In the research and development of unmanned aerial vehicles (UAVs), the traditional document-centric design methodology, which relied on physical prototyping and iterative test campaigns, has shown significant limitations in supporting the design and performance evaluation of UAV systems tailored to specific application scenarios. During my research, I identified two critical deficiencies in this conventional workflow: the lack of a unified model to support collaborative cyber-physical domain modeling, which prevented the uniform expression of task logic and dynamic responses; and the absence of a dynamic characterization mechanism for cyber-physical domain models, which constrained the overall improvement of power coupling reliability and delayed the optimization of development cycles. To overcome these challenges, I proposed an integrated design method for the whole process of UAV inspection, grounded in the theory of Model-Based Systems Engineering (MBSE). By leveraging structured modeling to cover the entire workflow from scheme design to physical implementation, my method establishes a full lifecycle collaborative optimization mechanism that integrates system design verification and performance optimization.
In this paper, I take the pipeline inspection mission as a typical application case to demonstrate the effectiveness and feasibility of my proposed approach. The method facilitates the in-depth integration of cyber and physical domains, providing a technical route for the performance evaluation of UAV systems in specific scenarios. Through the implementation of the MBSE methodology, I successfully constructed an information-domain architecture model of the UAV inspection process, achieving the coupled mapping mechanism between the behavioral logic and the physical structure of the system. Furthermore, I developed a multi-physics simulation model on the MWORKS platform to realize the dynamic verification and closed-loop mapping of the inspection UAV from the cyber domain to the physical domain. My research is systematically unfolded from the information domain architecture design, to the construction of the physical domain model, and finally to the integrated design and simulation of the entire inspection process.
1. Introduction
The proliferation of new-generation information technologies has profoundly transformed the design and development paradigm of unmanned aerial vehicles, establishing them as critical instruments for supporting the digital transformation of the low-altitude economy. As a common low-altitude operation platform, the UAV possesses significant advantages in structure compactness and operational flexibility. However, the development of UAV systems involves complex couplings across multiple disciplines such as mechanical engineering, control technology, and materials science. These characteristics make the establishment of a unified system-level model essential for analyzing and optimizing the overall performance of the aircraft.
The current UAV research and development process is predominantly document-based, with the division of tasks from the overall system to subsystems relying on textual documentation. This approach often leads to design inconsistencies between the system-level objectives and the underlying physical components, severely compromising the traceability of parameter coordination and matching. The consequence is often an increase in the number of iteration cycles during system integration and performance evaluation, which significantly reduces overall development efficiency. Consequently, there is an urgent need to introduce a formalized, traceable modeling and analysis method that can address the entire system, bridging the gap between the information domain (task logic) and the physical domain (dynamic response).
In my research, I employed MBSE, a trans-disciplinary technical approach based on systematic thinking, which provides a robust framework for managing the full lifecycle of system design through formalized models. MBSE’s exceptional capability in documentation management allows for the early identification of potential defects during the system development phase, thus reducing physical prototype iterations. Complementing this, multi-domain unified modeling, characterized by its high modularity and open architecture, enables the seamless integration of mechanical, electrical, control, and environmental dynamics within a unified framework. This approach complements MBSE’s static modeling capabilities and supports the dynamic response analysis and performance optimization of UAVs in complex flight environments.
Therefore, my research focuses on proposing an MBSE-driven integrated design method specifically for the full process of UAV inspection, integrating cyber-physical domain convergence. The main contributions of my work are structured into three primary research areas, which are elaborated in the subsequent sections of this paper.
2. Theoretical Foundations and Modeling Methodology
To address the challenges posed by the design of complex UAV systems, I have established a comprehensive theoretical foundation covering both the system engineering methodologies and the multi-physics modeling principles. This foundation enables me to leverage the strengths of both static architecture modeling and dynamic simulation to achieve a truly holistic design approach.
