- These past 2 weeks, I mainly focused on doing testing. This involved testing the integrity of the item stand, determine if our system met our timing requirements, and doing end-to-end testing with the entire system (including the item stand and facial recognition). For testing the integrity of the item stand, I wanted to ensure that the stand could withstand a large amount of weight placed on it. Each hook should be able to hold up to 20 pounds and the Nema 34 motor should be able to rotate the maximum weight of 120 pounds that we expect. In addition to this, I tested if our system could quickly recognize users, and rotate to a specific user’s position within 5 second. I worked with my team on fixing the motor speed and acceleration as well as looking more into the blocking behavior of our wireless transmission. We ensured that the speed at which the motor rotated would be safe for users and we ensured that the entire check-in and check-out process occurred relatively quickly. I also helped with testing the facial recognition with my team members, helping to develop new approaches and implementations to improve the accuracy closer to our target of 95%. Other than testing, I worked on adding code to determine if an attacker attempted to steal user belongings from the item stand. This involved introducing a buzzer and writing code to make it alert user if anything was removed from the stand while not in a check-out process. The implementation of the buzzer does not work yet, but in further weeks, I hope to further test it.
- My progress is slightly behind. This week, after introducing the buzzed to our system, our system unfortunately broke. As a result, we had to spend time to determine which component led to the failure. With my team members, we determined that the wireless transceiver were not working, so we had to get new ones. The additional time to debug this issue and receive new parts delayed out end-to-end testing, causing us to have less time to tune our algorithms. I plan to do more testing next week to prepare for the final demo, ensuring that our facial recognition system can accurately detect new faces and that our item stand can correctly find good positions for users to place their items.
- I plan to spend some time next week to further test the system and try to satisfy any design requirements that were not met. This involves doing more end-to-end testing, and improving the speed at which users can complete the check-in and check-out processes. I also plan to re-introduce the buzzer code so that we can detect if attackers have stolen items on the stand.
- As I’ve debugged the project, I found it necessary to learn more about facial recognition. I have had to look into the differences between SVM versus euclidean distance classifiers, and I’ve learned about ways to normalize faces. In terms of hardware, I’ve had the opportunity to learn about how to calibrate load cells and do wireless transmission to communicate between two Arduinos. I also have written more code in Arduino, so I improved my skills in this aspect. In terms of debugging, I have learned to do unit testing on individual components to locate which parts of the system led to failure. I have also relied on online tutorial and forums to learn about new approaches to solving problems dealing with both hardware and software components of our system, like load cells, motor, and facial recognition.
Carnegie Mellon ECE Capstone, Spring 2024 | Doreen Valmyr, Ryan Lin, Surafel Tsadik