Case Studies of RflySim Ecosystem Publications¶
This page summarizes representative academic achievements and related preprints from the RflySim ecosystem, covering areas including unified simulation platform design, propulsion system modeling and optimization, hardware-in-the-loop testing and credibility assessment, UAV digital twins, swarm navigation, fault diagnosis, audio-visual perception, and visual interception. Preprints are explicitly identified in the “Journal/Conference” field.
2026¶
📄 Physics-Aware Multichannel Vector Quantization for Hybrid Digital Twin Modeling of UAV Systems¶
Authors: Tu, Jinhu; Nian, Xiaohong; Dai, Xunhua
Journal/Conference: IEEE Transactions on Aerospace and Electronic Systems
Link: 🔗 10.1109/TAES.2025.3639005 📥 View PDF 📚 BibTeX Citation
Keywords: Digital twins Vector quantization UAV physics-aware learning
📖 Expand to View Abstract
Big data technology has driven transformations in industrial manufacturing, and AI -driven digital twins have become key to the development of the low-altitude economy, with UAV systems serving as core carriers. However, heterogeneous sensor data, complex spatiotemporal distributions, and nonlinear interactions in system dynamics pose challenges to constructing reliable UAV digital twins. This paper proposes a UAV digital twin modeling and fusion framework that integrates physical priors with data-driven learning. A multichannel soft vector quantization mechanism is employed to achieve robust fusion of multi-source data, while physics-aware constraints and a hybrid fusion strategy are incorporated to improve model generalization. Validation on a large-scale dataset comprising simulation and real-flight data across multiple UAV flight missions demonstrates that the method outperforms traditional models in both fidelity and robustness. This study proposes a scalable UAV digital twin architecture with strong fusion capabilities and provides open-source tools to facilitate the development of self-evolving digital twin systems in dynamic environments.
📄 RflySimSaT: A Safety Assessment Platform for UAVs Based on Hardware-in-the-Loop Simulation¶
Authors: Dai, Xunhua; Tu, Jinhu; Chen, Yong; Quan, Quan
Journal/Conference: IEEE Transactions on Automation Science and Engineering
Link: 🔗 10.1109/TASE.2026.3651543 📥 View PDF 📚 BibTeX Citation
Keywords: Safety assessment HIL simulation UAV Testing
📖 Expand to View Abstract
UAVs are leading the development of future digital smart cities, but their safety issues have attracted widespread attention. Although existing robot simulators provide efficient and low-cost testing environments for UAVs, they lack comprehensive consideration of safety design and user testing requirements. To address this issue, this paper proposes RflySimSaT, a dedicated UAV safety testing platform. The platform covers safety factors throughout the full life cycle and integrates core modules such as multiple fault testing scenarios and high-fidelity dynamic models. Its modular architecture supports plug-and-play cross-platform closed-loop safety testing. Users need only provide the aircraft and flight controller to complete the full workflow, including development and deployment, using the platform. This paper validates the credibility and generality of the platform through multiple test cases and also provides supporting manuals, cases, and fault datasets. The platform is open-source, and its code repository has been published.
📄 High-Precision Modeling of UAV Electric Propulsion for Improving Endurance Estimation¶
Authors: Dai, Xunhua; Liu, Wei; Chen, Yong
Journal/Conference: Drones
Link: 🔗 10.3390/drones10020080 📥 View PDF 📚 BibTeX Citation
Keywords: Electric propulsion Propeller modeling Endurance estimation UAV
📖 Expand to View Abstract
This paper proposes an electric propulsion modeling framework for improving UAV endurance estimation accuracy. It combines dimensional analysis based on the Buckingham π theorem with data-driven parameter fitting to predict propeller thrust, power, and motor current under different inflow conditions using limited experimental data. The method is validated using the UIUC propeller database, a self-built inflow test bench, and real-flight tests. It achieves both prediction accuracy and computational efficiency across multiple propeller and motor configurations, and the model and implementation code are publicly available.
📄 Audio-Visual Fusion Sim2Real Platform for Anti-UAV Detection and Tracking¶
Authors: Nian, Xiaohong; Liu, Haolun; Dai, Xunhua
Journal/Conference: Drones
Link: 🔗 10.3390/drones10030190 📥 View PDF 📚 BibTeX Citation
Keywords: Audio-visual fusion Anti-UAV Detection and tracking Sim2Real
📖 Expand to View Abstract
This paper proposes an audio-visual fusion Sim2Real platform for anti-UAV detection and tracking. It combines time-difference-of-arrival (TDOA)-based acoustic source localization with deep-learning-based visual detection, achieving complementary perception through spatiotemporal synchronization, in which acoustics guide visual attention and vision corrects acoustic estimates. The platform supports algorithm validation under configurable occlusion conditions such as smoke and fog, and evaluates sim-to-real transfer performance through real quadcopter experiments, improving target tracking continuity in visually degraded scenarios.
