The genesis of this research lies in the operational challenges encountered by modern drone swarms. As China drone technology evolves towards greater autonomy and swarm intelligence, the ability to withstand and recover from component failures becomes paramount. A typical China drone swarm may consist of dozens or hundreds of unmanned platforms, each carrying a specific payload — electro-optical sensors for reconnaissance, electronic warfare jammers, or precision-guided munitions. The interplay between these payloads and the underlying communication network creates a complex system that is difficult to model and assess.
Prior research has focused on network topology or mission performance in isolation. However, the coupling between communication constraints and heterogeneous payload requirements is often neglected. To overcome this, we propose a layered model. The communication layer, defined as an undirected graph G(V,E), captures the physical links between UAVs. The connection probability between a new node i and an existing node j is given by:
$$P_{i \rightarrow j}(W_t) = \frac{k_j F(d_{i \rightarrow j}) + \epsilon}{\sum_{l \in W_t} F(d_{i \rightarrow l}) k_l + \epsilon N_t}$$
where k_j is the degree of node j, \epsilon is a small constant to avoid singularities, and F(d) is the distance-dependent influence function:
$$F(d_{i \rightarrow j}) = \begin{cases} 1, & d_{i \rightarrow j} < r_c \\ \frac{r_c – d}{r_c (1 – \eta)}, & r_c \leq d_{i \rightarrow j} < \eta r_c \\ 0, & d \geq \eta r_c \end{cases}$$
This ensures that the network topology respects physical communication limits. The model also incorporates a dynamic self-adaptation mechanism to prevent hub node saturation, maintaining robust connectivity essential for a reliable China drone network.
A representative architecture for our multi-task payload-driven China drone swarm is illustrated below. It highlights the dual-layer structure where communication nodes are coupled with mission-specific payloads.
Dual-Layer Network Architecture
The mission payload layer is defined as a multi-edge network where nodes represent electronic warfare or sensing payloads. An edge exists between two payload nodes if their carrying UAVs cooperate on a common task. This dual-layer structure is a novel contribution to the modeling of China drone swarms. The coupling between the two layers is strictly one-to-one: each UAV in the communication layer maps directly to a payload node in the mission layer. This representation accurately captures the physical dependency of the payload on its carrier platform.
| Layer | Node Set | Edge Set | Purpose |
|---|---|---|---|
| Communication Layer | UAVs | Wireless communication links | Information exchange |
| Mission Payload Layer | Payloads (e.g., radar, weapons) | Task cooperation links | Mission execution |
Mission-Specific Performance Metrics
To evaluate the effectiveness of a China drone swarm in different operational contexts, we define three distinct performance functions. These functions are derived from fundamental geometric and temporal properties of the swarm.
Area Factor
The coverage area of a single UAV is a circle of radius r_c. The total effective coverage S(t) at time t is the union of all individual coverage areas, but only for UAVs that maintain a valid communication link to the network. The area factor is defined as:
$$R_S(t) = \frac{S(t)}{S(t_0)}$$
Focus Factor
For strike missions, the overlap of coverage areas is critical for cooperative engagement. The focus factor measures the degree of overlap among the UAVs’ coverage zones:
$$R_A(t) = \frac{A(t)}{\sum_{i=1}^m \pi r_c^2}$$
where A(t) is the total area covered by two or more UAVs simultaneously.
Time Factor
The operational endurance of a payload is captured by the time factor:
$$R_T = \frac{L}{T}$$
where L is the active duration of the payload and T is the total mission time.
