Carnegie Mellon ECE Capstone, Spring 2025: MM Demangone, Taylor Koda, Aneesha Bhattacharjee
This week, I worked on testing the smallest pill compartment to reach the 95% accuracy threshold. Currently, it’s at 83%, so I’m still working on that. I also did additional hardware testing and readjusted our design further to accommodate accurate pill dispensing by adding funnels to the top and covering the discs on the front to prevent the pills from flying out. Same as was in the presentation, conducted 59 trials on the smaller compartment.
Mainly worked on the final presentation this week and alsso did some more debug testing on the web app and looked into 3D printing some mroe insets.
The most significant risk that could jeopardize the project is the SDcard becoming corrupted right before the final demo. We think that we should boot up a spare, just in case it does become corrupted. No further changes were made to our design that we haven’t already mentioned. Everything is working now, except for the load cell amplifier, which we are continuing to work on.
We performed the following unit tests:
We performed the overall systems test:
Findings and design changes from tests:
This week I worked on the web app as well as the load cell amplifier. I modified the displays of some notifications on the web app, and added the email feature to be sent every time a user adds a new medication. I also flushed out all of the bugs for the automated email feature. For the load cell amplifier, I had soldered it and tested it on both the RPi5 and the arduino. Additionally, I worked on the poster, finishing the product pitch, system architecture, system design, and system evaluation. We still are having trouble with the load cell amplifier, which puts us a bit behind, but we have already talked to Professor Qin about it, and he said it was alright if it needs to be taken out. Next week, I plan to work on the load cell amplifier, finalize the presentation and paper, and prepare for the final demo.
This week our team worked on the physical housing. MM bought supplies, Taylor installed pipes, and Aneesha tested the new design. Aneesha did testing on the dispensing and MM did testing on the software application. Taylor did testing on both the hardware and software systems and spent time designing the new dispense system. Aneesha and MM worked together on testing the load cell on an arduino. The whole team worked on the presentation. The biggest challenge we faced was the load cell amplifier still not working. MM ordered another load cell amplifier, and is still waiting for it to arrive. However, at this point, it may need to be removed from the design entirely. Additionally, pills kept getting stuck in our original funnel design, so we decided to switch to a pipe-shape instead. This fixed our issue, and we have adapted our structure accordingly.
This week I had worked on the load cell amplifier issue by writing code for an arduino and testing the load cell with Aneesha. Unfortunately, this did not work, and we will need to change our design. Additionally, I figured out how to make our web application accessible on Google Chrome by creating a certificate and using wss. I fixed some coding bugs, such as a user’s personal alerts being the only alerts to pop up, sizing across different size screens (phone), and general bugs as they came up. I also went shopping to buy pipes for our project, as we had to do a change in design. Next week, I plan to solve some additional software bugs (overpopulation in google calendar when editing a form, not automatically saving the disposal time, being able to delete a pill). If the load cell amplifier arrives, I can work on that, and hopefully this issue no longer persists. Besides the load cell amplifier dilemma, everything is on track.
I have learned a lot of new tools and skills in this project. I gained familiarity with the RPi, as I have not worked extensively with it. Additionally, it was my first time working with the load cell, load cell amplifier, and speaker. I did learn those skills as well. There were a few software packages and deployment add-ons that I have never used before: celery, celery beat, installing certificates. I learned about all of these tools through the internet, primarily through google search or AI engines. I have links commented throughout the github of the resources that I referred to.
This week I continued debugging some of the software features like the healthcare provider view. I helped with the hardware components as well. Since the pill dispensing wasn’t the best for pills of the different sizes, I resized the slots in the discs so the right dose would fall out each time and added tubes to help with the jamming. I also made a CAD for 2 fillers for the discs. I am also preparing for the final presentation.
This week I have focused on doing a lot of integrated testing. MM and I were working on the load cell amplifier, which is not working as intended so i shifted my focus on running tests for the pill dispensing functionality. Initially the design was not working for all the pill sizes, so Taylor helped me resize and readjust some of the components so they work for the different pill. I 3D printed a small CAD model that Taylor made to fill two of the smaller discs. After that I have conducted 59 trials for each compartment with the different pill types, which we got from the equation in our design to ensure accurate pill dispensing of >95% . In doing so, I adjusted and stabilized the device as I went. I also added a slide-like structure so the pills fall into a central location. I also did slides for the final presentation and over the next week will continue testing and getting the hardware components ready for presentation and try to get the load cell working.
This week I helped gather some suplies by making some runs to tech spark. Then I also worked on adding like a healthcare provider view to the web app. Tmr I will finish the healthcare provider view on the web app and also add a walkthrough. So get excited!!! 😊 💊
The most significant risk that could jeopardize the project is the load cell amplifier. Since there is so much noise, we are currently unable to get any good reading (as shown in the picture below). From the research we’ve done, its likely due to the fact that HX711 load cell amplifiers work best for load cells that measure higher weights and that soldering can add to noise since we need a specific type of wire. However, we had to solder the wires because the load cells wires were so fragile, they kept on breaking. Aneesha and MM are looking into other solutions to explore such as force sensors. We decided to replace the battery pack with another buck booster converter to achieve better accuracy on the voltage that we need for a good reading. Additionally, we ordered another pack of SD cards, since we do not have any spares left. Both of these cost less than $40, keeping us under budget. The team will work on testing (accuracy of pills being dispensed, website re-sizing to devices, mobility needed to operate device, etc) as listed in the design report section 7. More specifically, we will pull up the website on a variety of screen sizes and go through the motions of the website to make sure everything is rendering appropriately. We will conduct about 60 dispense trials across a variety of pill sizes and record the accuracy of the dispensing.
