A Graph-Enhanced Deep-Reinforcement Learning Framework for the Aircraft Landing Problem

TL;DR

This work formulates the Aircraft Landing Problem as a Markov Decision Process and introduces a graph-enhanced deep reinforcement learning framework that leverages a graph neural network encoder and an actor-critic policy to learn landing sequences. A safety-aware sequential assignment layer enforces wake-vortex separations, time-window constraints, and precedence relations while maintaining real-time performance. The approach uses a graph-based state representation, a multi-objective reward, and adaptive exploration to balance runway utilization, safety, and delay penalties. Experimental results on standardized Paris-Orly datasets demonstrate near-optimal runway throughput with substantial reductions in computation time (up to 99.95% faster than MIP) and real-time solution generation (often under 1 second), indicating strong potential for industrial deployment and real-time rescheduling in air traffic management.

Abstract

The Aircraft Landing Problem (ALP) is one of the challenging problems in aircraft transportation and management. The challenge is to schedule the arriving aircraft in a sequence so that the cost and delays are optimized. There are various solution approaches to solving this problem, most of which are based on operations research algorithms and meta-heuristics. Although traditional methods perform better on one or the other factors, there remains a problem of solving real-time rescheduling and computational scalability altogether. This paper presents a novel deep reinforcement learning (DRL) framework that combines graph neural networks with actor-critic architectures to address the ALP. This paper introduces three key contributions: A graph-based state representation that efficiently captures temporal and spatial relationships between aircraft, a specialized actor-critic architecture designed to handle multiple competing objectives in landing scheduling, and a runway balance strategy that ensures efficient resource utilization while maintaining safety constraints. The results show that the trained algorithm can be tested on different problem sets and the results are competitive to operation research algorithms. The experimental results on standard benchmark data sets demonstrate a 99.95% reduction in computational time compared to Mixed Integer Programming (MIP) and 38% higher runway throughput over First Come First Serve (FCFS) approaches. Therefore, the proposed solution is competitive to traditional approaches and achieves substantial advancements. Notably, it does not require retraining, making it particularly suitable for industrial deployment. The frameworks capability to generate solutions within 1 second enables real-time rescheduling, addressing critical requirements of air traffic management.

Figures (10)

• Figure 1: Architecture of the proposed GNN-based Actor-Critic network for aircraft landing scheduling. The model combines graph neural network layers for feature extraction with parallel actor and critic heads for landing time prediction and value estimation.
• Figure 2: Graph neural network architecture showing message passing between aircraft nodes. The GNN encoder transforms raw aircraft features into a learned representation that captures both spatial and temporal relationships.
• Figure 3: Actor-critic network architecture for learning landing policies. The actor network outputs landing time distributions while the critic estimates state values to guide policy improvement.
• Figure 4: Delay distribution over 10,000 episodes of training for all four scenarios.
• Figure 5: Total costs across training episodes for all four scenarios demonstrating cost optimization behavior.