Simulation-Based Evolutionary Optimization of Air Traffic Management

Alessandro Pellegrini, Pierangelo di Sanzo, Beatrice Bevilacqua, Gabriella Duca, Domenico Pascarella, Roberto Palumbo, Juan José Ramos, Miquel Angel Piera, Gabriella GiganteIEEE Access 2020 · OpenReview

Abstract

In the context of aerospace engineering, the optimization of processes often requires solving multi-objective problems involving mixed variables and non-differentiable quantities. Air Traffic Management (ATM) is particularly complex due to the critical role of human controllers. This paper proposes a distributed architecture combining Agent-Based Modeling and Simulation (ABMS) with Evolutionary Computing (EC) to optimize ATM control plans. By utilizing fast-time simulations and real-world traffic data, the framework identifies Pareto-optimal solutions that balance safety, efficiency, and human workload.

Introduction

Air Traffic Management (ATM) is a high-stakes socio-technical system where technical performance is inextricably linked to human decision-making. Optimizing ATM is notoriously difficult due to the complexity of interacting entities, safety criticality, and conflicting objectives. Traditional mathematical models often fail to capture the nuances of human behavior, such as stress and situational awareness. This paper presents a "digital twin" environment where new policies can be rigorously tested and optimized using computational intelligence before real-world implementation.

Methodology

The proposed framework integrates Agent-Based Modeling and Simulation (ABMS) with Evolutionary Computing (EC) to navigate the complex search space of ATM configurations.

Agent-Based Modeling and Simulation (ABMS)

The system models individual actors—Pilots, Aircraft, and Air Traffic Control Officers (ATCOs)—as autonomous agents. A key innovation is the Projection Module, which acts as a broker for complex events and interactions. This allows the simulation to capture emergent behaviors that simple rule-based models miss.

The Simulation Infrastructure and Agent-Based Model architecture.

Evolutionary Computing (EC)

To optimize ATM configurations, the framework utilizes the NSGA-II (Non-dominated Sorting Genetic Algorithm II). This multi-objective algorithm identifies a set of Pareto-optimal solutions—trade-offs where one metric (e.g., fuel consumption) cannot be improved without worsening another (e.g., controller workload). This approach explicitly includes human-factor variables like mental workload and expertise as optimization criteria.

Distributed Architecture

Running thousands of high-fidelity simulations is computationally expensive. To address this, the authors implemented a Master-Slave parallel architecture. An "Orchestrator" distributes batches of simulations across a computing cluster to evaluate the "fitness" of different configurations in parallel, ensuring the system is viable for strategic and pre-tactical planning.

Experimental Validation

The framework was evaluated using real-world traffic data from the Eurocontrol Demand Data Repository 2 (DDR2), specifically focusing on historical scenarios in Italian airspace (Brindisi, Milan, and Padua) from 2013–2019.

Key Results

  • Validation Accuracy: The simulation's results for flight timeliness and sector occupancy deviated by less than 10% and 7%, respectively, when compared to historical data.
  • Optimization Viability: In high-complexity scenarios, such as the transition to free routing, the EC algorithm successfully identified configurations that minimized both occupancy and timeliness delays.
  • Computational Efficiency: Using a 136-core cluster, the system proved capable of handling the high-load demands of evolutionary optimization, making it a viable tool for planning hours to days before flights.

Conclusion

The integration of ABMS and EC provides a robust platform for the offline "what-if" analysis of ATM procedures. By decoupling simulation and optimization, the architecture remains modular and extensible. Future work will explore "any-time optimization" to reduce latency, potentially allowing the system to provide real-time assistive technology to controllers during live operations.

Summary updated

Abstract

In the context of aerospace engineering, the optimization of processes may often require to solve multi-objective optimization problems, including mixed variables, multi-modal and non-differentiable quantities, possibly involving highly-expensive objective function evaluations. In Air Traffic Management (ATM), the optimization of procedures and protocols becomes even more complicated, due to the involvement of human controllers, which act as final decision points in the control chain. In this article, we propose the use of computational intelligence techniques, such as Agent-Based Modelling and Simulation (ABMS) and Evolutionary Computing (EC), to design a simulation-based distributed architecture to optimize control plans and procedures in the context of ATM. We rely on Agent-Based fast-time simulations to carry out offline what-if analysis of multiple scenarios, also taking into account human-related decisions, during the strategic or pre-tactical phases. The scenarios are constructed using real-world traffic data traces, while multiple optimization variables governed by an EC algorithm allow to explore the search space to identify the best solutions. Our optimization approach relies on ad-hoc multi-objective performance metrics which allow to assess the goodness of the control of aircraft and air traffic regulations. We present experimental results which prove the viability of our approach, comparing them with real-world data traces, and proving their meaningfulness from an Air Traffic Control perspective.
BibTeX
@inproceedings{pellegrini2020simulation,
  author = {Alessandro Pellegrini and Pierangelo di Sanzo and Beatrice Bevilacqua and Gabriella Duca and Domenico Pascarella and Roberto Palumbo and Juan José Ramos and Miquel Angel Piera and Gabriella Gigante},
  title = {{Simulation-Based Evolutionary Optimization of Air Traffic Management}},
  booktitle = {IEEE Access},
  year = {2020},
  url = {https://openreview.net/forum?id=-_XtVvSSSjc},
}
Figure