Hand Masters Reference Page
Presentation

Download a copy of the last presentation I gave.
Prototypes
Prototype 4 (03/22/05)

The 4th prototype will be similar to the 3rd, but will now incorporate actuation. It will use gears at each joint in order to allow for actuation. Also, ball bearings will replace the journal bearings currently being used. And small, gear-reduced motors will be placed directly above the finger segments. Additionally, this prototype will be made out of steel instead of aluminum. This should act as a useful testbed for determining how feasible gears are for transmitting forces.

Sensing and Actuation
At this time, there are no plans to include sensors in this prototype. Small motors with gear reducers will be used to actuate the joints. If the gear-reduction is too high, the device will not be backdriveable.


Prototype 3 (03/17/05)

The 3rd prototype is a testbed for a number of concepts, including how thick the device could be and how adjustable the device needs to be. The device also incorporates mechanical joint limits to prevent hyperextention of any of the joints, and uses metal shafts to apply forces to the phalanges (as opposed to straps or rings).

Download a video of the 3rd prototype.

Sensing and Actuation
No sensing or actuation was included in this prototype.


Prototype 2 - Block-and-Tackle with Bearings

The second prototype was designed specifically to test a block-and-tackle setup on a small scale. A block-and-tackle is a transmission that provides a mechanical advantage. A cable is wrapped around a pair of pulleys (rollers in our case). The more times it is wrapped around the pair, the greater the mechanical advantage. One end of the cable is fixed relative to one of the pulleys. The other end is free to move relative to the pulleys.

The first prototype included a low quality block-and-tackle mechanism that used small rollers in journal bearings. The result was that the rollers had lots of friction and as a result the block-and-tackle was "sticky". We decided that we should build a higher quality block-and-tackle mechanism to determine if such mechanisms were feasible at this scale. This prototype revealed a number of cable management issues, but also showed that a block-and-tackle mechanism is feasible at this scale. The use of ball bearings in this design (rather than the journal bearings used in the previous design) greatly improved the performance of the rollers. We will need to take more care managing the cables to prevent them from crossing over one another and from becoming slack.

The block-and-tackle prototype I built and the 3D model are illustrated at the right in both an open and closed configuration. If you look closely at the pictures you can see the miniature ball bearings. The bearings, which are the smallest ones I could find, have an OD of 0.125" and an ID of 0.040".

Lessons Learned
(1) Block-and-tackle mechanisms are feasible at small scales provided good bearings are used.
(2) Cable management is important.








Prototype 1

The first prototype (illustrated in the pictures to the right) was intended to test out some basic concepts for connecting two finger segments. Primarily, we wanted a design that would not apply forces through the finger joints (only torques around them). We also incorporated a rudimentary block-and-tackle mechanism in order to actuate the joint.

In order to avoid applying forces through a joint, I designed the adjacent segments of the prototype to rotate about a remote center. If the remote center is placed coincident to the center of rotation of the joint, then all forces on the device will be transformed into torques about the device's (and therefore the joint's) center of rotation. Thus, the joint between the segments will experience no loading. The actual device and a model of the device are shown in both a closed and open state. The device is stuck to my finger using double-sided tape.

This design sounds good in theory, but there are two practical problems with creating a remote center. The first is that it is difficult to accommodate all of the possible finger sizes out there...it would be incredibly difficult to develop a design with an adjustable center of rotation. The second is that a finger joint doesn't actually rotate about a fixed point relative to each segment. Instead, there is a rolling contact. This means that it is very difficult to keep the remote center of the device aligned to the finger joint's center of rotation. And if the two aren't exactly aligned, then the system is overconstrained (both the finger and the device attempt to dictate the point about which they are rotating).

We implemented the remote center using a slot with rollers in it. The slot in the distal segment of the device (the part closer to the fingertip) is constrained by two rollers in the proximal segment. Since there we are using journal bearing, the rotating pieces that allow motion between the slot and the rollers isn't very smooth. The same is true for the rollers in the block-and-tackle mechanism.

Lessons Learned
Based on the first prototype, we learned a number of lessons:
(1) Journal bearings are not good enough for a block-and-tackle setup nor for the rolling mechanism.
(2) The rolling mechanism probably isn't going to be smooth enough unless we can grind/coat the contact surfaces of it.
(3) It is very difficult to align the center of rotation of the device with the rotation center of the finger joint. If they aren't aligned, then the system is overconstrained and neither the finger segments nor the device can rotate.








Summary of Meeting on June 23, 2004
In today's meeting we established some of the basic criteria for the project.

