TreeDoc

Inspiration

Commercial LiDAR hardware is slow, expensive, and above all, difficult to deploy. Yet, 1 in 3 smartphone users already carry powerful LiDAR scanning devices right in their pockets (via the iPhone 12 Pro/Pro Max and later). We saw this gap firsthand when one of our team members spent a semester surveying trees and natural landscapes in the Duke Forest, a process that required fleets of student volunteers to manually map the environment. Realizing this technology was already accessible, we set out to democratize environmental scanning by building a simple, DIY, autonomous LiDAR mapping platform. All you need is our app, a few 3D-printed mounts, and you're ready to go.

What it does

TreeDoc is an autonomous mobile scanning platform that generates high-fidelity LiDAR depth maps. By mounting iPhones on a motorized pivot atop an RC chassis, TreeDoc comprehensively scans its environment while autonomously navigating the terrain. This eliminates the need for manual, time-intensive human scanning and drastically cuts down research costs.

How we built it

We engineered a custom dual-axis scanning rig by combining a high-precision BLDC (brushless DC) motor with an ODrive motor controller to swivel the iPhone LiDAR modules. This assembly was mounted onto a repurposed children's RC ambulance toy chassis. We drove the entire hardware stack using a Raspberry Pi powered by a 6S LiPo battery, bringing it all together with custom 3D-printed enclosures and a lot of direct-to-board soldering.

Challenges we ran into

Syncing LiDAR point clouds across multiple iPhones was a massive hurdle. Apple's ARKit lacks robust native synchronization, which created a "double image" effect. Instead, we developed our own calibration pipeline to seamlessly stitch the incoming data streams together.
Virtually no components were compatible with each other. We had to splice cables, re-wire pins, and solder our microcontrollers directly to the proprietary RC car board.
Our project literally started smoking when tiny electrical errors from assembling everything at once fried our motor controller. This experience reinforced the hard lesson of iteratively wiring and testing our circuitry.
Sourcing parts with the exact dimensions we needed during the hackathon was difficult, teaching us that upfront planning and adaptability are necessary in the event of suppliers falling through.

Accomplishments that we're proud of

Successfully reverse-engineering and hijacking a children's toy ambulance RC control system.
Writing a robust data processing pipeline that can ingest real-time LiDAR streams from iOS to drive physical hardware.
Achieving highly accurate, usable depth maps from consumer hardware.
Successfully soldering and integrating a variety of formerly incompatible components into a unified, mobile robot.

What we learned

Building TreeDoc taught us the harsh realities of hardware integration. We were able to grow our skillset in mobile hardware design, especially in routing power and control signals for BLDC motors and ODrive controllers on a moving chassis. We also learned that keeping a strict inventory of materials and planning for component tolerances is just as important as the code itself.

What's next for TreeDoc

Our immediate next step is expanding the data pipeline to automatically convert the raw LiDAR point clouds into clean 3D meshes that researchers can easily export and save.

Overall, we want TreeDoc to be highly accessible. We plan to develop and distribute 3D-printable files so anyone can easily convert their iPhone into an affordable but quality scanner. Unlike alternatives on the App Store (like Record3D) that lock LiDAR streaming behind a paywall, our scanning software is completely open-source and free.