2.1 MBSE Methodologies and Tools
MBSE is a formalized engineering approach that uses models as the primary means of information exchange, rather than traditional documents. In my research, I systematically evaluated several mainstream MBSE methodologies, which are summarized in Table 1, to select the most suitable approach for the UAV system design process.
| Methodology | Key Characteristics | Core Advantages |
|---|---|---|
| Harmony | A V-model-based process emphasizing the identification of functions from behaviors and their mapping to physical architecture | Provides a clear and structured workflow for the entire system development lifecycle |
| OOSEM | An object-oriented approach integrating top-down functional decomposition | Supports parallelization and efficient integration across hardware and software development |
| OPM | A methodology using a unified formalism for system structure, behavior, and state evolution | Enhances model consistency through a single, unifying language |
After comparative analysis, I selected the Harmony methodology for structuring my research. Its emphasis on a behavior-driven design approach aligns well with the functional decomposition requirements of a UAV inspection mission. To implement this methodology, I utilized MWORKS.Sysbuilder, a modeling tool developed by Suzhou Tongyuan Soft Control Technology Co., Ltd. This platform provides robust support for the Harmony process and is compatible with the Systems Modeling Language (SysML), enabling a visual and formalized representation of the system’s information domain architecture.
2.2 Multi-Physical Domain Unified Modeling
For the physical domain modeling, I adopted the Modelica language. Modelica is an open-source, object-oriented language designed for easy modeling of complex physical systems. Its non-causal modeling approach, based on differential algebraic equations, allows for the simulation of multi-domain systems in a unified environment. The key characteristics of Modelica, as utilized in my research, are:
Object-Oriented Modeling: Modelica supports hierarchical model construction, component connection, and inheritance mechanisms, facilitating model reuse and maintenance. Multi-Domain Unified Modeling: It enables the modeling and analysis of multi-domain systems (e.g., mechanical, electrical, control) within a unified framework, overcoming the decoupling difficulties encountered in co-simulation approaches. Non-Causal Modeling: The language describes component behaviors through equations, which can straightforwardly represent the physical topology and interactions within the system.
In the Modelica context, component interactions are described using connectors. These connectors define two main types of variables: potential variables and flow variables, which are used to transmit power and information across different physical domains. This mechanism allows for a truly integrated and dynamic simulation of the UAV system.
3. Information Domain Architecture Modeling of UAV Inspection Process
Adhering to the principles of MBSE and guided by the Harmony methodology, my first stage of research was dedicated to the systematic analysis and modeling of the information domain of the UAV inspection process. This involved decomposing the mission requirements, analyzing system behaviors, and defining the system architecture in a formal and traceable manner.
3.1 Mission Analysis and Requirement Modeling
The pipeline inspection task is defined by a specific mission plan, which outlines the entire flight operation from take-off to return. The mission begins with a pre-flight self-check and communication link establishment with the ground terminal. After receiving the mission parameters, the UAV takes off and navigates to the designated inspection starting point. It then follows the pipeline trajectory to conduct the inspection, performing specific actions such as a detailed inspection at critical waypoints, before returning to its origin. This process is broken down into several high-level requirements, which are further refined into a detailed two-level requirement hierarchy, as shown in Table 2.
| ID | Requirement Name | Requirement Level |
|---|---|---|
| 0.1 | The UAV should have operational capabilities | Level 1 |
| 0.2 | The UAV should have cruise capabilities | Level 1 |
| 0.3 | The UAV should have emergency handling capabilities | Level 1 |
| 1.1 – 1.6 | Flight control, take-off, return-to-home, hover, and emergency response | Level 2 |
| 2.1 – 2.6 | Relative positioning, command, detection, path planning, pipeline tracking, and terrain adaptation | Level 2 |
| 3.1 – 3.8 | Handling of abnormal attitudes, cruise anomalies, altitude errors, low energy, communication loss, etc. | Level 2 |
To capture these requirements, I constructed a SysML requirement diagram. This diagram not only organizes the requirements in a hierarchical structure but also establishes traceability links, ensuring that each system function can be traced back to a specific user need, which is essential for managing the complexity of the system.
3.2 Use Case and Activity Analysis
Based on the identified requirements, I modeled the system functionalities using SysML use case diagrams. The use case diagram provides a clear representation of the interactions between the UAV system and its actors, while also delineating the system boundary. In my model, the top-level use case of “UAV Pipeline Inspection” is decomposed into a set of lower-level use cases, such as “Execute Take-off”, “Perform Autonomous Cruise”, “Conduct Inspection”, and “Execute Return”, which are connected using 《include》relationships. This approach organizes the system functions into a coherent and hierarchical structure.