Within one year, we would like to have a device demonstrating the basic concept of an exoskeleton for one finger. The finger exoskeleton should have the ability to apply and measure forces to the fingertip. This would be expanded later to the other fingers. Ideally, the final device would include 20+ degrees of bilateral (both flexing and grasping) actuation, but for now, we can focus on unilateral (just flexing) actuation of only one finger (3 or 4 degrees of freedom).

There are two areas of interest for this device: rehabilitation and basic science. Both areas of research will study "reach to grasp" movements of stroke patients. While there is some overlap between these two areas of research, there are some requirements for the device that differ between them. In the case of rehabilitation, the constraints can be mushy and only flexure motions require actuation. In the case of a basic science research, it is requisite that the device be able to perturb the movements of the patient and measure the force response of the finger to such perturbations. Bilatural constraints may be necessary.

Also, rotations about the wrist are important, and the device must allow for such motions to occur.

Another thing to consider in making this device is the time it will take to setup. If we make an exoskeleton device that comes over the back of the hand, it will probably require every segment of every finger to be strapped into the device, which could be very time consuming. As opposed to a device that you just place your hand on, or that only has 5 slots/holes; one for each fingertip.

Other Hand Masters

Below is an overview of some of the hand-masters that have been developed. Some of them are force reflective, and some are not. I think it's worth exploring the ones that aren't in order to get an idea of what information is being measured (e.g., force, position, velocity) and how these quantities are being measured.
Vanderbilt University - Exoskeleton for Astronauts

In 1997, a group at Vanderbilt (B.L. Shields, J.A. Main, S.W. Peterson, and A.M. Strauss) wrote a paper about using an anthropomorphic hand exoskeleton to prevent astronaut hand fatigue during extravehicular activities. They don't have a website, though, and I couldn't find any other information about the project other than what was written in the paper.

The exoskeleton was designed to assist motion about the PIP and MCP joints of the index finger, middle finger, and last two fingers combined. They use an innovative linkage to enforce a remote center of rotation about each joint. Each finger's motion is independent of the other (except for the pinky and ring fingers), but the motion of the PIP and MCP joints is coupled by the linkage.

Sensing
Strain gages are used to measure the forces applied by each fingertip. These appear to be the only sensors used on this device.

Actuation
One motor is used to drive each of the three finger linkages. Each motor is attached to a lead screw about which a wire is wrapped. This other end of the wire is attached to a cam on the linkage. As the cam moves, the linkage is "opened" and the fingers are bent.

 		Linkage Design


Linkage Schematic


Carnegie Mellon University - Robotics Institute Hand Exoskeleton Project

The group at CMU is working on an EMG Controlled Orthotic Exoskeleton for the Hand.
The goal of this research was to develop an exoskeleton that could aid people who were unable to pinch object between their index finger and thumb. The user would indicate their intent through activation of another set of muscles (e.g., their biceps) and the device would supply a grasping force to the two fingers.

Sensing
EMG signals from the biceps are used to obtain the intent of the user. These signals were used used in two different manners. In the first, they were used as binary signals. If the EMG signal was above a certain threshold then the grasping force would be applied. Otherwise, there was no grasping force. In the other case, they used a variable grasping force, where the force applied was proportional to the EMG signal.

Actuation
Two pneumatic pistons are used to control motion about the DIP, PIP, and MCP joints.


Jet Propulsion Laboratory

There are two websites related to this project, but neither one provides much information:
(1) This site has one useful picture.
(2) This site has some low-resolution movies.

In Man-Equivalent Telepresence Through Four Fingered Human-Like Hand System, Jau presents an overview of a four fingered robotic hand and the associated feedback glove. The pinky finger is not actuated in this system. The mechanism is sewn to the back of a glove in order to transmit forces to the user's fingers. While the mechanism can be adjusted to fit any hand size, in order to ensure proper force feedback, placement of the mechanism on the glove had to be accurately done. This means that a different glove must be made for every different hand size. There are 16 joints on the glove mechanism.

Sensing
The glove uses strain gages to force sense each of the glove joints.

Actuation
Flex cables transmit motion and power from "backdrive actuators" (not sure what these are - probably just dc motors) to each joint of the glove.




Laboratoire de Robotique de Paris (LRP) Dextrous Hand Master (DHM)

In Computing Optimal Forces for Generalised Kinesthetic Feedback on the Human Hand during Virtual Grasping and Manipulation, Tzafestas et al. present an introduction to the LRP Dextrous Hand Master (DHM). I'm trying to obtain the original paper.

Sensing
The LRP-DHM is capable of measuring 14 finger joint angles. They don't say how the joint angles are measured, though.

