The rapid advancement of low-altitude economies and defense applications has placed China drone swarm technology at the forefront of both civilian and military innovation. Unmanned aerial vehicle (UAV) swarms, composed of multiple collaboratively operating China drone units, exhibit complex spatiotemporal dynamics that demand precise modeling for simulation and deduction. Our research addresses the critical need for a dedicated spatiotemporal data model that can organize, describe, and dynamically represent the multifaceted behaviors of China drone swarms in virtual environments. This paper presents our comprehensive approach to spatiotemporal feature analysis and model construction, specifically tailored for China drone swarm simulation scenarios.
Introduction
China drone technology has evolved from individual platforms into sophisticated swarm systems capable of coordinated tasks such as surveillance, disaster response, and precision agriculture. The simulation and deduction of China drone swarms require robust data models that capture dynamic state changes, multi-granularity representations, and intricate correlation relationships. Traditional spatiotemporal data models, designed primarily for static or slowly changing geographic entities, prove inadequate for the high-frequency updates and complex interactions inherent in China drone swarms. We identify three core challenges: first, the need for sub-second state synchronization among China drone units; second, the explicit representation of dynamic internal and external associations; and third, the smooth transition between macro, meso, and micro views during simulation.
To overcome these challenges, we propose the Spatiotemporal Object Data Model for China Drone Swarms (STDM-UAVS). This model integrates six fundamental elements: attribute information, real-time status, behavioral events, correlation relationships, multi-granularity views, and spatiotemporal benchmarks. By establishing a formal mapping to spatiotemporal hybrid Petri nets, our model provides a mathematically grounded framework for modeling discrete events and continuous processes that characterize China drone swarm operations. Furthermore, we develop a dynamic expression mechanism using CZML to visualize the swarm’s evolution in digital space.
Spatiotemporal Feature Analysis of China Drone Swarms
Our analysis begins with individual China drone features, which encompass morphological structure, behavioral capabilities, granularity expression, spatiotemporal changes, and environmental adaptability. Table 1 summarizes the key attributes extracted for individual China drone units.
| Feature Category | Description | Example Attributes |
|---|---|---|
| Morphological Structure | Physical geometry and dimensional properties | Body type, material, weight, rotor configuration |
| Behavioral Capabilities | Action repertoire and performance limits | Max speed, payload capacity, communication range |
| Granularity Expression | Representation detail across scales | Point vs. volume representation in simulation |
| Spatiotemporal Changes | Dynamic variations in position and state | Position, velocity, energy consumption over time |
| Environmental Adaptability | Suitability under diverse conditions | Wind resistance, temperature tolerance, EMI protection |
Extending these individual features to China drone swarms, we define a comprehensive spatiotemporal description framework comprising eight aspects: basic attributes, morphological structure, behavioral actions, orientation and position, internal and external relations, spatiotemporal benchmarks, resource management, and performance evaluation. This framework enables holistic characterization of China drone swarm dynamics across different operational contexts.
From this descriptive foundation, we extract four essential spatiotemporal features that uniquely characterize China drone swarms:
1. Spatial Morphological Features. These include geometric measures such as centroid position and average inter-UAV distance, as well as higher-level pattern recognition for formation types. For a homogeneous China drone swarm with n units, the centroid C and average distance d are computed as:
$$C = \frac{1}{n} \sum_{i=1}^{n} p_i$$
$$d = \frac{2}{n(n-1)} \sum_{i<j} -=""
For heterogeneous China drone swarms, we extend the centroid calculation to a weighted form: C = Σ w_i p_i, where w_i accounts for payload capacity, mission priority, or communication capability.
2. Structural Topological Features. These capture composition, spatial connectivity, and communication architecture within China drone swarms. Composition features dynamically track sub-swarm formation, role assignment, and unit addition or removal. Spatial topology uses Delaunay triangulation or k-nearest neighbor graphs to define proximity relationships. Communication topology, modeled as directed weighted graphs from protocol logs, represents actual link quality and latency.
3. Behavioral Capability Features. This feature category describes the collective actions and operational capacity of China drone swarms. Behavioral patterns include normal maneuvering, cooperative sensing, and anomaly detection, derived from sequences of position and velocity states. Capability features quantify performance limits such as maximum coverage area, mission completion time, and total endurance.
4. Granularity Expression Features. China drone swarms must be represented at macro, meso, and micro scales depending on the analysis objective. At the macro scale, the entire swarm is abstracted as a single point with aggregated properties such as centroid and mean velocity. The meso scale reveals sub-swarm structures and inter-group relations, while the micro scale presents individual China drone details including 3D models, real-time telemetry, and fine-grained behavioral events.
