This SBIR addresses the need for robust airborne software design processes that are scalable, efficient, and low-cost to ensure safety features and design assurance for unmanned aerial systems (UAS) and manned aircraft. Xwing’s proposed tool (Tracer) manages software requirements and traceability at all levels; from the system-level objectives and use cases to the software requirements, corresponding code and the test cases. Tracer enables stakeholders and engineers to identify bidirectional impacts of changes to requirements or system code.
Xwing is currently developing capabilities for detect-and-avoid (DAA) for autonomous aircraft, and deriving and generating requirements based on DO-365. Those requirements were used as a Phase I test case for Tracer. The use case scope focused on tracking for detection of threats. The use case served to collect valuable usability experience and information, gather feedback and improve Tracer functionality.
The work in Phase I was performed along two main workstreams: 1) development of the requirements tracing tool, and 2) refinement through implementation of a DAA use case. Phase II will continue these workstreams, and add a third : development of a framework for certification of AI-based systems. These certification frameworks will be applied to vision-based DAA, expanding on conventional deterministic DAA previously explored.
The current certification process for avionics is not adapted to complex algorithms and to “AI-based” algorithms such as deep neural networks used in vision-based DAA. Despite their lack of explainability such algorithms can exhibit high performance. Xwing recognizes that such algorithms will enable the automation of functions where human judgement was previously necessary. It may also reduce the size, weight, and power required by some sensors. Frameworks added to Tracer in Phase II will help realize these benefits.
The Tracer tool, along with its application to DAA requirements and the extension to AI-based systems, supports the System-Wide Safety Project at NASA Ames Research Center. Since Tracer is a distributed tool, it has the potential to be a useful for any project where NASA is a stakeholder. Users of the tool could include a NASA program manager tasked with ensuring that requirements are met. Another opportunity is to build pre-packaged sets of requirements and test cases that are pre-approved for airworthiness, thus expediting new flight tests.
The Tracer software is intended for sale to manufacturers of aerospace who wish to achieve levels of certification for those systems, which may range from lightweight small UAS up to manned Part 25 aircraft. Other potential customers are manufacturers of different safety-critical systems (e.g., automotive, naval, medical) that have a strong software component (e.g., autonomous systems)