OpenCV

Arduino

Raspberry Pi

.NET

ROS

ROS2

Predictive and Generative Modeling of Mixed Dimensional Aerogels with Programmable Properties

A UR5e robotic arm loads aerogel samples into an Instron Series 6000 Universal Testing System to perform compression testing.

In this work I expanded and improved the autonomous platform for unsupervised testing of aerogel samples. The main goal was to close the loop. By this I mean, while using the stage, positions had to be pre-programmed, and thus the number of samples that could be run was limited. This platform allowed for open-loop operation, where samples can be continuously loaded in batches, and tested by the robot.

There were two major changes that made this work possible. First, we constructed an auto-feeder, which can store up to 12 samples at once, and continuously feeds samples down to the loaded position, from which the robotic arm picks them up. Our samples are held on custom holders, which includes a label and bar code so that it is both computer and human readable. When it is time to load a new sample, the label of the loaded sample is read, and then tested. Second was closing the loop with our Instron testing machine. We purchased a digitsl IO board so that we could send triggers signling test start and test end. This enabled a closed loop operation of our autonomous platform.

A control schema had to be designed and implemented to ensure the consistent and safe operation of this system. Our root node was a computer, which handled the synchronization of action by each system. When the Instron system was ready for a new sample to be tested, an image was taken of the loaded sample position to check if a new sample was prepared. The label on the sample also contained embedded data about the sample, including its unique sample ID for storing test data. If a sample was ready to go, a signal was sent to a microcontroller which signalled the robotic arm to move the sample into the testing position. A microcontroller was required to amplify the 5V signal to 24V to be read by the robotic arm's IO box. After the compression cycles were complete, and test data stored, a signal was sent from the Instron to the microcontroller for the arm to remove the sample. This loop continues until there are no remaining samples, at which time the program shuts down.

Depending on the number of cycles run, and the percent compression required by the user, compression testing can take anywhere from 5-15 minutes to complete per sample. This redundant manual task of loading samples and beginning the test was a prime task to implement an autonomous solution. Saving the human-labor of researchers, allowing them to perform much more critical and conceptually rigorous tasks.

Demo video coming soon...

Machine intelligence accelerated design of conductive MXene aerogels with programmable properties.

This goal of this project was to interface our UR5e robotic arm with a Series 6000 Instron Universal testing system to perform unsupervised compression testing of aerogel samples. A challenge, was finding a means of communicating between the Instron and UR5 robotic arm without having a digital IO board for the Instron system. We instead utilized an audio output as an analog signal to communicate the end of testing, and for the robotic arm to load a new sample. We used a stage with predetermined positions for a discrete number of samples to be tested. Testing was performed in batches.

This platform enabled the unsupervised testing of samples in batches. Saving manual human-labor, and increasing the experimental throughput of our lab.