Kaden Du

Watch the Demo Here

DRIFT was a drone designed to deploy tracking units on high-speed vehicles, intended for aid in police chases. The project was designed for Northeastern University's Senior Capstone Design competition, and we were awarded 1st place in Spring 2025. The drone was constructed using mainly OTS components, and utilized computer vision (YOLO) to track and follow vehicles, then following an approach algorithm, where a magnetic tracking unit was deployed to attach to a vehicle. The vehicle could then be monitored remotely over a LoRa interface. ️

My main responsibility for this project was designing our custom GPS tracking unit. The device included a GPS module with ceramic patch antenna, which would communicate data over I2C to the MCU. The position data was then forwarded over LoRa via the on board chip antenna to a receiver unit, and data was passed over USB to a host computer, where location was displayed. The device was designed to act as both transmitter and receiver, and included battery charging electronics, UART-USB conversion, and on board power conversion to support the main payload. Once the board was designed, I took s-parameter measurements using a VNA, and optimized a matching network to guarantee LoRa efficiency. All of the firmware for the device was written by me in C++, utilizing the STM32 HAL libraries. The final version of the board was tested to have a range of greater than a kilometer with clear line of sight (not tested until failure), but had some issues when the path was obstructed.

CSTAR is a robot designed to detect structural damages in concrete, namely parking structures, and was designed through Generate Product Development in spring of 2024. The core functionality of the robot is the acoustic sensing systems, in which on board microphones capture the sound of a mechanical "sounding tool" attached to the bottom of the robot. To keep any on board functionality lightweight, the audio data is simply forwarded over Bluetooth to a host, where the signal is processed (mainly by taking the fft and analyzing frequency). As the robot should be autonomous, it utilizes LiDAR and ultrasonic sensors, using SLAM algorithms to navigate and avoid obstacles. The robot tracks its position, also trasmitting to the host in order to create a map of structural integrity. The robot is battery powered and features on custom power electronics for regulation.

I was the Electrical Lead for this project. My role was architecting the electrical system, and overseeing all electronics and software design for this project. As well, mentorship was a significant part of my role. Many team members were new to electornics and PCB design, so I taught these newer memebers the design process through this project. This was one of my favorite projects to work on, both due to the novelty of the applicaiton, and a great team to collaborate with and mentor.

RoboUmp is an autonomous baseball umpire robot, designed to call balls and strikes using computer vision. The robot utilizes a camera to capture video of the pitcher and batter, then processes the video using a YOLO model to detect the ball and its trajectory. Using this data, the robot determines whether the pitch was a ball or strike, and communicates the call via audio and the on-board touchscreen.

I was the Project Lead for RoboUmp during the fall of 2023. My role was mainly organizational, planning for the team, communicating with the client, and ensuring seamless integration between teams. This project taught me a lot about mechancal, electrical, and software integration, adn how to manage a large, multidisciplinary team.