ML + robotics research

UAV Sandbox

A simulation platform for testing how drones sense obstacles, assess risk and plan routes.

My role: Experiment design, sensor pipelines and ML tooling
Output: Simulation platform and repeatable experiments
Tools: Python · PX4 / Gazebo · PyTorch / ONNX
Status: Active research · Simulation only

Explore the research repository

Case study updated 2 October 2026

UAV Sandbox — AI-assisted development. Concept illustration of a drone, simulated substation, LiDAR arcs and a planned route. — AI-assisted development · Concept cover. View full image

What I built

Simulation: PX4/Gazebo flight with A* routing and local replanning.
Sensor tools: LiDAR and camera recording, replay and inspection.
ML workflow: controlled scenarios, dataset preparation and repeatable model evaluation.

How I used AI

I used AI coding tools to develop and refine the platform, then reviewed the code and checked its behavior in simulation.

Current status

The research platform is implemented. Formal learned-model comparisons are ongoing; real-airframe validation has not been performed.

Example output

Repository A-star route preview showing raw obstacles, inflated planning buffers, waypoints and a return route. — Actual A* planning preview · Simulation baseline, not a learned-model result. View full image ↗

See the code, experiment workflow and current research status.

Start a conversation

Good questions lead to useful work.

Research, building something, or comparing notes.