The rapid advancement of unmanned aerial systems has placed large fixed-wing drones at the forefront of next‑generation logistics and low‑altitude air mobility. Integrating these aircraft with manned civil aviation in shared airspace demands rigorous safety assessment. In our work, we develop a quantitative framework for collision risk and separation minima specific to fixed‑wing drones, and propose a methodology for designing protection areas along predetermined track lines. This paper summarizes our key findings, including a 4D trajectory prediction model, a conflict risk template aligned with navigation performance standards, and a set of separation values applicable to low‑altitude high‑density operations. Our results support the evolution of smart low‑altitude airspace and intelligent logistics innovation.
Current Status and Problem Statement
The industrial ecosystem for fixed‑wing drones is maturing rapidly. Under the guidance of civil aviation authorities, research consortia have been formed to promote technological breakthroughs and standardization. Our team has identified three critical gaps that must be addressed before fixed‑wing drones can routinely operate in mixed airspace with manned aircraft. First, the navigation and flight performance characteristics of fixed‑wing drones differ significantly from those of conventional airliners – they rely on GNSS with ground augmentation, fly at altitudes between 3 000 and 6 000 m, and exhibit unique data‑link vulnerabilities. Second, existing collision risk models (CRM) are tailored for manned aircraft and do not fully capture the navigation accuracy and automatic‑flight behavior of fixed‑wing drones. Third, no consensus exists on separation standards for mixed operations in low‑altitude, high‑density environments. Our research directly tackles these problems by establishing a 4D trajectory prediction method, a conflict geometry model, and a reference set of safety intervals.
Technical Approach
We adopted a three‑part methodology comprising 12 progressive steps. The core challenges are: (1) accurate 4D trajectory prediction for fixed‑wing drones; (2) design of a conflict risk model that incorporates flight‑phase performance and navigation uncertainty; and (3) computation of separation minima that guarantee an acceptable collision probability. The overall technical roadmap is summarized in Table 1 below.
| Key Problem | Solution Steps | Expected Output |
|---|---|---|
| Precise 4D trajectory prediction for fixed‑wing drones | Data preprocessing, clustering, spatiotemporal characterization | High‑accuracy 4D trajectory with position, time, and velocity |
| Conflict risk model for fixed‑wing drones | Analyze flight‑phase performance, design conflict template based on ANP | Collision probability expression with spatiotemporal variables |
| Safety interval for mixed operations | Simulate encounter scenarios (head‑on, overtaking, climbing/descending) | Separation minima for fixed‑wing drones relative to manned aircraft |
4D Trajectory Prediction for Fixed‑Wing Drones
We processed ADS‑B data from actual fixed‑wing drone operations to construct a reliable 4D trajectory prediction model. After clustering and cleaning the raw data, we built a multilayered processing architecture that accounts for each flight phase. The model outputs spatiotemporal positions that are then validated against historical flight logs. A representative comparison for an approach segment is given in Table 2.
| Waypoint | Predicted Altitude (ft) | Actual Altitude (ft) | Predicted Time (UTC) | Actual Time (UTC) |
|---|---|---|---|---|
| A1 | 19 426 | 19 426 | 08:00:00 | 08:00:00 |
| A2 | 18 825 | 18 820 | 08:01:48 | 08:01:50 |
| A3 | 17 726 | 17 730 | 08:03:17 | 08:03:15 |
| A4 | 17 726 | 17 726 | 08:03:17 | 08:03:17 |
| A5 | 1 390 | 1 388 | 08:08:37 | 08:08:39 |
| A6 | 12 776 | 12 780 | 08:10:07 | 08:10:05 |
| A7 | 11 949 | 11 952 | 08:11:08 | 08:11:07 |
| A8 | 6 883 | 6 883 | 08:16:02 | 08:16:02 |
The total predicted flight time was 16 min, altitude change from 19 426 ft to 6 883 ft, and lateral distance about 140 km. The root‑mean‑square error in altitude was less than 10 ft, and the time offset never exceeded 2 s, confirming the high fidelity of our model for fixed‑wing drones.