The internal logic and control flow of the system were then detailed using SysML activity diagrams. These diagrams illustrate the sequence of actions, decision nodes, and parallel behaviors. For instance, the flight control activity diagram starts with a mission initialization node. The system then checks conditions, proceeds through take-off, climbs to the target altitude, follows the planned cruise path, performs the inspection, and finally decides whether to return or continue. This activity diagram provides a precise specification of the functional behavior that must be realized by the physical components.
3.3 System Architecture and Parameter Modeling
Following the behavioral analysis, I conducted a structural analysis to define the high-level system architecture and its components. Using SysML block definition diagrams and internal block diagrams, I decomposed the UAV system into key subsystems: the airframe, flight control system, sensor system (including GPS, IMU), motor, rotors, and environmental perception modules. These diagrams clearly depict the composition relationships and interaction interfaces among the components, providing a static blueprint of the physical system.
In the final step of the information-domain architecture phase, I established a parameter model to define and validate the system’s key parameters, such as the total mass. As shown in the parameter model, the total mass constraint is defined as the sum of the masses of the airframe, rotors, and motors. This constraint is then used in a validation check against a predefined target value, ensuring that the design meets the specified criteria. This formal integration of parameters into the system model closes the loop between the logical architecture and the quantitative design, laying a baseline for the subsequent physical domain modeling.
4. Multi-Physical Domain Dynamic Response Analysis
In the second phase of my research, I transitioned from the static architecture model to the dynamic physical domain. The goal was to construct a high-fidelity simulation model of the UAV’s dynamics to analyze its response under various environmental conditions, with a specific focus on wind disturbances. This model serves as a crucial “plant” model for testing the control system and validating the integrated design.
4.1 Coordinate System and Kinematics
The modeling began with the definition of coordinate systems. I use an inertial frame to describe the UAV’s position and velocity, and a body-fixed frame to describe its orientation and the forces acting upon it. The attitude of the UAV is described by the Euler angles Φ=[φ, θ, ψ]^T, representing the roll, pitch, and yaw angles. The rotation matrix from the body frame to the inertial frame, denoted as $R_{B}^{E}$, is a function of these angles and is essential for transforming forces and velocities between the two frames. The transformation matrix is given by:
$$R_{B}^{E} = \begin{bmatrix} \cos\theta\cos\psi & \sin\phi\sin\theta\cos\psi – \cos\phi\sin\psi & \cos\phi\sin\theta\cos\psi + \sin\phi\sin\psi \\ \cos\theta\sin\psi & \sin\phi\sin\theta\sin\psi + \cos\phi\cos\psi & \cos\phi\sin\theta\sin\psi – \sin\phi\cos\psi \\ -\sin\theta & \sin\phi\cos\theta & \cos\phi\cos\theta \end{bmatrix}$$
4.2 UAV Dynamic Modeling
To establish the UAV dynamics model, I made several simplifying assumptions: treating the UAV as a rigid body with symmetrical mass distribution, neglecting the flexibility of the propellers, and approximating the aerodynamic forces as being directly proportional to the square of the rotor speed. The translational dynamics are described by Newton’s second law in the inertial frame, which can be expressed as:
$$\begin{bmatrix}\ddot{x} \\ \ddot{y} \\ \ddot{z}\end{bmatrix} = \frac{1}{m} \begin{bmatrix} F_B(\cos\phi\sin\theta\cos\psi + \sin\phi\sin\psi) \\ F_B(\cos\phi\sin\theta\sin\psi – \sin\phi\cos\psi) \\ F_B(\cos\phi\cos\theta) \end{bmatrix} – \begin{bmatrix} g \\ 0 \\ 0 \end{bmatrix} – \frac{1}{m}\begin{bmatrix} K_x\dot{x} \\ K_y\dot{y} \\ K_z\dot{z} \end{bmatrix}$$
