springs for the four fingers. The thumb uses one muscle-spring pair to close the distal
and middle sections of the thumb, and two muscle-muscle pairs to control the opening and closing of the proximal
section as well as the side-to-side rotation of the thumb knuckle. This equates to an additional 5 muscles and 1
spring in the thumb for a grand total of 13 muscles and 9 springs in the hand.
Copyright © 2014 Rochester Institute of Technology
Each hand muscle was designed using 1/2” latex tubing and 1/2" mesh in order to provide the needed force to
actuate the hand components. Under contraction, each muscle expands to roughly 5/8” which needs to be accounted
for when packaging the muscles side-by-side. Each muscle also needs to be roughly 10 inches long in order to
provide the needed length of contraction. Previous RIT MSD projects that
used McKibben muscles had two common problems when wiring the muscles
to each component. First, the cabling that they used tended to stretch and two,
it was very difficult to wire the muscles up while keeping the line tight. To
combat these problems. Dyneema braided fishing line was used to limit the
amount of stretch in the line since it is the most sensitive fishing line on the
market, and a cable tensioner was designed and incorporated into each muscle
and spring assembly. This tensioner consists of three components. The first is
the tensioner housing. This component is installed into the end of the muscle
and has a hole for a hex bolt to pass through. There is a milled slot so when
the hex bolt is tightened down, it cannot rotate. Additionally, there is another
slot milled through the center of the
tensioner body to allow for the center of
the bolt to be accessed. The second
component in the tensioner assembly is
a hex bolt. This is a standard hex bolt
with the addition of a small hole drilled
Figure 6: Teflon & Line Routing
in the center. Threading the fishing
braid through this hole and then tying a
knot creates an anchor point for the line and creates the connection between
the muscle and appendage that it’s intended for. The last component is a nut to
fasten down the bolt and keep it in place. When the tension in the cable needs Figure 7: Tensioner Exploded View
to be adjusted, the nut is loosened and the bolt is retracted enough so the hex
cap is free to rotate. The bolt is then rotated until the proper tension is
achieved. The bolt is then reinserted completely into the slot so the bolt is no longer free to rotate and finally the nut
is tightened down to keep the bolt from slipping out. Teflon tubing was also used as an outer covering for the fishing
line to help guide and route the cabling through the palm and fingers. This Teflon tubing also had the added benefits
of acting as an outer housing that reduces friction and helps protect the line from wear.
FOREARM DEVELOPMENT
The forearm was designed to provide the proper spacing between the hand and elbow as well as provide a
location to mount the 13 muscles and 9 springs needed to actuate the hand. Dimensions for the forearm were
originally planned to replicate that of a human, but both the length and the diameter of the forearm needed to be
increased to properly package all of the muscles. The forearm uses a center shaft that runs the entire length of the
forearm. Three plates are then mounted onto the shaft to provide the means to mount each of the springs and
muscles. The first plate is mounted at the end of the forearm and
provides a means for the muscles and springs to be anchored to. The
next plate is located on the hand end and provides a support for the
muscles to contract and relax in. This ensures that the muscles
remain in line with their anchor points. The third plate is mounted
slightly further down the shaft closer to the hand. This plate allows
the braided line to remain in line with the muscles when it is routed
into the Teflon tubing. Each of the three forearm plates uses the same
hole pattern with an inner circle and an outer circle. The inner circle
allows for 8 mounting positions and the outer circle allows for 16
mounting positions. Together, they provide enough mounting
locations for the 13 muscles and 9 springs with 2 empty spots that
can be utilized for more functionality in the future.
Much like a human, the forearm is connected to the hand with a
wrist. The wrist uses a double horseshoe design. The first shoe allows
for the hand to bend up and down. The second show allows for the
hand to rotate side-to-side. Originally, the wrist was intended to be
controlled with muscles, but in order to reduce overall complexity,
Figure 8: Forearm Assembly
this function was dropped and Nord-Lock washers were added at
each axis of rotation to allow the user to tighten down the bolts to lock the wrist into the desired position.
Proceedings of the Multi-Disciplinary Senior Design Conference
Page 5
ARM AND SHOULDER DEVELOPMENT
The arm and shoulder were originally designed with five degrees of freedom. These degrees of freedom would
include horizontal rotation at the shoulder, vertical rotation forwards and backwards at the shoulder, vertical rotation
to the side at the shoulder, vertical rotation at the elbow, and forearm roll at the elbow. The vertical rotation to the
side at the shoulder was eliminated in order to simplify the design while still allowing the design to meet customer
and engineering requirements.