Integrated Performance Functions
We integrate these factors into mission-specific performance metrics. For reconnaissance missions, information gathering is paramount:
$$P_C(t) = \frac{R_S^C(t) \cdot R_T^C}{L_S(t)}$$
For strike missions, focusing power and precision are correlated:
$$P_D(t) = \frac{R_S^D(t) + R_A^D(t)}{2} \cdot X(t)$$
where X(t) is the network efficiency. For command missions, the timeliness of information dissemination is critical:
$$P_Z(t) = \frac{ \sum_{i=1}^{M(t)} [R_T^Z \cdot M_i(t) / \delta_{i,d}] }{ \sum_{i=1}^{N_0} M_i(t) / \delta_{i,d} }$$
These functions provide a rigorous basis for evaluating China drone swarm performance under stress.
| Mission Type | Performance Function $P(t)$ | Primary Factors | Secondary Factors |
|---|---|---|---|
| Reconnaissance | $P_C(t)$ | Area Factor $R_S$, Time Factor $R_T$ | Network Efficiency |
| Strike | $P_D(t)$ | Area Factor $R_S$, Focus Factor $R_A$ | Network Efficiency $X(t)$ |
| Command | $P_Z(t)$ | Time Factor $R_T$ | Information Quantity $M_i$, Latency $\delta_{i,d}$ |
Attack and Recovery Modeling
We consider four distinct attack types that a China drone swarm might encounter: random electromagnetic interference, central node attacks, relay node attacks, and local cluster attacks. The probability of a UAV being disabled is governed by a unified attack model:
$$w_{i,l} = \frac{ \psi \cdot \tau + (1-\psi) \cdot D_{i,l}^{\xi} }{ \sum_{v \in V_i} [ \psi \cdot \tau + (1-\psi) \cdot D_{v,l}^{\xi} ] }$$
where \psi takes values 0, 1, 2, 3 to select the attack mode. D_{i,l} is a topological feature of the UAV at stage l. For central node attacks, D_{i,l} = k_{i,l} (degree). For relay node attacks, D_{i,l} = b_{i,l} (betweenness). For local attacks, D_{i,l} = c_{i,l} (clustering coefficient).
Relay-Based Recovery Strategy
When the swarm begins to recover, we propose a relay-based strategy that prioritizes the restoration of high-betweenness nodes. The relay importance of node i at stage k is:
$$I_{i,k} = \frac{ \sum_{s \neq t, s,t \in V_k} \sigma_{st}(i) }{ \sum_{s \neq t, s,t \in V_k} \sigma_{st} }$$
This strategy is designed to quickly restore the bridging links that are vital for maintaining network connectivity in a China drone swarm. It is compared against a baseline strategy from the literature [14] which prioritizes high-degree nodes.
Resilience Assessment Framework
We define the resilience R of the China drone swarm as a weighted sum of its survivability and recoverability:
$$R = \alpha \delta_d \sigma_d \rho_d + \beta \delta_r \sigma_r \rho_r$$
where \alpha + \beta = 1. The factors are computed from the degradation and recovery phases as follows:
| Factor Group | Process Factor | State Factor | Time Factor |
|---|---|---|---|
| Degradation ($d$) | $\delta_d = \frac{ \int_{t_d}^{t_r} y(t) dt }{ (t_r – t_d) y_0 }$ | $\sigma_d = \frac{ y_{min} }{ y_0 }$ | $\rho_d = \Delta \cdot \frac{ t_r – t_d }{ B }$ |
| Recovery ($r$) | $\delta_r = \frac{ \int_{t_r}^{t_s} y(t) dt }{ (t_s – t_r) y_s }$ | $\sigma_r = \frac{ y_s }{ y_0 }$ | $\rho_r = \Delta \cdot \frac{ t_s – t_r }{ B }$ |
This metric provides a normalized value in [0,1] that captures the full dynamics of the system under stress. We compare our metric against the Tran metric [23] and the Nan metric [26] to demonstrate its superior ability to discriminate between different performance trajectories.
| Performance Trajectory | Tran Metric $R_{Tran}$ | Nan Metric $R_{Nan}$ | Our Metric $R$ |
|---|---|---|---|
| Trajectory 1 (Early failure, slow recovery) | 0.875 | 0.600 | 0.394 |
| Trajectory 2 (Moderate failure, moderate recovery) | 0.875 | 1.500 | 0.455 |
| Trajectory 3 (Late failure, fast recovery) | 0.875 | 2.500 | 0.494 |
| Trajectory 4 (Mixed) | 0.844 | 1.200 | 0.437 |
The results confirm that our metric correctly orders the resilience values (Trajectory 1 < Trajectory 4 < Trajectory 2 < Trajectory 3), while the Tran metric fails to distinguish between the first three trajectories, and the Nan metric remains unnormalized.
