Improving the safety of autonomous unmanned aerial vehicle teams through the creation of a systematic robustness assessment process.
Assuring the safety of teams of autonomous unmanned aerial vehicles (UAVs) that carry out a safety-critical inspection task collaboratively is hugely challenging. There are uncertainties and risks associated with the operating environment, individual UAV failures, an inconsistent global perspective between teams, interference because of limited physical space, and unreliable communication.
This project will develop a process for a systematic robustness assessment of UAV teams. This will be underpinned by methods for the specification, generation and testing of collaborative inspection scenarios, enabling the progressive transition from simulation to lab-based operations and to real-world operations.
The SAFEMUV project has carried out a comprehensive analysis of the state-of-practice in the assurance processes for UAV-based applications. Based on insights derived from this analysis and meetings with regulatory bodies and their industrial partners, the team developed the concept of operations needed for executing the safety analysis process of the SAFEMUV demonstrator.
The team has also started implementing a prototype pipeline comprising a dedicated domain-specific language, a model transformation engine and a modular fuzzing platform through which stakeholders will be able to define multi-UAV system and scenario specifications. These specifications will be transformed into executable models for the simulator of the SAFEMUV demonstrator as well as the code required for the real multi-UAV application.
The team has performed a safety assessment of the airframe inspection operation with multiple UAVs and derived operational safety objectives by specialising the Specific Operation Risk Assessment (SORA) methodology proposed by JARUS.
The assessment outcomes have been shared with the competent authorities and other project partners for consultation. On the technical level, the team implemented and populated a knowledge graph with formalised UAV-based mission concepts, their interrelations, barriers and mitigations that enhance the scenario specification engine, and advanced the prototypical fuzzing platform with support for analysing the robotic team’s capacity to withstand faults both at UAV- and team-level.
The next steps involve revising the specialised methodology and derived objectives, incorporating feedback from project partners, and completing the scenario specification and fuzzing engines, making the necessary adaptations for their planned integration starting in Q4.