A water pump rated at 100 psi was selected to power the muscles. The pump is capable of moving four liters of
water per minute. The flow rate of the pump was not important since there was not a project requirement that the
arm had to move at a certain rate.
The muscles are an assembly of two nylon end caps, two hose clamps, an eye screw, a push-to-connect fitting,
silicon tubing, and mesh sleeving. The nylon end caps contain a barb that holds the tubing on and allows the hose
clamp to secure the tubing and mesh sleeving. On one end cap the eye screw is connected so the muscle can easily
be connected to transfer force. On the opposite end of the muscle the end cap connects to the push-to-connect fitting
allowing it to connect to the pump.
The joints consist of an aluminum pulley, an aluminum rod, and two Teflon sleeve bearings the sit in structural
aluminum. The aluminum rod sits in the sleeve bearings allowing it rotate freely. A pulley is mounted to the
aluminum rod. From here, the pulley connected to the necessary muscles by braided cable.
The shoulder had to allow for both horizontal rotation and vertical rotation forwards and backwards. To
accomplish the horizontal rotation the sleeve bearings sit in two horizontal aluminum plates. The aluminum rod that
goes through the lower plate has a pulley mounted to it. Opposing muscles mounted to the base power this joint. In
order to achieve the vertical rotation at the shoulder a pulley is mounted to an aluminum rod that connects to the
upper arm. Opposing muscles mounted to the body rod control this joint.
At the elbow both vertical rotation and forearm roll are required. The vertical rotation is done with opposing
muscles mounted to the upper arm. The forearm roll is also accomplished with opposing muscles mounted to the
upper arm. These muscles must transfer their force through bowden cables, however in order to power the pulley
connected to the forearm rod.
CONTROL DEVELOPMENT
Through the process of developing the mechanical structure of the Underwater McKibben Muscle manipulator,
the processes that would be needed to control the entire system began to separate into two subsections, the hand
control system and the arm control system. Deciding to build a hand
that is semi-dexterous added complexities to the system since the
human hand is capable of movements that cannot be replicated
without several muscles contributing to something that a tendon or
small muscle would do. The second part of this is attempting to have
as much control over each movement as possible so that the
manipulator can perform tasks as a human hand would. Because
there are several layers of depth in what the analysis of this amount
of control would incur, finding an alternative was imperative,
especially for a nine month project. By using flexible resistors/
sensors and analog current switches, the goal of moving the fingers
was realized.
The second control system that was to be applied to the
mechanical system was for the arm. This was a simpler design since
the valves that were used to actuate the muscles were either on or off
and could be turned on and off with a simple switch. However, to
keep in mind what the purpose of the project was, it was decided to
control this part of the system with a joystick. Given that this
Figure 9: Hand Control Analog Circuit
manipulator would be controlled remotely, having a joystick that
could move each part of the arm was an attractive alternative to fulfill
the customer requirements.
As aforementioned, the control system for the hand portion of the project consisted of a flex sensor and an
analog current switch. The flex sensor is a resistor that when flexed, has an increased resistance value. This
difference in resistance is what was used to get the valve to turn on and off. The sensor was then grounded in a
voltage divider and fed into an op amp which was then fed into a power transistor and that is what drove the valve.
The schematic for this is shown in Figure 9. This was necessary to have because the difference in the resistance that
Copyright © 2014 Rochester Institute of Technology
was varying had to have a large enough for the valve to recognize. The
configuration allows the valve to turn completely off when the sensor is not
flexed and turn on when the sensor is flexed.
There are three different arrangements for this circuit to control the
different parts of the hand. Having to control the open and closing of the
finger, the spreading of the fingers and the many movements of the thumb.
PUMP AND VALVE DEVELOPMENT
The hand and arm design utilizes two different types of muscle
combinations. The first is a muscle-spring combination, and the other is a
muscle-muscle combination. The muscle-spring combination is used to either
contract the muscle fully, or to relax the muscle completely. This requires one
3-way, 2-position valve for each muscle-spring combination. The 3-ways Figure 10: Flex Sensor Glove Control
refers the 3 ports. One port is for the pressure source, one is for the exhaust
port, and one is for the outlet (muscle). The 2-positions refers to the 2 states
of the valve. The valve is either at rest, which allows the muscle to exhaust, or the valve is actuated, which
pressurizes and contracts the muscle. When the muscle is deflated, the spring creates the return force for the
opposing motion. When a muscle-muscle combination is used, a single valve is needed for each of the two muscle
pairs. The valve used in this situation is a 5-way, 3-position valve. The 5-ways refers to the 5 ports. The first port is
for the pressure source, the next two ports are the outlet ports for each muscle, and the final two ports are the
exhaust ports for each muscle. The 3-position refers to the 3 states of the valve. The first position is at rest, which
each muscle is neither exhausting nor pressurizing. In this state the muscles are holding their pressure. The second
state allows the first muscle to pressurize while the second muscle exhausts, and the third state allows for the second
muscle to pressurize while the first muscle exhausts. This single valve allows for both muscles to work in tandem to
provide the opposing motion. The 3-way, 2-position valves are either fully pressurized, or fully exhausted, while the
5-way, 3-position valves allow for position control. The hand utilizes 9 muscle-spring pairs to control the closing
and spreading of each finger and the tip of the thumb, and 2 muscle-muscle pairs to control the rotation and closing
of the thumb. The arm uses 4 muscle-muscle pairs to control the rotation of the shoulder, upper arm, lower arm, and
elbow.
