Yulong Cao

When general-aviation aircraft leave a training airspace and several return to the circuit together, the tower controller must assign one of six allocation schemes to each aircraft within seconds. Historical records, however, resolve identical situations in different ways, so a purely supervised classifier trained on them saturates: on the SR-20 sample used here (= 20,000, six schemes) it reaches only 87.69 % accuracy and 86.85 % macro-F1. This paper moves the supervision target from the records to an explicit, auditable matching-rule model that encodes the operating regulations as a ternary structure with three-source evidence and a trend-strength gate; only its continuous coefficients are estimated by a conditional-logit fit. The rule posterior is distilled into a LightGBM student as a soft label. Against the deterministic rule standard of the held-out test split the student reaches 95.48 % accuracy, 94.99 % macro-precision and 95.04 % macro-F1, i.e., meets deployment requirements. The single-modality controls fail - a BiLSTM’s macro-F1 on the flight frames reaches 81.94 % and a Chinese-RoBERTa text classifier’s macro-F1 reaches 50.48 % - so the two modalities carry complementary information; a random forest’s macro-F1 reaches 90.89 % but is less auditable. Because every training label is generated by a named rule, the deployed matcher stays traceable, and retraining costs under one minute on a desktop CPU, which is the property a safety-critical tower function requires.

The increasing complexity of general aviation flight training poses significant challenges to tower controllers, who must reconcile safety separation standards with training efficiency under diverse and often unexpected conflict scenarios. This study proposes a BERT-BiLSTM-XGBoost multimodal feature fusion ensemble learning model that automatically matches optimal allocation schemes based on predicted conflict characteristics, thereby providing real-time decision support for tower controllers. The framework operates in three stages. First, a BERT model extracts high-level semantic features from textual data—including aircraft status descriptions, controller characteristics, pilot profiles, and airspace constraints—by leveraging bidirectional contextual encoding. Second, a Bidirectional Long Short-Term Memory (BiLSTM) network captures temporal dynamic features from sequential aircraft operation data. Third, the extracted semantic and temporal feature vectors are fused and fed into an XGBoost classifier, which performs the final allocation scheme matching. The XGBoost algorithm additionally provides feature importance scores, supporting the traceability requirements of tower control decision-making. Experimental results demonstrate that the proposed model achieves a matching accuracy of 97.2%, substantially outperforming baseline methods including Naive Bayes, linear SVM, and LightGBM. Moreover, the model reduces controller decision-making time by 75–80% in simulated tower control tests, validating its practical utility as an effective decision-support tool for general aviation tower operations.

By 2030, projected UAV operations may exceed one million concurrent flights in urban airspace, yet traditional fixed-distance separation methods fail to accommodate heterogeneous platforms. This paper introduces a three-tier hierarchical dynamic separation framework adapting minima across strategic (30–80 m category-specific baselines), pre-tactical (0.7–1.8× encounter-dependent scaling), and tactical (real-time 3D decomposition) timescales. Monte Carlo simulation across 100,000+ flight hours demonstrates 47% collision rate reduction versus fixed 30 m separation (0.008 vs. 0.015 per 1000 h, p < 0.001), 50% airspace utilization increase (18.4 vs. 12.3 UAVs/km³), 44% flight time penalty decrease (8.5% vs. 15.2%), and 99.97% ICAO-compliant TLS achievement (≤10⁻⁷ per flight hour) with real-time performance (78.5 ms for 20 UAVs). The framework provides an immediately deployable foundation for heterogeneous UAV traffic management.