Autonomous air traffic separation assurance through machine learning

Abstract

With the increasingly severe airspace congestion problem, the aviation industry is facing huge challenges, including the surge in pressure on air traffic controllers, significant increase in flight delays, and frequent flight conflicts. Efficient conflict detection and resolution technology is the primary task to ensure flight safety, which is not only particularly important for complex and high-density airspace environments, but also has significance for maintaining flight order, preventing aircraft collisions, alleviating air traffic pressure, and ensuring air traffic safety. This paper proposes a flight separation assurance model based on reinforcement learning (RL) technology, aiming to address the shortcomings of existing conflict resolution models in terms of state and action dimensions that cannot meet the needs of real-world control scenarios. Based on the air traffic environment model, agents are allowed to learn separation assurance policies through simulation interaction. Simulation experiments show that the model can search for optimal policies, which can provide assistance for air traffic controllers in decision-making.