In this formulation, m represents the total mass of the unmanned aerial vehicles, g is the gravitational acceleration, and $K_x$, $K_y$, $K_z$ are the air resistance coefficients along the respective axes. The total lift force, $F_B$, is the sum of the forces generated by the four rotors:
$$F_B = \sum_{i=1}^{4} F_i = C_L \sum_{i=1}^{4} N_i^2$$
where $C_L$ is the lift coefficient and $N_i$ is the speed of the i-th rotor. The rotational dynamics, which describe the angular motion of the UAV, are derived from Euler’s equation and are presented here in a simplified form:
$$\begin{bmatrix}\ddot{\phi} \\ \ddot{\theta} \\ \ddot{\psi}\end{bmatrix} = \begin{bmatrix} \frac{(I_y – I_z)}{I_x}\dot{\theta}\dot{\psi} + \frac{L}{I_x}(F_4 – F_2) – \frac{J_r}{I_x}\dot{\theta}N \\ \frac{(I_z – I_x)}{I_y}\dot{\phi}\dot{\psi} + \frac{L}{I_y}(F_3 – F_1) – \frac{J_r}{I_y}\dot{\phi}N \\ \frac{(I_x – I_y)}{I_z}\dot{\phi}\dot{\theta} + \frac{1}{I_z}(M_1 – M_2 + M_3 – M_4) \end{bmatrix}$$
Here, L is the distance from the rotor center to the UAV’s center of mass, $I_x$, $I_y$, $I_z$ are the moments of inertia about the respective axes, $J_r$ is the rotor inertia, and $M_i$ represents the moment generated by the i-th rotor. These equations define the core physical domain model of the UAV platform.
4.3 Wind Disturbance and Rotor Aerodynamics Modeling
To evaluate the UAV system’s performance in real-world conditions, I modeled the wind environment by combining a steady mean wind component with a stochastic turbulence component, based on the Dryden spectral model. The Dryden model defines the turbulence velocity spectral functions, and the transfer functions for the turbulence filters are presented in Table 3.
| Turbulence Component | Transfer Function |
|---|---|
| Along X-axis, $H_u(s)$ | $\sigma_u \sqrt{\frac{L_u}{\pi}} \frac{1}{1 + \frac{L_u}{V}s}$ |
| Along Y-axis, $H_v(s)$ | $\sigma_v \sqrt{\frac{L_v}{\pi}} \frac{1 + \frac{1}{\sqrt{3}}\frac{L_v}{V}s}{(1 + \frac{L_v}{V}s)^2}$ |
| Along Z-axis, $H_w(s)$ | $\sigma_w \sqrt{\frac{L_w}{\pi}} \frac{1 + \frac{1}{\sqrt{3}}\frac{L_w}{V}s}{(1 + \frac{L_w}{V}s)^2}$ |
The effect of the wind on the rotors is a complex aerodynamic interaction. When a rotor operates in a wind field, it experiences an additional induced velocity, $V_d$, which affects the total aerodynamic force and torque. The total thrust of a rotor, $F_{ri}$, is then not only a function of its rotational speed but also of the airflow velocity. This is crucial for accurately predicting the UAV’s behavior under wind perturbations.
4.4 Control System Design
To ensure stable flight and accurate path tracking, I implemented a hierarchical control strategy. For the outer-loop guidance, I used the Line-of-Sight (LOS) guidance law, which calculates a desired heading angle, $\varphi_d$, to steer the UAV towards the desired path, minimizing the cross-track error, $e_y$:
$$\varphi_d = \varphi_p – \arctan\left(\frac{e_y}{\Delta}\right)$$
The inner-loop attitude and altitude control were implemented using a cascade of PID controllers. These controllers receive the desired attitude or altitude commands and generate control signals (e.g., changes in motor speed) to minimize the error. The combination of LOS for path guidance and PID for attitude stabilization provides a robust and effective control structure for the UAV system.
The resulting physical domain model, built using the Modelica language on the MWORKS platform, integrates the coordinate transformations, UAV dynamics, wind environment system, and control system, as shown in Figure. This figure presents the complete simulation model, and its parameters are detailed in Table 4.
| Parameter Name | Symbol | Value | Unit |
|---|---|---|---|
| Mass | m | 1.25 | kg |
| Arm Length | l | 0.38 | m |
| Rotor Radius | R | 0.16 | m |
| Rotor Inertia | Jr | 1 | kg·m² |
| X-axis Moment of Inertia | Ix | 0.096 | kg·m² |
| Y-axis Moment of Inertia | Iy | 0.096 | kg·m² |
| Z-axis Moment of Inertia | Iz | 0.016 | kg·m² |
4.5 Simulation Results and Dynamic Response Analysis
After developing the model, I conducted a series of simulation experiments to analyze the UAV’s dynamic response. First, I simulated the flight under ideal, no-wind conditions to establish a baseline performance. The results for the attitude and speed components are summarized below.
