Team EcoFlex -Design Plan

Initial Prototyping: 

We began with something we could create quickly and modify to attain a simple understanding of this soft material we plan to work with and how to best focus our efforts. We quickly learned at this point that using a 3d printed mold with EcoFlex brand silicon was a good place to start and from there we worked with the model to try to expand its functionality. Such expansions included a four-way soda bottle splitter made of 1/4″ tubing and glue and rigid valve connections in order to achieve isolation in each of the robot’s limbs. These parts, while functional, are obviously not as versatile as we’d like them to be and it is important to keep in mind that the 3d printed mold is subject to change in the future. 

Our first attempt at a 4-way splitter enclosure
Placing rigid valves inside the mold greatly helped us create a sealed air flow

Going Forward: Our plans for this project.

Stage 1Autonomous Air control: Simple control of individual valves and valve decompression. This is necessary to get right before we move any further with our project.

Inexpensive example which uses a diaphragm pump and solenoid valve. 

Stage 2 Movement based on manual control: We want this thing to ultimately move this robot with a sequence of inflations and any other movement necessary for this thing to crawl (that being a small rotation left or right in the limbs or something similar). It will be important that this thing is able to physically move with the controls we give it before we begin any sort of training.

Simple crawling soft robot

Stage 3 Intelligence/Movement Learning: It would be fantastic for us to apply a sort of learning method to this project. Our idea is to reward this robot whenever an inflation or rotation in its body results in movement. This way, in theory, we are creating an “unsupervised training model” in which it has a general goal of learning how to move. I have found examples in which an artificial neural network is used with an Arduino Uno microcontroller to create a “general purpose network for supervised or unsupervised training”. This of course is a pretty big task and the example I’ve found may be overkill for what we’re trying to accomplish. The point is that there are many ready made examples and resources for us to learn from and apply to our project.

To get an idea of how this might work in a virtual environment, check this out. Although this video demonstrates the use of a genetic algorithm (in this case it mimics biological evolution in the form of physical permutations), the same concept and iteration process can be applied to the motion of our robot.


Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Google+ photo

You are commenting using your Google+ account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )


Connecting to %s