Construction of the Spatiotemporal Data Model
Drawing on object-oriented principles and spatiotemporal cognition theory, we define the STDM-UAVS for China drone swarms as a six-element system. The formal definition is:
$$UAVS = \{ SID, PRO, STA, ACT, REL, MGV, STB \}$$
Where SID is the unique identifier, and the remaining elements correspond to attribute information, real-time status, behavioral events, correlation relationships, multi-granularity views, and spatiotemporal benchmarks respectively. Table 2 details each element and its role within the model.
| Element | Symbol | Description | Formalization |
|---|---|---|---|
| Attribute Information | PRO | Static (e.g., ID, manufacturer) and dynamic (e.g., mission type, deployment) attributes | PRO = {StaticProperty, DynamicProperty} |
| Real-time Status | STA | Instantaneous state at time t: composition, spatial state, mission progress, failures | STA = {SPS, PES, ITime} |
| Behavioral Events | ACT | Actions (e.g., formation change) and events (e.g., task completion) with time intervals | ACT = {Behavior, Event, STime, ETime} |
| Correlation Relationships | REL | Internal (inter-UAV) and external (swarm-to-environment) relations | REL = {OuterRelation, InnerRelation} |
| Multi-granularity Views | MGV | Abstract levels: macro, meso, micro | MGV = {MacroView, MesoView, MicroView} |
| Spatiotemporal Benchmarks | STB | Global (WGS84/UTC) and local (relative to swarm center) references | STB = {GSTB, LSTB} |
The dynamic attribute information within PRO is represented as a time-indexed list:
$$DynamicProperty = \{ PName, ValueList \}, \quad ValueList = \{ PValue, STime, ETime \}$$
The real-time status element integrates three sub-components:
$$SPS = \{ Position, Velocity, Extent, Semantic \}$$
$$PES = \{ TaskState, FailureState \}$$
Correlation relationships are further decomposed into source and target identifiers, relationship type, and a relation list that captures temporal validity:
$$XRelation = \{ SourceID, TargetID, RType, RelationList \}$$
$$RelationList = \{ RValue, STime, ETime \}$$
Our model establishes a direct mapping to spatiotemporal hybrid Petri nets, as demonstrated in Table 3, enabling formal verification and precise simulation of China drone swarm behaviors using tools such as UPPAAL.
| STDM-UAVS Element | Petri Net Component | Mapping Rationale |
|---|---|---|
| PRO (Attribute Information) | Place set (P) | Static attributes→discrete places (Pd); dynamic→continuous places (Pc) |
| STA (Real-time Status) | Initial marking vector (M0) | M0 represents instantaneous state at t=0 |
| ACT (Behavioral Events) | Transition set (T) | Discrete events→Td; continuous processes→Tc; time intervals→I |
| REL (Correlation Rel.) | Flow relation set (F) | F defines directed arcs between places and transitions |
| MGV (Multi-granularity Views) | Place/transition set | Abstraction level of places/transitions matches view granularity |
| STB (Spatiotemporal Benchmarks) | Time function (I) and spatial info (A) | I provides time reference; A provides spatial reference |
Dynamic Expression Modeling for China Drone Swarms
To bridge the gap between the abstract STDM-UAVS and real-time digital visualization, we develop a dynamic expression mapping to CZML (Cesium Language), a JSON-based format optimized for time-dynamic geospatial data. This mapping translates each of the six STDM-UAVS elements into corresponding CZML nodes, enabling data-driven rendering without format conversion.
For instance, the attribute information (PRO) maps to CZML properties nodes, with dynamic properties updated via time-tagged arrays. Real-time status (STA) maps to position, path, and model nodes for spatial and visual representation. Behavioral events (ACT) drive animation sequences and state transitions, while correlation relationships (REL) are encoded as label and billboard overlays to denote links between China drone units.
The multi-granularity view (MGV) mechanism is critical for adaptive visualization. At the macro level, the entire China drone swarm is represented as a single point with aggregated labels showing centroid altitude, average speed, and mission phase. At the meso level, sub-swarm groupings are highlighted through distinct color coding and connectivity lines. The micro level renders individual China drone platforms with full 3D models, telemetry streams, and real-time behavior annotations. View switching is automatically triggered by camera altitude thresholds: above 3,000 m activates macro view; between 500–3,000 m triggers meso view; below 500 m enables micro view.