Actuation
The LRP-DHM uses tendons to apply forces on each phalanx of the hand. Each of the 14 joints "is actuated through a tendon-sheath transmission by DC, disk motors, placed remotely from the hand. Miniature force sensors are placed on each phalanx in order to measure cable strain and permit the implementation of force/impedence control techniques."


Immersion CyberGrasp

In Development and Testing of a Telemanipulation System with Arm and Hand Motion, Turner et al. present an overview of the CyberGrasp System. The CyberGrasp can apply independent forces to the 5 fingertips. It is capable of applying 12 N of force continuously to a fingertip.

Sensing
The CyberGrasp uses the CyberGlove to detect the joint angles in the fingers. At the moment, I do not know how the CyberGlove detects joint angles.

Actuation
According to Turner et al., the cyberglove uses "a set of [high quality DC] motors, worn in a backpack, to apply tension to cables in teflon sheaths, which in turn apply forces to each finger."

Turner et al., also mention that the "principal performance constraint is static friction between the tendon and the sheath." Something to keep in mind if we are considering the use of bowden cables.

Relevant Patents
Below are four patents that seem most relevant to understanding how the CyberGrasp works.








Utah/MIT Dextrous Hand Master

In Optimized Fingertip Mapping: A General Algorithm for Robotic Hand Teleoperation, Rohling, Hollerbach, and Jacobsen present an overview of the MIT/Utah Dextrous Hand Master (UDHM). The UDHM consists of a carbon-fiber exoskeleton attached to an elasticized glove. The exoskeleton extends around all fingers except for the pinky finger, and has a total of 16 DOFs.

Sensing
Joint angles of each finger are determined using sets of four-bar linkages between adjacent finger segments. The angles between these sets of linkages are determined using hall-effect sensors.

Actuation
The UDHM appears not to be actuated at all, and is therefore not capable of providing force feedback to a user.

The paper also notes that the EXOS Dextrous Hand Master was based on the first iteration of the UDHM.


Scuola Superiore S. Anna Haptic Interface for the Hand

In Mechanical Design of a Haptic Interface for the Hand, Frisoli et al. present an overview of a new haptic interface for the hand. The device is only capable of actuating the index finger and thumb, but demonstrates an alternative to the hand-mounted exoskeleton design seen in the CyberGrasp and the Utah/MIT Dextrous Hand Manipulator.

Sensing
Unknown at this time.

Actuation
Brushed DC motors are used to actuate a set of tendons.






Rutgers University

In The Rutgers Master II - New Design Force-Feedback Glove, Bouzit et al. present a haptic interface for interactions with virtual environments. The glove is able to provide force feedback of 16 N to each of the fingers (except the pinky, which is not used in this device). The glove limits the range of motion of the fingers because of the placement of the cylinders. However, because of its design, the glove weighs only 80 grams (2.8 oz), and is therefore much less tiring to use than other exoskeleton gloves might be.

Sensing
Hall-effect sensors are used to detect the angle of the pneumatic actuator relative to the palm, and an infrared sensor is used to detect the translation of the piston inside of the air cylinder. The position of the fingertip can be determined based on the angle and distance measured.

Actuation
Pneumatic actuators are used to apply forces to all of the fingertips except for the pinky finger.


Keio University's Passive Exoskeleton Haptic Device

In Multi-Fingered Exoskeleton Haptic Device using Passive Force Feedback for Dexterous Teleoperation, Koyama et al. present a passive exoskeleton that uses clutches to implement constraints. The pinky finger and ring finger are not restricted by the device.

Sensing
Potentiometers.

Actuation
Clutches in series with springs are used to prevent motion of the fingertips.


EXOS

It appears that EXOS worked on three projects related to hand masters:

(1) Dextrous Hand Master:
I believe is the same as the Utah/MIT Dextrous Hand Master I. In which case, it is unactuated and uses a less accurate sensing scheme (see Optimized Fingertip Mapping: A General Algorithm for Robotic Hand Teleoperation) than the Utah/MIT Dextrous Hand mentioned above.

(2) Sensing and Force-Reflection Exoskeleton (SAFiRE).
I haven't found any papers on this yet.

(3) Hand Exoskeleton Haptic Display (HEHD).
I haven't found any papers on this yet.

 		Exos Dextrous Hand Master


Sensing and Force-Reflection Exoskeleton (SAFiRE)


Hand Exoskeleton Haptic Display (HEHD)


University of Tokyo Sensor Gloves I and II

There's little to no information that I can find about this lab or these devices. Even the website where I got these pictures is not working right now. Based on all of the wires shown in the third picture, it seems likely that this was another actuated glove. 		Sensor Glove I


Sensor Glove II




 

Last updated by Tom Worsnopp on April 05, 2007.