The basis for this project is the robotic hand article Andrew Terranova contributed to the Manual section of Popular Science magazine, based on the techniques outlined in Harvard's Soft Robotics Toolkit.
The goal of this robotic hand project is to make soft robotics accessible to the hobby roboticist. Some of the tools, materials and techniques included in the soft robotics toolkit may be somewhat challenging or expensive for a hobbyist. This project aims to use simpler techniques, common tools and less expensive material whenever possible.
The video below demonstrates the performance of the hand and the control board.
Andrew Terranova is a program manager, electrical engineer and an electronics and robotics hobbyist. He is an active member and supporter of the Let's Make Robots community.
Andrew is also a freelance writer. In addition to the Popular Science article for the robotic hand project, he has worked as a contributing editor for Make:, contributing several articles to the print edition of Make: Magazine and numerous posts and projects to their blog.
While discussing a topic for an article with Popular Science editors, the Harvard Soft Robotics Toolkit was mentioned. Andrew Terranova had seen the toolkit before, but had not considered making something with it. Andrew reviewed the information in the toolkit, and decided to adapt some of the concepts to make a robotic hand.
Some of the materials and fabrication techniques included in the toolkit seemed intimidating for a hobby roboticist. To make soft robotics more widely accessible, the goals of this project were:
The resulting design met these goals, and was published in the May issue of Popular Science.
Below is a video excerpt from Erin Kennedy's Robot Party Google Hangout from February 2015, where I introduce the project and explain some of the motivation behind it. The Robot Party is a community group of robot makers who meet online to demonstrate and discuss robotics projects.
Although there were no formal functional requirements, an important aspect of the project was to create an engaging exhibit to attract and interest people into soft robotics. On April 19th 2015, Andrew displayed the robotic hand at the Children's Museum of Somerset County NJ, where he volunteers, much to the delight of the children and not a few of their parents. It is hopeful that the project will continue to inspire curious people to try something in robotics in general, and soft robotics in particular.
The design for the actuators of the soft robotic hand is based on the concepts of fiber-reinforced actuators, but the materials and fabrication techniques are much simpler. Ribbed plastic hose, of the kind found on inexpensive foot pumps, is used as the reinforcing material.
A standard latex 350Q balloon, the kind used for making balloon animals and available at any party store, is used as the inner bladder.
There is a clear trade-off in reliability, as many of the fingers eventually failed... sometimes quite comically as the balloon expanded through the failed plastic of the constraining ribbed hose (see Actuator Testing and Performance section). Nevertheless the low cost and ease of fabrication make this a reasonable option where reliability is not a key factor. It is fairly easy to make a new finger to replace a broken one as needed.
Each actuator forms a finger, which was combined to make a completed hand by fastening to a toy armor gauntlet that happened to be available.
The subsections will outline the design of the individual actuators, the hand as a whole, and the control board.
Balloons of the type used for twisting animals and other fun things for kids seemed the clear choice of material for the inner bladder. The 350 type balloons (3 inches wide, 50 inches long) were chosen over the 260 type (2 inches wide, 60 inches long) most commonly used for twisting balloon shapes. The 350 balloons felt more durable than the 260s; the latex may have greater wall thickness. These balloons are easily available and suited the purpose well.
The selection of the reinforcing material, however, took considerable trial and error. PVC tubing of various sizes and types, and even latex reinforced work gloves were tried. The standard PVC tubing split easily, while the braid reinforced type was too hard to bend. The gloves did not provide enough structure. In the end, what worked best was a ribbed plastic hose that came from an inexpensive foot pump. This is the blue hose shown in the picture below.
The hose has a slight natural bend to it, similar to a human finger when held at rest. Stress relief holes are used on each side of the hose between each rib to reduce the chance of the hose splitting. Slices between each rib along the upper side of the curve, extending to the stress relief holes, allow the finger to bend when the balloon is inflated.
The tip of the finger is blocked so the balloon cannot expand in that direction. So when the balloon is inflated with pressurized air, the only way the balloon can expand is by bending the finger.
The design is a simple analog for a human hand. For simplicity, a plastic toy armor gauntlet was used as the platform for the five actuators. One happened to be available. A simple wooden board or other platform would have served, but wouldn't have looked as nice.
The thumb is opposed to the fingers as in a human hand. This allows for some gripping capability. However, because all the fingers are fixed with only the ability to flex as their movement, the effectiveness of the gripper is limited. In a human hand, there are many more degrees of freedom, of course.
