Generative AI–Assisted Debugging of Safety-Critical Cockpit Systems in Software-Defined Vehicles

Modern software-defined vehicle architectures use digital cockpits for infotainment, which deliver speed information, warning telltales, rear-visibility imagery, and other outputs to the driver. These outputs are federally regulated due to the nature of the information they convey and are classified as safety-critical. As these systems integrate complex system-on-chip (SoC) platforms, high-speed display links, and other complexities across multi-brand portfolios, debugging failures has become a data-intensive, highly manual task. Current approaches rely on engineers spending hours analyzing fragmented logs and traces, followed by extensive reproduction efforts, and ultimately on the experience of a small group of expert engineers. This paper proposes a Generative AI-assisted framework to address the limitations of current approaches and to support the diagnosis and resolution of safety-critical cockpit issues in production-vehicle programs. The framework is built on augmenting GenAI-based debugging with existing engineering workflows to increase efficiency and support the diagnosis and resolution of problems at scale. The framework proposes four application areas: Using representative case studies from digital cockpit programs, the paper qualitatively illustrates how these capabilities can reduce manual analysis effort, improve reuse of institutional debug knowledge, and shorten time-to-insight while maintaining human control, data governance, and regulatory traceability. The paper concludes by evaluating future work proposals and identifying areas for development in the Gen AI-assisted debugging effort.