📄 Skeletal Virtual Tube Generation for Urban UAV Swarm Navigation¶
Authors: Dai, Xunhua; Liao, Yuting
Journal/Conference: 2026 IEEE 20th International Conference on Control and Automation (ICCA)
Link: 🔗 10.1109/ICCA69928.2026.11618197 📥 View PDF (Author Manuscript) 📚 BibTeX Citation
Keywords: Virtual tube Generalized Voronoi Diagram Urban navigation UAV swarm
📖 Expand to View Abstract
This paper proposes a virtual tube generation framework based on generalized Voronoi diagram skeletons for UAV swarm navigation in complex urban airspace. The method extracts geometric skeletons of the environment at different altitude layers, constructs three-dimensional global routes through cross-layer topological associations, and generates continuous safe tube spaces. Experiments demonstrate that the method balances obstacle safety clearance and planning efficiency, providing path and spatial constraints for large-scale swarm navigation in urban environments.
📄 Safe Swarm Navigation in Constrained Environments: A Dynamic Tube-Based Distributed MPC Approach¶
Authors: Dai, Xunhua; Liao, Yuting; Chen, Yong
Journal/Conference: Drones
Link: 🔗 10.3390/drones10030177 📥 View PDF 📚 BibTeX Citation
Keywords: UAV swarm Dynamic tube Distributed MPC Collision avoidance
📖 Expand to View Abstract
This paper proposes a method combining dynamic virtual tubes and distributed model predictive control for safe UAV swarm navigation in constrained environments. By adjusting tube boundaries to accommodate dynamic obstacles, the method maintains connectivity of traversable space while using predicted risks to trigger constraint activation, thereby reducing unnecessary conservative constraints and computational overhead. Experiments validate the method's collision avoidance capability, space utilization, and scalability in high-density swarm scenarios.
📄 Digital Twin-Guided Spatiotemporal Graph Representation Learning for Reliable UAV Fault Diagnosis Under Complex Flight Conditions¶
Authors: Tu, Jinhu; Nian, Xiaohong; Dai, Xunhua
Journal/Conference: IEEE Internet of Things Journal
Link: 🔗 10.1109/JIOT.2026.3709310 📥 View PDF 📚 BibTeX Citation
Keywords: Digital twin Fault diagnosis Spatiotemporal graph Health management
📖 Expand to View Abstract
This paper proposes a twin-consistent spatiotemporal graph diagnostic network (TC-SGDN) to address the problem of confusion between maneuvering disturbances and fault characteristics under complex flight conditions. The method constructs a spatiotemporal graph using digital twin nodes and physical sensor nodes, describes sensor correlations and deviations through dynamic health baselines, and suppresses non-fault disturbances through consistency attention and temporal encoding. Experiments on real and simulated UAV data validate its diagnostic accuracy, robustness, and cross-condition generalization capability.
📄 Planar-Sector LOS Guidance for Interception of Agile Targets with Lifting-Wing Quadcopters¶
Authors: Liu, Linkai; Yang, Kun; Zou, Han; Min, Chen; Lv, Shuli; Wang, Shuai; Quan, Quan
Journal/Conference: arXiv preprint (v2, 2026)
Link: 🔗 arXiv:2606.10639v2 📥 View PDF 📚 BibTeX Citation
Keywords: Lifting-wing quadcopter PS-LOS guidance Visual interception Delay compensation
📖 Expand to View Abstract
This paper proposes a planar-sector line-of-sight (PS-LOS) guidance method and supporting control system for lifting-wing quadcopters, intended for image-based interception of agile targets. The method preserves maneuverability by relaxing conventional conical line-of-sight constraints and combines delay-compensated extended Kalman filtering with coordinated-turn control to provide continuous target state estimation and control commands adapted to lifting-wing dynamics. Outdoor flight experiments demonstrate interception capability against irregularly maneuvering targets.
