The rapid evolution of urban air mobility (UAM), particularly in densely populated regions like China, is fundamentally reliant on the safe and efficient integration of unmanned aerial vehicles (UAVs) into low-altitude airspace. A critical bottleneck for scaling high-density operations lies in the design of vertiport terminal airspace—the structured volume where drones execute their final approach, landing, take-off, and initial departure maneuvers. Conventional air traffic management principles are ill-suited for the unique constraints of urban environments, characterized by complex obstacle fields, high traffic demand, and the need for automated, predictable flows. This necessitates the development of novel, optimized airspace structures specifically tailored for drone operations. This paper addresses this challenge by proposing a systematic methodology for the optimal structural design of UAV approach and departure procedures within urban obstacle environments in China.
The proliferation of China UAV drone applications, from logistics and inspection to air taxi services, imposes unprecedented demands on urban airspace. Existing research has explored various airspace concepts, from free-flight and layered structures to more organized designs like circular holding patterns and funnel-shaped corridors. While these provide foundational ideas, significant gaps remain. Many models address only arrival or departure in isolation, lacking an integrated framework for mixed operations. Key structural parameters are often predefined statically, without mechanisms for dynamic adjustment based on real-time traffic density or specific urban morphology. Crucially, the impact of external urban obstacles and the enforcement of precise safety separations within the structure are frequently underrepresented. These limitations hinder both the efficiency of airspace utilization and the safety assurances required for widespread adoption of China UAV drone services in complex cities.
To overcome these limitations, this research introduces a holistic optimization framework. The core contribution is a parameterized, multi-layer funnel-shaped hybrid airspace model that unifies approach and departure flows. An airspace efficiency metric is formulated as a weighted sum of five key performance indicators: capacity, safety, flight efficiency, airspace utilization, and system buffering capability (measured by the number of holding points). A Genetic Algorithm (GA) is then employed to search the high-dimensional parameter space—including layer radii, heights, and vertical spacings—to identify the configuration that maximizes this composite objective function. The optimization explicitly accounts for static urban obstacles, pruning infeasible flight paths and holding points to ensure operational safety. Through extensive simulation-based analysis under various urban scenarios and operational rules, this methodology demonstrates its effectiveness in generating safe, efficient, and high-capacity airspace structures for China UAV drone operations.
1. Problem Description and Urban Airspace Challenges
The design of terminal airspace for urban drone operations presents a multi-faceted optimization problem constrained by the physical environment, vehicle performance, and operational safety rules. The primary challenge is to structure a finite volume of airspace around a vertiport to sequence, merge, and separate continuous streams of arriving and departing drones while avoiding conflicts both amongst themselves and with surrounding obstacles.
Urban environments in China are typically characterized by dense and irregular obstacle fields, including high-rise buildings, towers, and other infrastructures. These obstacles impose hard constraints on the available airspace, potentially blocking certain approach/departure paths and reducing the feasible design space for holding patterns. A successful design must inherently incorporate obstacle avoidance, not as a post-processing step, but as an integral part of the structural definition. Furthermore, the operational paradigm for China UAV drone fleets often involves a high degree of automation with predefined flight paths. This requires the airspace structure to be highly structured yet flexible enough to accommodate varying traffic levels. Key conflicting objectives arise: maximizing throughput (capacity) often requires more holding points and larger airspace, which may increase flight distances (reducing efficiency) and potentially lower airspace utilization. Conversely, a very compact design might be efficient but offer insufficient buffering, leading to congestion and decreased capacity. Safety, measured by the minimum separation maintained between drones, is a non-negotiable constraint that directly interacts with all other objectives.
Therefore, the problem can be formally stated as: Given a vertiport location, a set of surrounding urban obstacles, drone performance characteristics (speed, climb/descent rates), and mandated safety separation standards, determine the geometric parameters of a multi-layer funnel airspace structure that optimizes a composite measure of operational performance.