The muscles in the arm require more pressure than the muscles for the arm since they are required to create
much higher forces. As a result, the hand and arm each use a separate pump to provide the needed pressure and flow
rate for each sub-system. The hand uses a 60 psi output pump, and the arm uses 100 psi output pump.
SYSTEM RESULTS
The resulting system assembly is shown in Figure 11. The arm has two degrees of freedom, having the ability to
rotate around the base 70° and up and down at the shoulder 120°. The elbow also has two degrees of freedom and
can lift and roll the forearm 90° relative to the upper arm. All four fingers close into a fist and spread wide at the
knuckle. Cable tensioning issues late in the project resulted in removing a degree of freedom from the thumb, which
left it with the ability to open and close at the proximal and distal
appendages.
The full manipulator assembly was tested above and
underwater. In both scenarios, it was able to grasp and lift a onepound weight, diving stick, golf ball, baseball, tennis ball, and brick,
which were objectives outlined at the beginning of the project.
The maximum lifting capacity for the system was never
measured since project display and
planning did not leave time for
testing. It’s assumed that hand
position,
the
cable-to-pulley
connection, and bicep muscle size are
the determining factors for this
specification since the large muscles
used to rotate the shoulder up and
down are capable of lifting hundreds
of pounds, which was validated
Figure 11: System Assembly Rendering
through testing.
Figure 12: Hand Rendering
Page 7
Proceedings of the Multi-Disciplinary Senior Design Conference
Figure 13: Grasping a Brick
Figure 14: Grasping a Baseball
CONCLUSIONS AND RECOMMENDATIONS
The Underwater Mckibben Muscle Manipulator assembly ran without failure at an all day campus festival for
creativity and drew enough attention to win the Xerox Corporation Sponsor Award for Innovation and Creativity.
This final test proved that the overall design of the system is done with high quality. Throughout the development
process however, initial ideas, goals, and designs were changed in order to keep the project on schedule.
The controls subsystem was one such area where the initial ideas were simplified. Future work should be
pursued to enhance the controls of this manipulator. Ideas to do this include force and position feedback,
proportional valves, an exosleeve to have the manipulator also mimic arm movements, and integrating a
microcontroller.
Additionally, the hand could be further improved by allowing wrist movement, since the wrist is currently
locked into place. Also, consideration should be given to redesigning the thumb. The thumb is a very intricate part
of the human anatomy and not enough time was spent during the development process on achieving a fully
functioning thumb. Furthermore, finger actuation could be improved by creating individual control of the proximal
and middle/distal appendages and replacing the retraction springs for muscles.
As mentioned before, the fishing line had the tendency to snap and therefore, could be upgraded to provide
more durability.
It’s also recommended to create more forearm roll. Having 180° of rotation would add a great deal of
functionality for lifting objects from the floor. In addition, another degree of freedom could be added to the arm
allowing it to lift to the side.
REFERENCES
[1] "Compare SpecsManipulators." Kraft Telerobotics. N.p., n.d. Web. 12 May 2014.
[2] "Dexterous Hand." Shadow Robot Company. N.p., n.d. Web. 12 May 2014.
[3] "Humanoid." Festo Pneumatic & Electric Automation Worldwide. N.p., n.d. Web. 12 May 2014.
[4] Liu, Hao, Wei Fan, Lin Yu, Guang-Zheng Peng, and Tao Wang. "Study On A New Dexterous Hand Actuated
By Pneumatic Muscle Actuators." Proceedings of the JFPS International Symposium on Fluid Power 2008.7-2
(2008): 521-26. Print.
ACKNOWLEDGMENTS
A very special thanks to Boeing, Kathleen Lamkin-Kennard, Rick Lux, Rob Kraynik, Jan Maneti, and the RIT
ME machine shop staff for without their generous support this project would not have been possible.
Copyright © 2014 Rochester Institute of Technology
Download