Autonomous conflict resolution for high-density urban operations using deep reinforcement learning

Master thesis (2023)

Authors

Ricardo Santana Aerospace Engineering

Contributors

Alexei Sharpanskykh Air Transport & Operations - Aerospace Engineering (supervisor 1)
Faculty
Aerospace Engineering
Copyright: © 2023 Ricardo Santana
Artificial intelligence Deep Reinforcement Learning Conflict Resolution Urban Air Mobility
More Info
expand_more

Published Date

03-04-2023

Language

English

Reuse Rights

Other than for strictly personal use, it is not permitted to download, forward or distribute the text or part of it, without the consent of the author(s) and/or copyright holder(s), unless the work is under an open content license such as Creative Commons.

Faculty

Aerospace Engineering

Abstract

Low-altitude, high-density air traffic is expected to grow in the coming decades with several companies being certified to initiate urban operations for both freight and passenger transport. However, traditional human-centered Air Traffic Control operations (ATCos) are not scalable to handle the increased demand to maintain safe separation between aircraft. Thus, new autonomous solutions to resolve potential conflicts between aircraft must be developed, allowing humans to supervise and understand machine actions. We propose a centralized Deep Reinforcement Learning (DRL)-based framework to provide speed advisories to vehicles, ensuring safe and efficient operations. We used Proximal Policy Optimization (PPO) to train the intelligent controller and show that our framework is capable of handling challenging merging point settings. We evaluated our model extensively using a custom OpenAI Gym environment and proved that it can achieve a 99% success rate in conflict resolution across multiple merging point configurations.