Unfortunately, I have not had enough time to complete my robot but here is what I made, what I learned, and where I want to go with it.
The original idea for my robot was inspired by a video that I saw on youtube while searching different types of soft robots. I came across a video of a robot that could successfully pick up ping-pong balls and perfectly shoot each ball into all the cups in a game of beer pong. I was amazed how the gripper didn’t even look like it was holding the ball but how it so successfully held onto it. I was also amazed at how far and how precise the balls could be launched with just air. So, I started researching the method in which they were doing this and discovered this particle jamming method was developed by Cornell University in 2010.
I started researching and thinking about how this could be done for the least amount of money because I enjoy the challenge of replicating complex functionality quickly and cheaply even if it may take a little more work. I stumbled across a tutorial for a universal gripper and started building.
Balloon – Used to hold the coffee grounds and inflate and deflate when suction and inflation happens. Using a balloon is important because the rubbery material enables a strong friction grip.
3/4″ PVC Pipe – Used for air flow from air pumps
Coffee Grounds – Used as the particles for the particle jamming. I got ground coffee with fairly large grounds so because having larger grounds allows for more space and more air between the particles. This in turn allows for particles to more easily be distributed when air is introduced and also allows for more air to be allowed to be sucked out of the particles resulting in more control with suction.
Plastic Funnel – The funnel focuses the direction of the balloon inflation to be downward rather than to the side. The funnel allows greater control of the direction of gripping.
Small piece of cloth – Used to keep the coffee grounds from coming out of the balloon when the pump is pulling air in.
Duct Tape – Used for attaching tubing and securing the balloon open.
Air Mattress Pump x2 – One pump is used for suction the other is used for inflation. I originally wanted to use 1 pump but was unable to think of a good system to use both with only 1 pump since the dc motors within the pumps were only unidirectional.
1ft Vacuum Tube x2 – Used so I could keep the pumps on the table while I positioned the gripper over the object. The tube just allowed for a greater range of movement while also restricting air direction.
PVC Ball Valves x2 – The ball valves were used to control the type of air flow to the balloon gripper. By using a valve for the suction flow and the inflation flow I could restrict flow from one direction and allow flow from another direction.
T-Connector – Directs the flow of the air into the gripper
- Fill Balloon with ~ 1/3 can of coffee grounds
- Trim down funnel so that it is just slightly smaller than the PVC so it can fit into the PVC tube.
- Pull end of balloon through funnel and over the small opening of the funnel.
- Place cloth over the opening of the balloon
- Duct tape the cloth around the outside of the funnel securing the cloth over the balloon and the balloon to the funnel
- Put the funnel into the PVC pipe so that the cloth and balloon are inside the pipe
- Duct tape the PVC to the the funnel
- Connect the PVC to the T-connector
- On either side of the T-connector attach a PVC ball valve
- For each T-Connector route the vacuum tubing to either the inflation or suction pump, 1 suction, 1 inflation.
Unfortunately, I could never actually get the robot working. I was able to get the balloon to inflate and deflate and as a result get the particle jamming effect but I was unable to pick anything up. I believe that this was because of the lack of suction from my air pump. It just couldn’t hold the object strongly enough to keep them in place when lifted. It is a great prototype and I learned a lot about air flow and material characteristics.
There were a couple challenges I encountered throughout the project. Firstly, I did not know of a good way to get anything into a balloon while simultaneously inflating it. This caused me a lot of frustration and resulted in a lot of spilled coffee throughout the process. Another challenge I ran into was how much coffee to add into the balloon to get the optimal amount of space, suction, and grip. I went through a couple iterations and learned that there was a max to how much should be put in the balloon. Below is a picture of one of the failed prototypes.
Another challenge I had was figuring out how to create a bi-directional system without two pumps. I originally thought that I would be able to use a single pump and single output for suction and inflation by just reversing the dc motor in the pump. I learned after purchasing the pump that the suction actually comes out of the side and the inflation comes out a different outlet. This led me to try a couple different methods of connecting both section to the same flow pipe but that didn’t end up working out. I also had challenges with trying to figure out a good valve mechanism to control air flow. While I decided to use two bi-directional ball valves I realize now that if I was to do it again, it would be better to use a three way ball joint valve. This would then only require rotating one valve to change the air flow.
This was an enjoyable project to build and I learned a lot through the build process but in the end I didn’t get it to do what I originally envisioned the robot to do. If I had more time to work on it, I would work on a better system for controlling air flow including an electronically controlled solenoid valve to rapidly cut or open air flow from a direction. I would also include small dc motor pumps instead of air mattress pumps because small dc motor air pumps are just as powerful and are easier to control using an arduino or other type of micro-controller. I would then be able to program the micro-controller to react to a variety of sensors and inputs to turn the system into a more autonomous robot. If I could get all the above implemented I would also want to work on building an arm to control the location of the gripper. The goal in the end would be to be able to control the gripper to do a variety of complex and important tasks like win at a game of beer pong by accurately shooting ping pong balls into plastic cups.
Thanks for reading