Reinforcement learning agents to tactical air traffic flow management

Abstract

Air traffic flow management (ATFM) is of crucial importance for the airspace control system, due to two factors: first, the impact of ATFM on air traffic control, including inherent safety implications on air operations; second, the possible consequences of ATFM measures on airport operations. Thus, it is imperative to establish procedures and develop systems that help traffic flow managers to take optimal actions. In this context, this work presents a comparative study of ATFM measures generated by a computational agent based on artificial intelligence (reinforcement learning). The goal of the agent is to establish delays upon takeoff schedules of aircraft departing from certain terminal areas so as to avoid congestion or saturation in the air traffic control sectors due to a possible imbalance between demand and capacity. The paper includes a case study comparing the ATFM measures generated by the agent autonomously and measures generated taking into account the experience of human traffic flow managers. The experiments showed satisfactory results.