2025¶
📄 RflyMAD: A dataset for multicopter fault detection and health assessment¶
Authors: Xiangli Le; Bo Jin; Gen Cui; Xunhua Dai; Quan Quan
Journal/Conference: The International Journal of Robotics Research
Link: 🔗 10.1177/02783649241305153 📥 View PDF 📚 BibTeX Citation
Keywords: fault detection health assessment multicopter dataset
📖 Expand to View Abstract
This paper releases RflyMAD, an open-source multicopter anomaly dataset developed by the Reliable Flight Control Team, to support research in fault detection and isolation, health assessment, and related fields. The dataset totals 114 GB and, in accordance with the ADS-33 standard, includes 6 flight states and 11 fault types, covering fault scenarios under different UAV maneuvering conditions. It comprises 5,629 flight cases with a total fault duration of 3,283 minutes, including semi-physical simulation, hardware-in-the-loop simulation, and real-flight data. The integration of RflySim simulation and real-flight data improves both the quantity and quality of the data. Each case contains multiple types of log and processed files and can serve as a benchmark for fault diagnosis methods. Transfer learning validates the supporting relationship between simulation and real-flight data. The dataset and tools have been open-sourced, and baseline methods, updated data, and additional fault types will be provided in the future.
📄 RflySim ToolChain: A Rapid Development and Validation Toolchain for Intelligent Unmanned Swarm Systems¶
Authors: Dai, Xunhua; Tu, Jinhu; Quan, Quan
Journal/Conference: Journal of Systems Engineering and Electronics
Link: 🔗 10.23919/JSEE.2025.000079 📥 View PDF 📚 BibTeX Citation
Keywords: swarm MBD hardware-in-the-loop Sim2Real
📖 Expand to View Abstract
The development of intelligent unmanned swarm systems is highly complex. Existing simulators and toolchains primarily focus on isolated technical elements, lacking systematic integration to support full-process technical and development requirements, and failing to effectively address the challenge of bridging algorithm simulation with actual deployment. There is an urgent need for a comprehensive solution covering the entire lifecycle. This paper proposes the RflySim ToolChain, specifically designed to facilitate the rapid development and validation of such systems. It adopts a model-based design approach, integrating modeling and simulation, low-level reliable control, and high-level swarm decision-making modules, covering the entire process from modeling and simulation to testing and deployment. This toolchain meets full-stack development requirements: its modular architecture and supporting SDK enable automated development workflows, while high-fidelity models and a reliable architecture achieve seamless transition from simulation to actual deployment. This paper also validates the practical effectiveness of the toolchain through a series of case studies.
📄 A Digital Twin-Level Modeling Framework for UAV Control Algorithm Optimization and Sim2real¶
Authors: Dai, Xunhua; Gong, Wanqi; Tu, Jinhu; Chen, Yong
Journal/Conference: Unmanned Systems (published online in 2025)
Link: 🔗 10.1142/S2301385027500464 📥 View PDF (Author Manuscript) 📚 BibTeX Citation
Keywords: Digital twin Control optimization LQR Sim2Real
📖 Expand to View Abstract
This paper proposes a digital twin-level modeling framework for UAV control algorithm optimization and simulation-to-real-world transfer, integrating multicopter dynamics, propulsion and aerodynamic characteristics, ground interaction, sensor simulation, and hardware-in-the-loop validation. Each subsystem is validated using experimental data, and an LQR controller is designed by combining simulation data with a small amount of real-flight data, then compared with PX4 and manually tuned PID controllers under wind disturbances. The results demonstrate that the framework helps improve control performance and reduce tuning and real-flight testing costs.
📄 Line-of-Sight-Constrained Multicopter Interceptability¶
Authors: Yang, Kun; Bai, Chenggang; Quan, Quan
Journal/Conference: Journal of Guidance, Control, and Dynamics
Link: 🔗 10.2514/1.G008527 📥 View PDF 📚 BibTeX Citation
Keywords: Multicopter Line-of-sight constraint Interceptability Visual interception
📖 Expand to View Abstract
This paper investigates whether a multicopter is capable of completing an interception while keeping the target within the camera field of view. It proposes definitions and calculation methods for global line-of-sight-constrained interceptability and the degree of interceptability (DOI). By combining the target's optimal strategy, relative-motion dynamics, and available thrust directions, analytical criteria are established, and design recommendations for camera mounting angles and aircraft configurations are provided. Numerical simulations and real-flight experiments validate the effectiveness of the theoretical criteria and configuration design method.
2023¶
📄 Lifting-wing Quadcopter Modeling and Unified Control¶
Authors: Quan, Quan; Wang, Shuai; Gao, Wenhan
Journal/Conference: arXiv preprint (v1, 2023)
Link: 🔗 arXiv:2301.00730v1 📥 View PDF 📚 BibTeX Citation
Keywords: Lifting-wing quadcopter Unified control Control allocation HIL simulation
📖 Expand to View Abstract
This paper introduces a lifting-wing quadcopter that combines the vertical takeoff and landing capability of multicopters with the forward-flight efficiency of fixed-wing aircraft. A dynamic model that simultaneously considers rotor thrust and wing aerodynamic forces is established, and a unified controller covering hover, transition, and forward-flight phases is designed. Optimization-based control allocation is used to coordinate rotor and aerodynamic actuators, reducing energy consumption while maintaining control capability. Hardware-in-the-loop simulation is used to validate the aircraft configuration and control method.
