The Hardware

We started with humble components but ambitious goals:

• ESP32 DevKit as our brain (thanks to the IEEE workshop!)
• MPU6050 6-axis IMU sensor for tracking orientation
• 3 push buttons for full mouse functionality
• Breadboard foundation for rapid prototyping

Originally, we intended to use a Raspberry Pi based system, like the Raspberry Pi Zero, but with challenges with connecting to the device, we pivoted to using the ESP32. While a released product would feature much nicer case, we used a breadboard for easier testing and demonstrating the functionality. To improve the form, we used wires, rather than jumper cables for the breadboard.

The Software Revolution

The real work happened in the code. We handcoded the design and implemented sophisticated algorithms for nice cursor tracking.

• Complementary Filter: Fuses gyroscope with custom curves for rock-solid tracking
• Adaptive Bias Correction: Automatically compensates for sensor drift using Kalman-lite approach
• Non-linear Response Curves: Separate mathematical curves for cursor vs scroll modes
• Momentum System: Captures those subtle, slow movements with 0.88 decay rate
• Health Monitoring: Automatic I2C recovery when sensors misbehave

We wrote the code in C/C++ using Arduino IDE, to have better low level access and customization. We used the ESP32 BLE Mouse library to wirelessly connect our device as a Bluetooth HID mouse.

Team Coordination and Work

With only 2 ESP32s for 4 team members during the hackathon crunch, we had to get creative with collaboration. Additionally, with such a linear hardware project, where the connection required the gyroscope which required the buttons, it was hard to work together on some of the sections. Despite this, we divided the tasks along roughly obvious categories and worked together as much as possible. With our extra team power, we created a product website and polished it.

Precision vs. Usability

Our biggest technical challenge was achieving the perfect balance between sensitivity, stability, and the gyroscope's own limitations. Too sensitive meant the mouse was too jittery. Too slow meant sluggish frustration. We solved this with:

• Multiple sensitivity modes (Low/Medium/High)
• Smart deadzone algorithms to eliminate micro-movements
• Exponential smoothing for smooth tracking

We originally wrote all of our code by hand, which worked well, but the cursor was not as smooth as any commercial product. After building the working demo code ourselves, we decided to have AI write some custom curve algorithms for us as a last minute experiment to improve the controls. After several failed attempts with Gemini, which possibly made the code worse, and Copilot, which just made it wrong in different ways, Claude rewrote the majority of the code in an iterative experimentation process that resulted in a much smoother mouse, which has ended up being our final submission. So, the majority of our code will flag as AI produced, because these final rewrites changed the code significantly, but we originally wrote all of the code ourselves. If we had more time, we might have been able to code the customized calibration curves ourselves, but even working with AI, we needed to revise the code several times to achieve the best functionality.

Gyroscope Accuracy

Initially the MPU6050 gyroscope wasn't quite meeting our precision expectations. Our code (pre-AI) worked, but we ran into issues with the gyroscope precision, where small movements would be ignored and the gyroscope would reset its heading. This made for a very difficult challenge that we just didn't have time to fix. However, with some help from AI and several rounds of testing lasting into the early morning before the submission deadline, we developed custom calibration controls to mitigate this error.

Library Troubles

The T-vK ESP32 BLE Mouse library saved us a lot of work, which we immediately used for exploring hardware alternatives, improving the calibration, and building a website. However, it wouldn't originally run for us because of some type conversion errors in the library itself, and we almost had to rewrite the entire library from scratch. Thankfully, our team found the problem, fixed the library by hand, and compiled the code correctly for the rest of the team, all while multitask attending some of the great workshops hosted at the hackathon.

We Shipped a Complete Product!

In just 36 hours, we created a WORKING LIVE PRODUCT that:

• Seamlessly connects to all computers that support Bluetooth normally (Mac, Windows, most Linux...)
• Registers with the computer identically to a traditional mouse (so no weird connection issues)
• Includes expected UX features like drag & drop and scroll modes (makes for a nicer experience)
• Self-calibrates and recovers from errors automatically (all with a sarcastic attitude)
• Provides intuitive gesture-based control (no more flat mice!)

While we technically had ~36 hours for the competition, with planning, failed attempts with the Raspberry Pi, and a health(ish) sleep schedule, we completed the project in more like 12 hours, but even during this time, we made space to attend intereting workshops, the career fair, meals, and a few rounds of video games to keep the environment fun.

Technical Breakthroughs We Achieved

• Advanced sensor fusion: rivaling commercial motion controllers
• Zero-driver installation: works instantly via Bluetooth HID
• Power optimization: for extended battery life
• Robust error recovery: that keeps working even when sensors glitch
• Multi-modal operation: supporting cursor, scroll, and precision modes

Although we didn't have a battery pack to connect the mouse to, the mouse is wireless, so it can be plugged into the wall, a computer, or a portable battery.

From Zero to Hardware Heroes

Most of our team entered this hackathon with minimal hardware experience, but the motivation to try something new and have fun along the way. Because of a lack in embedded hardware mentorship, we ended up learning everything on the fly ourselves as we built the mouse. Through countless pinout diagrams, datasheets, and trial-and-error moments, we now genuinely understand:

• Breadboarding fundamentals and circuit design
• Embedded software development and real-time constraints
• Sensor fusion mathematics and signal processing
• Bluetooth protocols and HID device implementation

The Art of Ambitious Project Management

We learned that ambitious projects require not just technical skills, but strategic thinking about scope, team coordination, and time management under pressure. Even a linear project can be broken down into collaborative steps, and free team members can continue working on other aspects of the project design.

Version 2.0 Vision

• Custom 3D-printed ergonomic enclosure - no more breadboard aesthetic
• Higher precision IMU for even smoother tracking
• Rechargeable battery integration with USB-C charging
• Haptic feedback for button press confirmation
• Multiple device support - switch between computers seamlessly

Market Potential

AeroMouse actually addresses real pain points in the computer mouse market

• Presentation professionals who need freedom from podiums
• Home entertainment enthusiasts controlling media centers
• Accessibility users who benefit from ergonomic alternatives
• Remote workers seeking comfortable computing positions

A commercialization plan is way beyond our current scope, but if we can make a marketable product (not just wired to a breadboard) this would be possible to manufacture and distribute, an exciting possibility.

Technical Roadmap

Some other interesting ideas that we had are:

• Machine learning integration for specific gesture recognition
• Multi-user calibration profiles for improved callibration
• Extended gesture vocabulary (swipes, rotations, etc.)
• Cross-platform mobile app for advanced configuration