The relevant parameters and their typical relationships are summarized below:
| Design Parameter | Symbol | Typical Influence on Performance |
|---|---|---|
| Upper Layer Radius | $$R_u$$ | ↗ Increases buffering capacity, may ↗ flight distance. |
| Lower Layer Radius | $$R_l$$ | ↘ Reduces final approach time, affects bottleneck spacing. |
| Vertical Spacing | $$\Delta H$$ | ↗ Improves safety separation, ↘ may reduce capacity rate. |
| Approach Point Height | $$H_{ap}$$ | Critical for sequencing arrivals/departures. |
2. Methodology: An Optimization Framework for Airspace Structure
2.1 Multi-Layer Funnel Airspace Model
We propose a structured, multi-layer funnel-shaped airspace model that integrates both arrival and departure procedures. The model is defined by several concentric cylindrical layers and key points, as illustrated conceptually in the figure. The structure is parameterized to allow optimization:
- Assembly Layer: The outermost buffer zone. Arrival streams from multiple fixed-direction entry routes merge here.
- Upper Holding Layer: A cylindrical layer where drones are sequenced after the assembly layer.
- Lower Holding Layer: A smaller cylinder beneath the upper layer for final sequencing before the final approach.
- Approach Point (AP): A single, critical three-dimensional point directly above the vertiport. All arriving drones must pass through this point before vertical descent, and all departing drones ascend vertically to occupy it before horizontal departure. It is a shared, mutually exclusive resource.
- Vertiport & Departure Paths: The ground-level landing pads and dedicated departure routes leading out of the core airspace.
The geometric design variables for optimization are the radius and altitude of the upper holding layer ($$R_u, H_u$$), the radius and altitude of the lower holding layer ($$R_l, H_l$$), and the altitude of the approach point ($$H_{ap}$$). The altitudes of the assembly layer and departure routes are fixed based on the highest obstacle and operational ceiling. Holding points within each layer are spaced evenly around the circumference, with the number $$N$$ being a function of the layer radius $$R$$ and the minimum required horizontal separation $$d_{sep}$$ between points:
$$ N = \left\lfloor \frac{2\pi R}{d_{sep}} \right\rfloor $$
This creates a direct link between the structural size and the system’s instantaneous holding capacity.
2.2 Performance Metric & Objective Function
The optimality of an airspace structure is evaluated using a weighted multi-criteria objective function $$Z$$. Five key performance indicators (KPIs) are normalized and aggregated:
1. Capacity (C): The total number of drones successfully completing both landing and take-off within a saturated simulation period.
$$ C = \sum_{i=1}^{N_{total}} I_{land}(i) + \sum_{j=1}^{N_{total}} I_{dep}(j) $$
where $$I$$ is an indicator function.
2. Safety (S): The minimum 3D Euclidean distance observed between any two drones during the entire simulation. A larger value indicates a greater safety margin.
$$ S = \min_{i \neq j, t} \sqrt{(x_i(t)-x_j(t))^2 + (y_i(t)-y_j(t))^2 + (z_i(t)-z_j(t))^2} $$
3. Efficiency (E): The maximum total flight distance traveled by any drone within the terminal airspace, from entry to landing (or take-off to exit). Minimizing this is desirable.
$$ E = \max_{i \in D} \int_{t_{enter,i}}^{t_{end,i}} \sqrt{\dot{x}_i^2 + \dot{y}_i^2 + \dot{z}_i^2} dt $$
4. Airspace Utilization (U): The ratio of the effective volume used by the funnel structure to the volume of its bounding cylinder. It measures geometric compactness.
$$ U = \frac{ H_{u-l}(R_u^2 + R_l^2 + R_u R_l) / 3 + \text{Lower Volume} }{ \pi R_{max}^2 H_{total} } $$
where $$H_{u-l}$$ is the vertical distance between layers, and $$R_{max} = \max(R_u, R_l)$$.
5. Holding Point Count (N): The total number of waiting points in the upper and lower layers, representing the system’s buffering capability.
$$ N = N_u + N_l = \left\lfloor \frac{2\pi R_u}{d_{sep}} \right\rfloor + \left\lfloor \frac{2\pi R_l}{d_{sep}} \right\rfloor $$
The composite objective function $$Z$$ to be maximized is:
$$ Z = w_C \cdot \tilde{C} + w_S \cdot \tilde{S} + w_E \cdot (1-\tilde{E}) + w_U \cdot \tilde{U} + w_N \cdot \tilde{N} $$
where $$\tilde{C}, \tilde{S}, \tilde{E}, \tilde{U}, \tilde{N}$$ are the normalized values (0 to 1) of each KPI, and $$w$$ are their respective weights. For this study, emphasizing capacity and robustness for high-density China UAV drone operations, the weights are set as: $$w_C=0.35, w_S=0.15, w_E=0.10, w_U=0.10, w_N=0.30$$.
