Logo
← Back to projects

Autonomous Drone Mapping

Applications not yet open

Simulation-first drone stack pairing SLAM with reinforcement learning path planning, built to transfer onto real hardware.

Project Manager
Nathan Miller

Computer Vision Process Control Robotics
Ran
Running since Spring 2026
Commitment
4-6 hrs/week
Openings
5-7

What you need coming in

C++ and Python proficiency including NumPy and PyTorch. Previous robotics experience, including SLAM, via FIRST or school organizations. Experience with Linux CLI, VMs, and Docker. ML/AI coursework such as CS373 and CS471, or similar project-based experience.

We aim to develop a simulation-first autonomous drone pipeline that maps real-world environments. The primary goal is completing a drone navigation stack using CAD modeling, physics simulations, SLAM, and reinforcement learning path planning. Once the drone system works, we plan to coordinate and optimize a swarm of drones to map virtual spaces simultaneously. Research on these systems is independently scattered; our innovation lies in combining them into a unified project optimizing for coverage completeness, energy efficiency, and flight time.

The overarching objective is a codebase transferable to physical hardware. By the end of the semester we plan to demonstrate a single drone autonomously exploring unknown simulated environments while constructing real-time 3D maps. We use gym-pybullet-drones for rapid RL iteration and Gazebo with PX4 SITL for the flight stack — tools that ease the transition from simulation to physical robotics in future semesters.

Technical elements

  • SLAM pipeline: ORB-SLAM3, FAST-LIO2 (LiDAR-based), Nav2, or MATLAB Navigation Toolbox for real-time localization with loop closure
  • RL navigation: PPO, SAC, DQN, or DDQN algorithms
  • Multi-drone coordination: decentralized frontier-based exploration with Hungarian algorithm task allocation and ORCA collision avoidance
  • Simulation stack: gym-pybullet-drones for algorithm development, Gazebo for system integration

Published research advancing robotics SLAM systems, including ORB-SLAM3 and FAST-LIO2, is the culmination of years of doctoral work. This project distills realistic milestones while maintaining technical rigor, giving members research-adjacent experience that bridges classroom knowledge and industry-level robotics engineering.