No-Wind Disturbance: In the absence of wind, the UAV’s attitude angles exhibited a small transient fluctuation during take-off (0-10s), but quickly converged to a stable state with a steady-state error near zero. The yaw angle also stabilized with a minor initial offset. This shows the control system can effectively stabilize the UAV in ideal conditions. The velocity components showed smooth convergence to the desired cruise speed, with minimal coupling between the axes. The UAV dynamics model demonstrated good stability and controllability in the ideal environment.
With Wind Disturbance: When the wind disturbance model was enabled, the UAV’s attitude and velocity responses changed significantly, and the performance metrics in both conditions are compared in Table 5.
| Performance Metric | No-Wind Condition | Wind Disturbance Condition |
|---|---|---|
| Roll/Pitch Angle Fluctuation | ±0.02 rad | ±0.3 rad |
| Settling Time | 5-8 s | 10-15 s |
| X-axis Velocity Peak | 5 m/s | 5.5 m/s |
| Speed Component Coupling | Minimal | Significant |
| Overall Stability | Excellent | Good (bounded response) |
The simulation results clearly demonstrate that the multi-physics model can effectively predict the motion response of the unmanned aerial vehicles under varying conditions. Although wind disturbance causes larger fluctuations in both attitude and velocity, the designed control system successfully suppresses these disturbances, maintaining bounded responses with zero steady-state error. This confirms the robustness of the physical domain model and its control system for the inspection task.
5. Integrated Design and Simulation of the Full Inspection Process
The final phase of my research focused on integrating the information-domain architecture model with the physical-domain simulation model to create an integrated design and simulation environment. This represents the realization of the cyber-physical domain convergence, enabling a closed-loop verification of the entire inspection process. The implementation of this integrated framework is a key contribution of my work, as it connects system-level logic directly to the physical behavior.
5.1 Task Analysis and Scenario Design
The inspection task was modeled as a hybrid system, combining continuous state evolution with discrete task-triggering events. In this formulation, the UAV’s state vector, Z(t), includes its position, velocity, and attitude, and the continuous dynamics are governed by the previously developed physical domain model. The discrete task events, such as “reaching a waypoint” or “completing a photo shot”, are triggered by the UAV’s state. To model these events, I introduced an event indicator function, $\eta(t)$:
$$\eta(t) = \sum_{i=1}^{N} \delta(X(t) – P_i)$$
where $P_i$ represents the position of the i-th critical inspection target. This hybrid model allows for the integrated simulation of the continuous flight dynamics and the discrete task logic within a unified framework. To accurately simulate a realistic mission, I also designed a complex 3D terrain model, which describes the surface of a mountain environment. The terrain is constructed by superimposing smooth large-scale undulations, sharp discontinuities (like cliffs), and linear features (like ridges). This provides a challenging and realistic flight scenario for evaluating the UAV system’s performance.
5.2 Integrated Simulation Verification on the MWORKS Platform
The central element of this phase is the joint simulation on the MWORKS platform, where the information-domain and physical-domain models interact dynamically, as illustrated in the accompanying figure. In this joint simulation architecture, the Modelica physical-domain model continuously provides flight state information (such as position, velocity, and altitude) to the SysML activity diagram. The information-domain model, based on this real-time state information, determines the current task logic and makes high-level decisions, such as initiating take-off, switching to cruise mode, or issuing a command to stop and hover for a detailed inspection. These commands are then sent back to the physical domain model, which adjusts its control targets to execute the new mission phase.
This closed-loop interaction between the decision-making logic and the physical dynamics is the essence of my proposed cyber-physical model. It ensures that the system architecture and control logic are not designed in a vacuum, but are continuously validated against the underlying physical system’s capabilities and responses. This approach provides an environment for testing and optimizing the integrated UAV system design.
