The project’s 10-week timeframe was extremely tight, and we were tasked with solving complex problems. We had to start from scratch, learning everything from Raspberry Pi programming to database setup, front-end development, and data transmission. I invested considerable time in developing a streamlined, AI-assisted learning & working flow that allowed for rapid testing and iteration of our hardware. As the team leader, I also organized regular work sessions to address challenges collectively, fostering a supportive environment that kept the team motivated. Despite the daily hurdles, including persistent software bugs, no one gave up. Time management was another major challenge, but we managed it well. Our sponsor was highly impressed with our dedication and outcomes, and we were the only team to have our paper accepted for the GHTC conference. Our project exceeded expectations in both scope and completeness, earning us second place in the final evaluation among all sponsored projects in our department. Reflecting on this journey, I am proud of my relentless learning ability and my "Make it happen" mindset, which drove me to seek external help and expand my knowledge, successfully navigating the entire product development lifecycle. This experience also deepened my appreciation for how AI can significantly empower designers and engineers.

Honestly, when we tested the third-generation prototype in the field, it didn’t perform as well as we had anticipated. Several factors contributed to this outcome. First, due to budget constraints, the IR sensors we used lacked precision. Reducing the monitoring frequency to meet energy consumption requirements further decreased sensor sensitivity, leading to some inaccuracies (as detailed in Experiment 3 of our paper). Second, material limitations meant that our prototype did not adequately address waterproofing, wind resistance, and high-temperature endurance. Third, the cost of the third-generation prototype was around $200 per unit, which needs to be reduced. Fortunately, by switching to a cheaper motorized Pi camera and a lighter microprocessor, we could potentially lower costs by nearly 50%. Despite these challenges, our project explored innovative solutions that were highly original. In an era dominated by Machine Learning models and LLMs, our decision to use a combination of DSP and computer vision, which are more cost-effective and user-friendly, was validated by the GHTC committee’s recognition. This project has laid a solid foundation for future iterations and improvements.