Using Swarm AI to Map a Cave Network

CS310: Dissertation

Date: December 2018 - April 2019

Cave exploration is dangerous and time-consuming. This project demonstrates how swarm AI could be used in a group of autonomous flying drones to navigate and explore a cave as efficiently as possible. The project also includes realistic cave environment generation.


  • Realistic cave environment generation - Starting from Simplex noise, a cellular automata and several flood fills create unique, realistic cave environments for the simulation to run on.
  • Individual drone searching - A single drone can successfully explore every cell in the cave using a semi-efficient approach.
  • Multiple drone searching - Multiple drones can work together to more efficiently explore the cave by communicating information between them and avoiding exploring the same locations.
  • Visualisation - View the process of the drone's exploration. Ability to see what cells the drone has explored, potential frontier cells and the next target cell.
  • Statistics - View statistics of each drone including the distance they have travelled and percentage of the cave they have explored.