5.3 Results and Analysis of the Full-Process Simulation
The integrated simulation was conducted to verify the full inspection process. The results were analyzed by examining the UAV’s flight trajectory and control performance.
Flight Trajectory Analysis: Analyzing the UAV’s trajectory in different planes revealed important insights. The xoy plane trajectory shows that the UAV successfully tracked the pipeline route, maintaining a relatively consistent horizontal offset after an initial adjustment phase. The xoz plane trajectory shows the UAV’s ability to maintain a stable flight height of around 2.25 m during the main inspection segment, and then adjust to 3.25 m during a transition phase. The three-dimensional trajectory, presented in the table below, integrates these elements, showing a smooth, continuous flight path with a clear transition between different mission phases, confirming the effective coordination between the continuous flight dynamics and the discrete task logic.
| Trajectory Segment | Behaviors and Features |
|---|---|
| Initial Flight Phase | Minor lateral oscillations, altitude adjustment, and transition to cruise state |
| Cruise Inspection Phase | Stable horizontal tracking, consistent flight altitude of 2.25 m, uniform velocity |
| Transition Phase | Altitude change from 2.25 m to 3.25 m, smooth trajectory change, no abrupt behavior |
| Return Phase | Stable flight back to the starting point, consistent with the planned route |
Control Response Analysis: The control system’s performance was evaluated by analyzing the altitude and attitude control responses. The altitude response curve shows a smooth convergence to the desired setpoint of 2.25 m. During the mission transition, the UAV executes a controlled altitude change, initially decreasing and then increasing to the new setpoint of 3.25 m, with a brief, damped oscillation that is characteristic of a well-tuned PID controller. The attitude response, specifically the pitch angle, shows a small transient during take-off but is quickly damped. Similarly, during the mid-mission transition, the attitude changes are well-controlled and the system stabilizes quickly. The results demonstrate the robustness and precision of the proposed control strategy in managing the UAV’s altitude and attitude throughout the entire flight envelope.
These results validate my proposed integrated design method for the full process of UAV inspection. The method successfully achieves the dynamic verification and closed-loop mapping of the information and physical domains, providing a robust and effective pathway for UAV system design and performance evaluation in specific application scenarios.
6. Conclusion and Future Outlook
In this paper, I have proposed and validated an MBSE-driven integrated design method for the full process of UAV inspection, focusing on achieving cyber-physical domain convergence. My research addresses the critical limitations of traditional UAV design processes by establishing a unified modeling framework that spans the entire system lifecycle. The main conclusions of my work are summarized as follows:
First, I successfully developed an information-domain architecture model for the UAV inspection process based on the MBSE methodology. This model effectively characterizes the coupling mapping mechanism between the behavioral logic and physical structure of the UAV system, providing a formalized and traceable model foundation for the subsequent system analysis and dynamic optimization.
Second, I constructed a high-fidelity multi-physics simulation model on the MWORKS platform. The model incorporates UAV dynamics, environmental disturbances, and control systems, and can predict the motion response of the UAV under varying environmental conditions. Simulation results confirmed the model’s ability to effectively simulate the UAV’s behavior, especially in the presence of wind disturbances, demonstrating its relevance for engineering applications.
Finally, I established a closed-loop integrated design and simulation environment. By combining the information-domain architecture model with the physical-domain model, I was able to perform a full-process simulation of the inspection task. This integration allowed for the dynamic verification of system behavior, from high-level mission decisions down to the physical control responses, achieving the envisioned cyber-physical domain convergence.
The principal innovation of my research is the creation of a unified modeling and simulation platform that supports the multi-domain coupling and closed-loop verification of UAV systems. This unified platform can effectively support the design, development, and performance evaluation of UAV systems, especially for complex tasks in specific scenarios, while reducing the iteration cycle and enhancing the quality of the design.
In future work, I aim to extend this research in several directions. The current physical model can be enhanced by incorporating more complex environmental factors and higher-fidelity rotor aerodynamics. Furthermore, I plan to integrate intelligent optimization algorithms into the framework, enabling the simulation results to not only verify the design but also actively drive the automatic optimization of system parameters. Finally, I intend to explore the integration of this model-based design approach with hardware-in-the-loop and real-flight testing to further validate its applicability and robustness in practical engineering deployments.