2.3 Optimization with Genetic Algorithm and Obstacle Handling
The search for the optimal parameter set $$\mathbf{P^*} = [R_u^*, H_u^*, R_l^*, H_l^*, H_{ap}^*]$$ is performed using a Genetic Algorithm (GA). GAs are well-suited for this non-linear, simulation-based optimization with a potentially multi-modal fitness landscape.
- Encoding & Initialization: Each chromosome encodes the five design parameters as real numbers within predefined feasible ranges (e.g., $$R_u, R_l \in [20, 100]$$ m). An initial population is randomly generated.
- Fitness Evaluation: For each chromosome (airspace design), a high-fidelity discrete-event simulation is run. The simulation models the flow of drones according to the structured procedures, enforcing separation rules and resource conflicts (e.g., exclusive use of the Approach Point). The KPIs are extracted from the simulation results and combined into the fitness value $$Z$$ using the objective function.
- Obstacle Integration: Before simulation, a static obstacle-checking module is applied. It removes any holding point or path segment that intersects with modeled urban obstacles (represented as 3D cylinders or prisms). This ensures all evaluated structures are inherently safe from static terrain.
- Genetic Operations: The population evolves over generations through selection (tournament selection), crossover (simulated binary crossover), and mutation (polynomial mutation). Constraints on layer ordering ($$H_u > H_l > H_{ap}$$) and minimum vertical separation are enforced during these operations.
- Termination: The GA runs for a fixed number of generations or until convergence, outputting the parameter set with the highest fitness value as the optimal design.
The integrated optimization and simulation framework ensures that the final airspace structure is not only geometrically feasible but also operationally effective under realistic China UAV drone traffic and urban constraints.
3. Case Study & Simulation Analysis
3.1 Simulation Setup and Comparative Configurations
To validate the proposed methodology, a case study was conducted simulating a vertiport in a generic Chinese urban district with scattered high-rise buildings. The drone fleet was homogeneous, with a cruise speed of 3 m/s and vertical speeds of 5 m/s (climb) and 3 m/s (descent). The mandatory safety separation between drones in flight was set to 10 m, and the holding point separation was 20 m.
The GA was configured with a population size of 50 for 30 generations. The optimal design found (labeled Optimal-U for urban) had the following parameters: $$R_u = 84m, H_u=100m, R_l=24m, H_l=42m, H_{ap}=20m$$.
For comparison, six alternative airspace structures with distinct geometric features were manually designed and simulated under identical conditions:
| Config. Name | Geometric Characteristic | Design Intent |
|---|---|---|
| Cylinder | $$R_u = R_l$$ (Large) | Maximize holding space |
| Flat | Small vertical spacing | Minimize flight time |
| Tall | Large vertical spacing | Maximize safety buffer |
| Compact-Funnel | Small $$R_u$$ and $$R_l$$ | High efficiency |
| Inverted-Funnel | $$R_l > R_u$$ | Unconventional design |
| Open-Optimal | GA-optimized for open field | Baseline without obstacles |
3.2 Performance Comparison and Results
The performance of all seven configurations was evaluated. The results, normalized for clarity in the composite score, are presented in the table below. The proposed Optimal-U design demonstrates a superior balance across all metrics.
| Configuration | Capacity (C) | Safety (S) [m] | Efficiency (E) [m] | Utilization (U) | Hold Points (N) | Composite Score (Z) |
|---|---|---|---|---|---|---|
| Optimal-U (Proposed) | 94 | 17.83 | 543 | 0.270 | 28 | 0.782 |
| Cylinder | 45 | 39.00 | 570 | 0.541 | 4 | 0.423 |
| Flat | 80 | 10.32 | 564 | 0.391 | 9 | 0.601 |
| Tall | 43 | 24.82 | 639 | 0.201 | 23 | 0.514 |
| Compact-Funnel | 89 | 12.48 | 547 | 0.440 | 5 | 0.655 |
| Inverted-Funnel | 51 | 16.97 | N/A* | 0.544 | 32 | 0.521 |
| Open-Optimal | 113 | 16.83 | 543 | 0.301 | 32 | 0.801 |
* Operation failed due to structural inefficiency causing deadlocks.