Our dynamic expression approach is validated through a Cesium-based simulation of a China drone swarm conducting a cooperative ground surveillance mission. The simulation involves 32 heterogeneous China drone units (16 DJI Inspire 3 and 16 DJI Mavic 3 platforms) operating over a 10-minute period, encompassing formation changes, coordinated movements, and task transitions.
The simulation demonstrates that our model effectively captures the full spectrum of China drone swarm dynamics. When the camera is zoomed out, the swarm appears as a single coherent point with aggregate status indicators. As the viewpoint approaches, individual China drone units become discernible, and their spatial relationships become measurable. At the closest zoom, each China drone’s three-dimensional model, trajectory path, and real-time telemetry are fully rendered. The transition between task phases—such as switching from rapid transit formation to area coverage formation—is seamlessly handled through the behavioral event engine, which updates positions, velocities, and visual representations in real-time.
The correlation relationships module proves particularly effective for displaying communication links and spatial constraints. During the simulation, inter-UAV distances are continuously computed and compared against collision avoidance thresholds. When a China drone unit deviates from its nominal path, the model generates an alert event that is propagated through the relationship network, allowing the surrounding units to adjust their trajectories autonomously. This demonstrates the model’s capacity for real-time, event-driven dynamic updates.
Model Comparison and Evaluation
To contextualize our contribution, we compare STDM-UAVS against existing models commonly used for China drone swarm simulation: simple trajectory-temporal models and spatiotemporal hybrid Petri nets. Table 4 summarizes the qualitative assessment across key dimensions.
| Dimension | STDM-UAVS (Proposed) | Spatiotemporal Hybrid Petri Net | Simple Trajectory-Temporal Model |
|---|---|---|---|
| Primary Objective | Data organization and visual expression; convertible to formal models | Formal verification of behavioral logic with mathematical rigor | Historical trajectory storage and playback |
| Discrete-Continuous Fusion | Integration via behavioral events and real-time status elements | Supported through discrete and continuous places | Not supported; records state snapshots |
| Correlation Expression | Explicit multi-dimensional (spatial, temporal, communication, semantic) via REL | Implicit via arcs; precise but high abstraction | Indirect; must infer from proximity |
| Granularity Support | Built-in macro, meso, micro views with automatic switching | Requires external conversion for multi-scale visualization | Single-scale trajectory plotting |
| Ease of Use for Simulation | Direct mapping to CZML for immediate visualization | Needs transformation for visual output; complex operations | Simple for playback; inadequate for dynamic interactions |
| Applicable Scenarios | Core data model for simulation platforms and situational awareness | High-reliability system design and algorithm verification | Post-mission analysis and basic monitoring |
The comparison highlights that STDM-UAVS uniquely addresses the synergistic requirements of China drone swarm simulation: it combines comprehensive data organization with formal model compatibility, explicit association management, built-in multi-granularity support, and direct visualization integration. Existing models either focus solely on formal logic (neglecting visualization ease) or simple state recording (neglecting dynamic interactions and multi-scale expression).
Our experimental results confirm that the STDM-UAVS can manage the spatiotemporal information of a 32-unit China drone swarm across all three granularity levels without data redundancy or format mismatch. The event-driven update mechanism maintains a state refresh rate of 10 Hz, sufficient for typical swarm simulation scenarios. The correlation relationship module successfully tracked an average of 496 inter-UAV spatial relations and dynamically adjusted them as the swarm reconfigures.
Conclusion and Future Directions
We have developed and validated the spatiotemporal object data model for China drone swarms (STDM-UAVS), which systematically addresses the limitations of traditional models in handling dynamic, multi-associative, and multi-granular characteristics. The model’s six-element architecture provides a comprehensive framework for organizing both static and dynamic information, while its mapping to Petri nets ensures formal verification capability. The dynamic expression mechanism bridges the gap between abstract data models and real-time visualization, demonstrating successful application in a Cesium-based simulation environment.
This work directly supports the advancement of China drone swarm technology by providing a standardized data foundation for collaborative mission planning, situational deduction, and effectiveness evaluation. The model’s explicit correlation relationships enable new analytical possibilities, such as emergent behavior detection and resilience assessment. Its multi-granularity views facilitate decision-making from strategic command (macro) to individual operator control (micro).
Looking forward, we plan to extend the model for large-scale heterogeneous China drone swarms with multiple payload types and autonomous decision capabilities. We will also explore integration with artificial intelligence frameworks, enabling the model to serve as the core data infrastructure for reinforcement learning-based swarm optimization and graph neural network-driven state prediction. By establishing a closed loop from sensing to modeling to decision-making, our research aims to accelerate the deployment of intelligent, resilient China drone swarms in real-world operations.