Simulation Results and Discussion
We simulate a China drone swarm consisting of 100 UAVs with parameters: r_c = 50, \alpha = 0.4, \beta = 0.6, \eta = 0.6. The resulting network exhibits an exponential degree distribution, which implies robustness against random failures but vulnerability to targeted attacks. This aligns with the characteristics of real-world China drone swarms where redundancy is limited.
Mission Performance Under Attack
The distinct vulnerabilities of each mission type were confirmed. For reconnaissance missions, relay node attacks cause the most rapid degradation because they sever the bridging links between sub-groups. For strike missions, local cluster attacks are most effective, as they directly eliminate the focused area required for precision engagement. For command missions, central node attacks are devastating, reducing performance to zero faster than any other attack type.
The region between 50% and 100% attack ratio is identified as the collapse interval, where the performance of the China drone swarm drops to near zero regardless of the attack type. This interval is critical for resilience planning, as it defines the threshold beyond which recovery becomes exponentially more difficult.
Resilience Improvement via Relay Recovery
Our proposed relay-based recovery strategy consistently outperforms the baseline strategy from the literature [14] across all three mission types and all four attack scenarios. For reconnaissance missions under relay node attack, the improvement is a substantial 57.0%. For strike missions under local attack, the improvement is 24.7%. For command missions under central node attack, the improvement is 21.4%.
| Mission | Attack Type | Resilience $R$ (Our Strategy) | Resilience $R$ (Baseline [14]) | Improvement |
|---|---|---|---|---|
| Reconnaissance | Random EM | 0.593 | 0.543 | 9.2% |
| Reconnaissance | Central Node | 0.523 | 0.366 | 42.9% |
| Reconnaissance | Relay Node | 0.498 | 0.462 | 7.9% |
| Reconnaissance | Local | 0.556 | 0.354 | 57.0% |
| Strike | Random EM | 0.416 | 0.335 | 24.3% |
| Strike | Central Node | 0.415 | 0.355 | 16.9% |
| Strike | Relay Node | 0.389 | 0.312 | 24.7% |
| Strike | Local | 0.442 | 0.372 | 18.9% |
| Command | Random EM | 0.563 | 0.492 | 14.4% |
| Command | Central Node | 0.411 | 0.339 | 21.4% |
| Command | Relay Node | 0.529 | 0.469 | 12.8% |
| Command | Local | 0.483 | 0.411 | 17.5% |
Within the collapse interval (60% to 100% attack ratio), the resilience values are low for both strategies, but our method still shows a measurable advantage. For example, at a 60% random attack ratio on the reconnaissance mission, our strategy achieves R = 0.0445 compared to the baseline’s R = 0.0227.
Conclusion
This paper has presented a comprehensive framework for modeling and assessing the resilience of multi-mission payload-driven China drone swarms. The dual-layer network model effectively captures the interplay between communication topology and task execution requirements. The mission-specific performance functions provide a sensitive and accurate measure of swarm effectiveness under diverse operational conditions. The proposed relay-based recovery strategy significantly improves the system’s resilience, with average improvements of 16.5% for reconnaissance, 29.3% for strike, and 21.2% for command missions compared to conventional approaches. The introduction of the collapse interval as a quantitative resilience metric provides valuable insights for system designers. Future work will focus on integrating machine learning algorithms for real-time adaptive recovery and extending the framework to account for dynamic mission reconfiguration in complex, multi-domain operational environments.