Using Reinforcement Learning for Unmanned Aerial Vehicle Conflict Resolution in a CTR Environment
Joaquin Vico Navarro, Juan Antonio Vila Carbó · IntechOpen eBooks · 2025
Growing demand for Unmanned Aircraft System (UAS) operations in urban areas within Controlled Traffic Regions (CTRs) around airports poses a significant challenge to UAS traffic management. Current U-space regulations allow temporarily segregating sections of the CTR airspace to carry out these operations, and UAS must react to changes in the airspace structure of the CTR, which restrict flights into segregated areas by terminating the flight. This work faces the problem of UAS performing less restrictive operations of UAS inside a CTR. These operations are assumed to be defined by a flight plan that does not necessarily enclose them into a segregated area. This is the case, for example, of UAS flying between two heliports inside a CTR. Achieving such operational level in a CTR is based on the concept of Dynamic Airspace Reconfiguration (DAR), which allows the definition of temporary no-fly zones (NFZs) to preserve safety of manned aviation and lets ATC dynamically restructure them upon a safety event. The proposed solution enables UAS to react to airspace restructuring without necessarily terminating the flight and to self-manage Conflict Resolution (CR) to keep traffic separations without the intervention of ATC. This is achieved using a Multi-Agent Reinforcement Learning (MARL) system with decentralized decision-making that provides UAS guidance after a CTR restructuring and keeps separation with all types of traffic in the CTR while accomplishing the predefined mission. This offers several benefits, including dynamic rerouting for UAS, increased efficiency, and greater scalability with reduced reliance on a centralized traffic control unit.