Analysis of Key Findings:
- Superior Balance: The Optimal-U design does not necessarily lead in any single KPI but achieves the highest composite score (Z). It successfully trades off the extremely high capacity of the open-field design (which is infeasible in cities) against the safety risks of the “Flat” design and the very low capacity of the overly cautious “Cylinder” and “Tall” designs.
- Impact of Obstacles: Comparing Optimal-U with Open-Optimal highlights the performance cost of urban constraints. Capacity drops from 113 to 94, and holding points reduce from 32 to 28 due to obstacle pruning, yet safety and efficiency are maintained.
- Structural Intelligence: The GA-derived parameters (e.g., $$R_l=24m$$, $$H_l=42m$$) show a non-intuitive optimal point. A moderately compact lower layer with sufficient vertical spacing from the approach point proves more effective than either extremely wide or narrow designs.
- Robustness of Funnel Shape: The pure “Cylinder” (no radial gradient) and “Inverted-Funnel” perform poorly, confirming that a convergent funnel shape is fundamentally efficient for sequencing flows towards a single point, a critical requirement for China UAV drone vertiport operations.
3.3 Sensitivity Analysis: Safety Separation & Environment Density
Further analyses were conducted to test the robustness of the methodology under varying conditions critical for China UAV drone regulation.
A. Effect of Safety Separation Rules: The optimization was rerun with stricter (15m en-route / 30m holding) and more relaxed (5m / 10m) separation rules.
– Stricter Rules: Resulted in an optimal design with fewer holding points (N=16) and a lower capacity (C=83), but a higher safety margin (S=21.28m). The structure became more vertically stretched to accommodate the increased spacing.
– Relaxed Rules: Yielded a design with a larger radius and many more holding points (N=60), boosting capacity to 127, but with a lower safety margin (S=10.00m).
This demonstrates the method’s ability to adapt the airspace geometry to the prescribed safety standard, clearly quantifying the inherent trade-off between capacity and separation assurance.
B. Effect of Obstacle Density: The environment was modified to represent a sparser (open plaza) and a denser (central business district) setting.
– Sparse Obstacles: The optimal design increased the upper layer radius to its maximum, utilizing the available space to add holding points, slightly improving capacity.
– Dense Obstacles: The optimal design became more compact and taller, squeezing the operational volume between buildings. Capacity was maintained but at the cost of reduced buffering (lower N) and slightly higher risk (lower S).
This confirms the algorithm’s effectiveness in generating site-specific solutions, a vital capability for deploying China UAV drone networks across diverse urban landscapes.
4. Conclusion and Implications for China UAV Drone Operations
This paper presented a systematic, optimization-driven methodology for designing the terminal airspace structure for UAV approach and departure procedures in complex urban environments. By formulating a multi-layer funnel model, defining a comprehensive multi-objective performance metric, and employing a Genetic Algorithm coupled with obstacle-aware simulation, the method successfully identifies airspace geometries that optimally balance capacity, safety, efficiency, and resource utilization.
The case study results demonstrate that simply enlarging airspace does not guarantee higher performance; an intelligently shaped, moderately compact funnel structure outperforms various intuitive alternatives. The methodology is robust, adapting effectively to different safety regulations and urban obstacle densities, providing a powerful tool for planners and regulators. For the burgeoning China UAV drone industry, this approach offers a scientifically grounded path to design vertiport airspace that is safe, efficient, and scalable. It enables the customization of airspace structures to the unique constraints of specific cities, ensuring that the integration of UAM into China’s urban fabric is both technically sound and operationally viable. Future work will focus on integrating dynamic traffic demand, heterogeneous drone performance, and more complex, multi-vertiport network interactions.