AI-Driven Drone Development

Currently, I am developing a compact AI-driven drone with advanced autonomous capabilities. This project integrates custom PCB design, 3D printing, and sophisticated machine learning algorithms, all orchestrated by a Raspberry Pi acting as the flight controller. My aim is to create a drone that stands out in the market by addressing some of the most pressing limitations of current products.

Many drones on the market lack the ability to perform real-time person, face, and obstacle detection or avoidance. Those that do integrate AI often rely heavily on cloud-based processing, introducing latency and reducing operational reliability in remote or offline areas. Those that don't are limited to basic detection tasks—for example, identifying a single person in the frame—and their functionality can fail if primary sensors experience errors or malfunctions. This lack of redundancy and situational awareness makes existing solutions less robust and prone to failure.

In addition to these technical shortcomings, many current drones are prohibitively expensive, often costing thousands of dollars. Despite their high price, they offer little to no modularity, locking users into rigid designs that cannot be easily upgraded or customized to suit different needs or applications.

My goal is to tackle these challenges, to make a drone that features edge-based AI processing for reliable, real-time detection and decision-making. This along with various sensors will give the drone a full understanding of its environment, allowing for easy/reliable following capabilities. Along with this having modularity will allow for the system to be debugged and fixed easily, swapping individual parts if necessary or simply upgrading them.

To accomplish this many custom parts/algorithms are necessary, on the right you can see some of my work. These consist of the output of my current AI and some of the boards such as an ESC (Electronic Speed Controller), SIB (Sensor Integration Board), and PDB (Power Distribution Board). I am actively working on new versions of these components that add more functionality, for example, my new SIB will utilize the Hailo-8 AI chip for much faster inference and a more robust 9-axis system less vulnerable to environmental noise.

With these improvements, the drone should easily be able to handle following-mode tasks, and additional functionalities like weather data and advanced video capturing will prove a more robust system. When I have a more complete system, I will explain it in much better detail and most likely make a few videos on it.