The design of the control board was based on the fluidic control board in the soft robotics toolkit.
A few key differences to note are:
The following subsections provide step-by-step instructions for fabricating the individual actuators, the hand as a whole, and the control board.
Detailed steps to fabricate the actuators are below, but can also be found on the Popular Science website.
Detailed steps to fabricate the hand are below, but can also be found on the Popular Science website.
Detailed steps to fabricate the control board are below, but can also be found on the Popular Science website.
This is a complex project. Since your control board is a considerable investment in time and money, consider future expansion as you plan your build. For example, you may wish to buy an 8-port manifold and appropriate valves and fittings, so you can expand the board to control more actuators than then five you need for the robot hand project.
You can download the schematic below from the link at the bottom of the page. Clicking on the image below will also load the image in your browser, and you can save it locally from there.
The robotic hand was not designed to meet a specific set of requirements or capabilities. Rather, the idea was to create a soft robotics project that would be accessible to a wider audience. Performance of the soft robotic hand has been acceptable, based on the goals of the project.
The resulting design was published in the May issue of Popular Science.
The video below demonstrates the hand and control board operation.
The following subsections will discuss performance observations of the individual actuators, the hand as a whole, and the control board.
The video excerpt below is from Erin Kennedy's February 2015 Robot Party, where I discuss the reliability of the finger actuators. The Robot Party is an online community of robot makers who share their projects and discuss robotics ideas.
As mentioned in the design section, several different materials for the fingers were tried. The ribbed hose selected for the prototype worked the best. However, it is far from perfect. The fingers do fail after repeated use. Others are encouraged to investigate other materials and techniques.
Sometimes the tip of the finger blows out, as shown in the picture below, or it fails along the length. These failures are caused when the plastic fatigues and one of the cuts in the finger splits beyond the stress relief holes.
A 3D printed design to replace the ribbed hose is an interesting option. The growth of 3D printing has made a custom designed actuator a reasonable possibility for a hobbyist. The ability to manipulate design parameters and rapidly prototype for increased reliability is very attractive.
This project will be updated if a workable solution using 3D printing can be developed.
The video below is an excerpt from Erin Kennedy's February 2015 Robot Party. The Robot Party is an online community of robot makers who share their projects and discuss robotics ideas. In the video I demonstrate the manual and programmed modes of operation, and discuss the performance of the gripper.
As a practical gripper, the hand has very limited use. The fixed position of the fingers makes it much less flexible than a human hand, while the soft materials make it much less precise than a traditional robotic gripper. In effect... the worst of both worlds!
The hand can hold light objects like a ball or a light weight hand tool. It could be attached to a larger robot arm as an end effector, and used in situations where precision and grip strength are not as important as having a gentle, compliant grip.
Much of human communication is non-verbal. By using gestures that humans are familiar with, a robot can elicit an emotional response. For example, different responses from an observer are likely if the robot is gently waggling its fingers, taping impatiently, imitating a rude gesture, or ominously counting down.
In programmed mode, the several gestural positions and programmed motions can be amusing, and are instructive as an example of what soft robotics can do. Additional positions and motions can be programmed as well.
When operating in manual mode, the frequency of the Pulse Width Modulation (PWM) used to control the air valves can be changed via a potentiometer. The frequencies used are in the audio range and sound is emitted by the coils of the air valves during use. The frequency also visibly alters the movement of the fingers from a stuttering motion to a fairly smooth motion.
Within the limits of the freedom of motion, the hand can be made to "dance" and "sing" along to the control of an operator, which can be somewhat entertaining.
The video below is an excerpt from Erin Kennedy's February 2015 Robot Party. The Robot Party is an online community of robot makers who share their projects and discuss robotics ideas. In the video I demonstrate the control board, and also discuss how this project can be a jumping off point for people interested in using the soft robotics toolkit.
The control board performed very well. As mentioned in the design section, there are certain differences between the prototype and the fluidic control board design defined on the main Soft Robotics Toolkit.
The controls knobs and switches are all located on the outer edges of the board so they are easy to reach. As I built the prototype for myself, nothing is labeled. I did find that when I invited others to use it, the controls were fairly intuitive. Even children quickly learned how to operate the hand.
The use of an 8-port manifold allows for controlling more pneumatic actuators. Actuators could be added to move control a robotic arm or to add additional degrees of freedom to the existing hand.
Because the current control panel is fully populated, it may have to be replaced or modified if additional actuator control is desired.
The programmable nature of the Arduino Mega microcontroller allows for plenty of software expandability.