Our project is being used to help the Marcus Center for Therapeutic Cell Characterization and Manufacturing (MC3M). Their work includes research on therapeutic cell treatments and analyzing cells related to epigenetics. Their mission is as follows: To bring together clinicians, industry researchers, and project developers to tackle the issue of high quality cell manufacturing.

The process of pipetting, which involves transferring small quantities of liquids using a pipette, can be quite time-consuming and is highly susceptible to potential human errors, which can introduce inaccuracies and compromise the overall reliability of the results obtained.

One potential solution to address the time-consuming nature and human error associated with pipetting is the implementation of robotics to automate the process. By incorporating robotic systems into laboratory workflows, the pipetting process can be made autonomous, resulting in increased efficiency, accuracy, and overall reliability of experimental procedures. In our project, we used the UR3 robot arm, an electric pipette, and a grip (made in CAD to attach the arm to the pipette) to move and pipette across a well plate autonomously.

Although idea 4 did not meet all of the requirements, we decided to sacrifice ease of assembly to create a gripper that would prioritize compliance, stability, and a secure mount.

We started by writing a script that would find the average distance between each well and will move the robot that average distance. The problem with this is that it would only work if every well location was already known beforehand which makes it very hard for the code to work across multiple well plates. Later we changed our plan to use computer vision to find the location of each well using a Intel Real-Sense D415 camera that way the code can be applied to every well plate regardless of size. The downside of this is that the pipetting is less accurate, but the upside is that the the UR3 can pipette a well plate of any size

After several phases of prototyping, evaluating, and modifying, the model successfully meets all of the requirements. It is easy to mount and assemble, and holds the pipette well enough to not need the top piece for extra security.

In conclusion, our approach on combining computer vision and a corrective algorithm has proven to be a successful solution for pipetting a 96-well plate. Through multiple phases of prototyping, evaluation, and modification, we have developed a model that will hold the pipette so it can perform serial dilutions. The integration of computer vision technology has enabled us to accurately identify the target wells and precisely guide the pipette, ensuring efficient and reliable pipetting operations. Our success in developing this pipetting solution opens up exciting possibilities for automation in the laboratory setting. With the potential to adapt to various well plate sizes, our model paves the way for automating pipetting tasks across different experimental setups, thereby saving time, improving reproducibility, and increasing overall productivity.