Description
I initially worked on this in secret for maybe a month or so, and once I felt like I had a good base to present, I shared it at an intern report meeting and everyone really enjoyed the idea. I also purchased many of the objects used in the demo prior to announcing the idea to my company. This was because I did not want my idea to get caught up in waiting for the items needed to be approved and ordered. I wanted to make progress fast and I needed many components to get accurate dimensions for modeling. My overall strategy was first strategy (CADing, ROS2, sketching) then it was testing/RD&T (iterating, testing, updating and learning), and then polishing and programming last (preparing for demo, durability testing, consistency testing, printing backups, etc.)
Having previous barista experience working summers at a cafe in Miami during high school, I knew how picky some people can be about their coffee. Knowing this, I really prioritized making a cup of coffee that was actually good, and not just some gimmick. One thing I really wanted was to have the filter be cleaned first before making the coffee. This is a textbook rule in making drip coffee and almost any recipe you find will include this as a preliminary step. So I added a sequence to my program that first rinses the coffee filter with hot water, then removes the cup underneath before continuing to make coffee. It is at this point the user would add the coffee server, allowing the coffee to not have an odd filter/paper taste in their final product. In addition, I also added a bloom phase to the program. The bloom is simply wetting the coffee grounds to release trapped CO2 from the grounds, allowing for a bolder and more complex coffee. In addition, it has the effect of creating a foam when the first pour is done, which is eye catching.
The process that took the longest amount of time and iterating was the 90° gearbox. I think I made about 5 different designs until I was satisfied with the final product. Another challenge was finding a water dispenser that does not have a child safety lock. Since many hot water dispensers require two buttons to be pressed at once or other various ways to prevent a child from burning themselves, this posed a challenge to the single robot I had to use. Luckily I was able to find one on Amazon and the lever push style that was discovered was more effective and efficient than the typical button press versions often found in Japanese convenience stores.
Once I had modeled everything, I exported the models to meshes, and defined my URDF for the 4-axis robot, prismatic gripper, and gearbox movement. I used the simulated robot and demo environment to get an idea of where all the components should be placed relative to the accessible reach of the robot. I initially used simulation to make a plan in MoveIt and export that to the real robot, but I found it easier to use the real robot and use teaching to create the desired program.
I had initially started by brewing one cup of coffee at a time, but once I learned that the company wanted to present my demo at Robot Technology Japan 2026, I started thinking about more widespread distribution. Rather than creating one cup at a time and having people need to wait in line, I deduced it would be better to redesign the holder for the V60 to fit a coffee server underneath. This design update would allow me to serve up to 5 people at a time with one run of the program. I did some testing and found the desired ratio of rotation sequences and fill time to create a palatable 500ml pot of coffee.
I had the opportunity to present this demo alongside my automated pipette demo at Robot Technology Japan, with full ownership/authority for both. In total, I served over 100 cups of coffee, running the program 20 times, and used 1kg of coffee grounds. In addition to creating a program to actually brew the coffee, I also made a separate program that would just loop the movement of the cup pouring in a circle. This had the effect of allowing customers/clients to see how the coffee was made while drinking it if they were not there to actively see it brewed. I ran this loop program for a total of 18 hours without any problems or disruptions. It made me really happy to see people walk by and take videos of the project I made, or laugh and say how interesting of an idea it was. Also when people said the coffee it made was actually good, I got joy. There were a variety of people served, even a person who said they formally worked at a kissaten (Japanese coffee shop) approved of the pour and commended it.
What I initially intended to be just for my coworkers turned out to be shared and experienced by strangers. I was at first nervous and wary of the trust they put in an intern like me, but I stepped up to the plate and really went from zero to one all on my own accord.