Abstract

Modern air traffic management (ATM) systems rely on human-centric coordination between controllers and pilots, a framework currently challenged by increasing flight densities and diverse aircraft types in European airspace. This saturation results in heightened congestion, safety risks, and environmental impact. To address these limitations, this paper proposes a transition toward a "next-generation" ATM through enhanced airspace abstraction and situational awareness. This study contributes an analysis of contemporary ATM complexity and introduces a digital assistance system for conflict detection. Utilizing Long Short-Term Memory (LSTM) networks, the system evaluates temporal dynamics in surveillance data to classify error patterns. Training was conducted on large-scale models comprising thousands of flights, utilizing a 20–10–1 architecture with leaky ReLU and sigmoid activations, optimized via the ADAM algorithm and binary cross-entropy loss. Numerical findings demonstrate the efficacy of LSTMs in predicting critical disruptions, including adverse weather, cyber-attacks, and human-factor-related emergencies.