2021¶
📄 RflySim: A Rapid Multicopter Development Platform for Education and Research Based on Pixhawk and MATLAB¶
Authors: Shuai Wang, Xunhua Dai, Chenxu Ke and Quan Quan
Journal/Conference: 2021 International Conference on Unmanned Aircraft Systems (ICUAS)
Link: 🔗 10.1109/ICUAS51884.2021.9476786 📥 View PDF 📚 BibTeX Citation
Keywords: Education Pixhawk MATLAB/Simulink HIL
📖 Expand to View Abstract
This paper proposes and open-sources the RflySim platform—a rapid development platform for UAV education and research based on Pixhawk/PX4 and MATLAB/Simulink. Leveraging model-based development principles, the platform accelerates physical deployment through software-in-the-loop (SIL) and hardware-in-the-loop (HIL) simulations. With this platform, beginners and developers can directly use MATLAB/Simulink to design low-level controllers (e.g., attitude and position control) and high-level applications (e.g., decision-making and autonomous flight), and then deploy them onto multicopter autopilot systems without touching low-level C/C++ code. The usability and efficiency of the platform are validated through three demonstration cases.
📄 RFlySim: Automatic test platform for UAV autopilot systems with FPGA-based hardware-in-the-loop simulations¶
Authors: Xunhua Dai, Chenxu Ke, Quan Quan, Kai-Yuan Cai
Journal/Conference: Aerospace Science and Technology
Link: 🔗 10.1016/j.ast.2021.106727 📥 View PDF 📚 BibTeX Citation
Keywords: Automatic test HIL simulations FPGA Autopilot systems
📖 Expand to View Abstract
Unmanned aerial vehicle (UAV) autopilot systems are safety-critical systems, with increasingly stringent requirements for reliability and safety. However, testing complex autopilot control systems is a time-consuming and costly task, requiring extensive outdoor flight tests throughout the development phase. This paper proposes an indoor automatic test platform aimed at significantly improving UAV development efficiency and safety levels. First, a unified modeling framework suitable for various aircraft types is constructed to facilitate the sharing of common modeling experiences and failure modes. Subsequently, a real-time simulation platform is developed using automatic code generation and FPGA-based hardware-in-the-loop (HIL) simulation technologies to ensure the credibility of simulations at both hardware and software levels. Finally, an automatic test framework is proposed to traverse test cases and evaluate results within real-time flight simulations. The verification phase validates the platform's accuracy and credibility by comparing simulation results with experimental data. Successful application cases involving multicopters confirm the practicality of this platform.
📄 Simulation Credibility Assessment Methodology With FPGA-based Hardware-in-the-Loop Platform¶
Authors: Dai, Xunhua; Ke, Chenxu; Quan, Quan; Cai, Kai-Yuan
Journal/Conference: IEEE Transactions on Industrial Electronics
Link: 🔗 10.1109/TIE.2020.2982122 📥 View PDF 📚 BibTeX Citation
Keywords: Simulation credibility FPGA Hardware-in-the-loop Assessment methodology
📖 Expand to View Abstract
This paper proposes a simulation credibility assessment methodology based on an FPGA hardware-in-the-loop platform. By emulating sensor chips, the control system under test maintains consistent software and hardware operating environments in simulation and real experiments. On this basis, assessment metrics are designed from the perspectives of performance, time-domain response, and frequency-domain response, and different metrics are normalized and fused into an overall credibility evaluation. Simulations and experiments on a multicopter system demonstrate the effectiveness and practicality of the platform and assessment methodology.
2019¶
📄 An Analytical Design-Optimization Method for Electric Propulsion Systems of Multicopter UAVs With Desired Hovering Endurance¶
Authors: Dai, Xunhua; Quan, Quan; Ren, Jinrui; Cai, Kai-Yuan
Journal/Conference: IEEE/ASME Transactions on Mechatronics
Link: 🔗 10.1109/TMECH.2019.2890901 📥 View PDF 📚 BibTeX Citation
Keywords: Electric propulsion Design optimization Hovering endurance Multicopter
📖 Expand to View Abstract
This paper proposes an analytical design optimization method that meets target hovering endurance requirements, addressing the reliance on experience-based selection and repeated trial-and-error in multicopter propulsion system design. By separately establishing models for propellers, motors, electronic speed controllers, and batteries, the overall optimization problem is decomposed into several solvable subproblems to obtain the optimal parameters for each component, which are then matched with actual devices from product databases. Experiments and statistical analyses validate the effectiveness of the method and demonstrate its practical value for online propulsion system selection services.