Conflict Risk Model Based on Flight Performance of Fixed‑Wing Drones
We built a conflict geometry that considers the Actual Navigation Performance (ANP) of fixed‑wing drones. Using the concept of a “protective volume,” we defined relative positions between two fixed‑wing drones with respect to the intersection point of their predetermined tracks. The mathematical formulation is given below.
Let i denote the drone index, and let the intersection point O define the coordinate axes x, y, z with a correction tolerance ε. The position of drone i at time t is:
$$
\mathbf{P}_i(t) = (x_i(t), y_i(t), z_i(t))^\mathrm{T}
$$
The relative distance between two drones 1 and 2 is:
$$
S_{12}(t) = \sqrt{ (x_1 – x_2)^2 + (y_1 – y_2)^2 + (z_1 – z_2)^2 }
$$
When considering the angular deviation from the predetermined track, we defined the half‑angles α and β as the angles between the line connecting the drones and the track intersection:
$$
\alpha = \arccos\left( \frac{ \mathbf{P}_1 \cdot \mathbf{P}_2 }{ |\mathbf{P}_1| |\mathbf{P}_2| } \right)
$$
$$
\beta = \alpha + \varepsilon
$$
To incorporate the temporal dimension, we introduced a time offset tΔ ∈ [0, tn−tn−1) and defined the spatiotemporal distance function:
$$
S_{12}(t, t_\Delta) = \sqrt{ \left[ x_1(t) – x_2(t+t_\Delta) \right]^2 + \left[ y_1(t) – y_2(t+t_\Delta) \right]^2 + \left[ z_1(t) – z_2(t+t_\Delta) \right]^2 }
$$
The rate of change of this distance is:
$$
\dot{S}_{12}(t, t_\Delta) = \frac{d}{dt} S_{12}(t, t_\Delta)
$$
This spatiotemporal formulation allows us to evaluate conflict probabilities dynamically. We computed the lateral and vertical protection boundaries using the accepted collision probability (ACP) value of 1 × 10−7 per flight hour, as recommended by international standards for fixed‑wing drones operating in joint airspace.
Safety Intervals for Mixed Operations Involving Fixed‑Wing Drones
We conducted a series of Monte‑Carlo simulations that modeled encounter scenarios between a fixed‑wing drone and a category‑C manned aircraft (e.g., A320). The airport structure and procedures used for the experiments are described in Table 3.
| Parameter | Value |
|---|---|
| Drone type | Fixed‑wing (modified Y‑5, AT‑200) |
| Manned aircraft type | Category C (B737/A320) |
| Altitude range | 1 000 – 6 000 m |
| Navigation accuracy (ANP) | 0.1 NM (lateral), 50 ft (vertical) |
| Acceptable collision probability per flight | 1 × 10−7 |
For the same airport terminal‑area geometry, we computed the lateral protection radius required to meet the ACP target. The conventional CRM for manned aircraft yielded a protection radius of 5.937 km. After incorporating the ANP‑based conflict model for fixed‑wing drones, the required radius reduced to 4.875 km. This reduction is attributable to the superior lateral navigation accuracy of fixed‑wing drones when operating under GNSS guidance with ground‑based augmentation. The resulting unshaded area in the airspace, which can be used for drone route planning, increased by approximately 22% compared to the manned‑only scenario. These quantitative intervals provide a practical reference for designing predetermined track protection zones in low‑altitude, high‑density airspace.
Conclusion
Our study has established a comprehensive methodology for determining protection areas along predetermined track lines for large fixed‑wing drones operating in integrated flight with manned aircraft. The key achievements are:
- A high‑fidelity 4D trajectory prediction model tailored to the navigation and performance characteristics of fixed‑wing drones, validated against real flight data.
- A spatiotemporal conflict risk template that incorporates ANP and ACP standards, providing a mathematical foundation for separation analysis.
- Quantitative safety interval values (lateral radius of 4.875 km) that can be directly used for designing fixed‑wing drone protection zones in low‑altitude, high‑density environments.
The results not only advance the theoretical understanding of fixed‑wing drone safety but also offer practical tools for air traffic controllers and regulators. Future work will extend the framework to account for multiple drone categories, varying payload weights, and non‑segregated airspace operations, thereby supporting capacity assessment and dynamic route optimization for the emerging smart‑low‑altitude ecosystem.