The study of airspace sector demarcation can help controllers to deal with complex air situations, provide reference for air traffic control services, reduce the workload of controllers, and ensure safe and efficient airspace operation. Based on complex network theory, this paper proposes a sector division method based on a mean shift clustering algorithm and a Voronoi diagram. Firstly, a flight conflict network is constructed based on the structural characteristics of the airborne flight situation, combined with the three-dimensional velocity barrier method and the aircraft protected area model. Secondly, six network topology indexes, such as the total node degree, average point strength and network density, are selected to construct the flight-situation-assessment index system, and the busy and idle time segments of the airspace are divided according to the comprehensive network indexes; finally, based on the historical flight data, the airborne aircraft in the busy and idle time segments are clustered using the mean shift clustering algorithm and the Voronoi diagram method, respectively, so as to obtain a sector division scheme. The simulation results show that this method can equalize the network topology indexes and provide a reference for the scientific allocation of controllers’ energy.

As the demand for air transport continues to increase, air traffic congestion in the terminal area is becoming more and more serious. In order to assist the controller in efficiently handling the symmetrical activities of aircraft take-off or landing and alleviate traffic congestion, this paper proposes a method for identifying traffic congestion situations based on complex networks and a multiclass random forest algorithm with symmetrical characteristics. First, the approach points, departure points, waypoints, and navigation stations are used as nodes, the flight paths as edges, and the busyness of the paths as edge weights to construct a traffic network model for the terminal area. On this basis, five congestion situation recognition indicators are selected from the perspective of network topology, and a symmetric multiclass random forest algorithm is proposed to recognize the congestion situation. Finally, this method is compared with the situation recognition method based on the traditional random forest algorithm. The results of the simulation experiment show that compared with the traditional random forest algorithm, the proposed recognition model improves the recognition accuracy by 17.5%, can better handle symmetry information, and can accurately determine the traffic congestion situation in the terminal area.

Designing arrival routes in terminal airspace requires deciding how inbound streams should be grouped and where the resulting merge points should be placed. This paper searches convergence topology, merge-point positions and population-aware path decoding jointly at the network scale, treating the three as a single design problem. The problem is formulated as a bilevel optimization: an outer simulated annealing loop searches convergence topologies and merge-point locations, while an inner A* algorithm with population density embedded in its edge cost generates the track segments between structural nodes. Sound energy is accumulated once per unique structural segment, independently of how many arrival routes traverse it. The method is demonstrated as a single-case proof of concept on the north-landing configuration of Runway 35 at Shanghai Pudong International Airport. Over a preference sweep from distance-only to noise-only weighting, the optimized networks outperform the published procedures in both metrics for $w_N\in [0.0,0.6]$; at higher weights the proxy value falls further but the route length exceeds the baseline. The noise reduction forms a statistically flat plateau rather than a single working point: at $w_N=0.3$ the model-based, population-weighted noise proxy (the modelled population above a 65 dB single-event $L_{\mathrm{AE}}$ threshold) is reduced by 69.3% and the total route length by 3.4%, and the differences between neighbouring weights within the plateau are smaller than their run-to-run scatter. Attributing the reduction sequentially, a 3° continuous-descent profile alone lowers the proxy value by 29.1%, and the horizontal redesign of the convergence structure accounts for the remainder at 3° CDO. A sensitivity study across structural, algorithmic, population and acoustic assumptions shows that the reduction in the proxy is robust, whereas the route-length reduction holds only while the structural constraints are not substantially tightened.