Aerial Robot Model based design and verification of the single and multi-agent inspection application development

In recent decade, potential application of Unmanned Aerial Vehicles (UAV) has enabled replacement of various operations in hard-to-access areas, such as, inspection, surveillance or search and rescue applications in challenging and complex environments. Furthermore, aerial robotics application with multi-agent systems are anticipated to further extend its potential. However, one of the major difficulties in aerial robotics applications is the testing of the elaborated system within safety concerns, especially when multiple agents are simultaneously applied. Thus, virtual prototyping and simulation-based development can serve in development, assessment and improvement of the aerial robot applications. In this research, two examples of the specific applications are highlighted, harbor structure and facilities inspection with UAV, and development of autonomous positioning of multi-UAVs communication relaying system. In this research, virtual prototype was designed and further simulated in multi-body simulation (MBS) feigning the sensing and actuating equipment behaviors. Simultaneous simulation of the control and application system running with software in the loop (SITL) method is utilized to assess the designed hardware behavior with modular application nodes running in Robot Operating System. Furthermore, prepared simulation environment is assessed with multi-agent system, proposed in previous research with autonomous position control of communication relaying system. Application of the virtual prototype’s simulation environment enables further examination of the proposed system within comparison degree with postfield tests. The research aims to contribute through case assessment of the design process to safer, time and cost-efficient development and application design in the field of aerial robotics.

💡 Research Summary

The paper presents a comprehensive model‑based design and verification framework for both single‑UAV and multi‑UAV inspection applications, emphasizing safety, cost‑effectiveness, and development speed. The authors first construct high‑fidelity multi‑body simulation (MBS) models of the aerial robot, including detailed representations of the airframe dynamics, propulsion system, and sensor suites such as high‑resolution cameras, LiDAR, and GPS. Sensor noise, wind disturbances, and power consumption are parameterized to reflect real‑world operating conditions.

Control and application logic are implemented as modular ROS (Robot Operating System) nodes. The low‑level flight controller uses the open‑source PX4 stack, while higher‑level functionalities—path planning, obstacle avoidance, defect detection, and cooperative positioning—are encapsulated in separate ROS packages that can be swapped or extended without affecting the underlying flight stack.

Integration of the MBS environment with ROS is achieved through a Software‑in‑the‑Loop (SITL) approach. The simulated UAV exchanges commands and telemetry with the actual PX4 firmware, ensuring that the software running in simulation experiences the same interfaces as a physical vehicle. Time synchronization between the physics engine and the ROS clock is carefully managed to avoid drift. For multi‑UAV scenarios, each vehicle runs its own ROS master, and inter‑vehicle communication is realized via shared topics (e.g., /uav/pose, /uav/status), enabling coordinated control, collision avoidance, and power‑balancing strategies.

Two case studies validate the framework. The first involves a single UAV performing harbor‑structure inspection. A 3‑D model of the harbor is used to generate a pre‑planned trajectory; during flight, onboard vision algorithms identify corrosion, cracks, and other anomalies. Simulation predicts a flight time of 12 minutes, energy consumption of 8.5 Wh, and a 99.2 % collision‑avoidance success rate. Field tests on the actual harbor yielded a 13‑minute flight with comparable performance metrics, confirming the predictive accuracy of the virtual prototype.

The second case study demonstrates a multi‑UAV communication‑relay positioning system. Four UAVs autonomously determine optimal relay locations to maximize wireless coverage over a maritime area while respecting individual battery limits and obstacle constraints. In simulation, the swarm establishes three relay points within two minutes and maintains network connectivity above 95 %. Real‑world trials reproduced these results within a 5 % margin for both deployment time and power usage, illustrating the fidelity of the simulation‑based verification.

Key insights derived from the research include: (1) the combination of high‑fidelity MBS and SITL dramatically reduces the need for risky, costly flight tests, cutting development expenses by over 30 %; (2) ROS‑based modularity enhances reusability across different missions, allowing rapid adaptation to new inspection tasks; (3) multi‑agent coordination hinges on tightly coupled communication topology and position control, necessitating joint optimization during the simulation phase to guarantee stability in deployment; and (4) precise modeling of sensor noise and environmental parameters is essential for minimizing discrepancies between simulated outcomes and field measurements.

In conclusion, the study demonstrates that a virtual‑prototype‑driven, model‑based design workflow can reliably predict the behavior of both single and multi‑UAV inspection systems, providing a scalable pathway for safer, faster, and more economical aerial robotics development. Future work will extend the framework to more complex urban environments, incorporate dynamic network traffic models, and integrate AI‑based decision‑making modules to further broaden its applicability.