Parsa Rezaei

Work · NASA MINDS competition

Autonomous Aerial Docking and Recharge

NASA MINDS 2024 - Finalist

A heavy-lift carrier UAV acting as an airborne charging hub and data relay, letting micro-drones dock in flight to recharge, offload sensor data and borrow GPU cycles.

Parsa Rezaei beside the heavy-lift carrier quadcopter, showing its carbon arms, GPS module and flight controller
Parsa Rezaei beside the heavy-lift carrier quadcopter, showing its carbon arms, GPS module and flight controller

The problem

Micro-drone endurance is measured in minutes, which caps any mission built on them. Landing to recharge surrenders the position and the time.

Docking two aircraft in flight is a precision control problem where GPS is nowhere near accurate enough.

What I did

Built a vision-based docking system using OpenCV and visual-servo PID loops, closing the loop on what the camera sees instead of on position estimates.

Coordinated the swarm over ArduPilot and MAVLink with a 5 GHz Wi-Fi mesh plus LoRa telemetry, so control and bulk data used appropriate links.

Ran distributed compute across Jetson Xavier NX and Orin Nano nodes under k3s, turning the docked swarm into a compute cluster rather than just a charging stop.

Result

Mission endurance extended from minutes to hours, with the swarm doubling as a distributed compute and relay platform. Finalist at NASA MINDS 2024.

Get in touch
7 figures

From the report

Bench photographs and measured data from the project's own report.

IR lock-on range falls off sharply as approach angle grows
GPS-only tracking failed 37% of dockings; sensor fusion nearly eliminated it
LIDAR cut post-touchdown bounce height substantially
Where telemetry latency goes once Kyber encryption is in the loop
The LED array geometry the vehicle homes on, seen from above
VS1838 infrared receiver used for the terminal approach
ESC temperature logged across a docking cycle