Multi-Agent Quadrotor Hardware and Simulation
Description
Created a 3D-printed, crash-resistant, LLM-enabled quadrotor with high customizability to provide a platform for multi-agent coordination and AI agent research within the MAGICC Lab
Key Takeaways
- Designed parametric CAD parts to construct a 3D printed quadrotor and spherical protection cage
- Created Gazebo simulations of multiple quadrotors using ROS 2 communication and control
- Integrated an AI agent with ROS2 for natural-language control of multi-agent simulations
Timeline
Duration: May 2025 - August 2025
Total time: ~350 hours
Time commitment: ~30 hours a week for 12 weeks
Result
See the project presentation and documentation at the website 3D Printed Quadrotor
Technical Skills
Engineering Design
Learned how to pioneer a new system through iterative design and testing. Created CAD models from ideas and sketches.
Manufacturing - 3D Printing
Learned how to manufacture and fabricate various parts and components using various tools. Examples: custom parts with 3D printing, crimping and soldering, and more.
AI Agent Development
Learned how to create a custom AI agent using LangChain, LangGraph, Hugging Face, and Ollama
Software
Developed ability to create a robotic system in ROS 2 using Python. Introduced to Ubuntu and became proficient operating on a Linux system. Practiced good coding management and collaboration skills using GitHub. Created reproducable code by implementing a custom Docker container.
Robotics/Electronics
Gained experience with Gazebo and Rviz simulations. Learned how to use PX4 flight controllers (Pixhawk and Pixrace Pro). Used a Raspberry Pi to run ROS2 on a Linux OS to control the quadrotor mid-flight via ssh on a ground station laptop.