As companies strive for greater levels of autonomy, more capability will be required of automotive ICs, and the challenge of ensuring functional safety is exacerbated. The mass public trusts companies to deliver safe products to the market, but can industries deliver on that promise given the demand for rapid innovation and complexity within the automotive ecosystem and supply chain? The scope of functional safety extends beyond the product boundaries to systems of interlinked devices representing the complete transportation network. From IP to automobile, each product plays a role in the overall functional safety of the transportation network. New paradigms and methodologies are required to ensure functional safety across all levels of the automotive ecosystem.
T panel, we will explore the challenges of addressing functional safety as new technologies and tools are introduced. Can we work together? Safety has been commonly used as a marketing lever. Should safety continue to be viewed as a differentiator and data kept private?
• What sort of standardization is required to ensure functional safety beyond ISO26262 and SOTIF?
• Should safety reports, metrics, and audits be regulated?
• Should interfaces and V2X data models be standardized? Should industry engage in interoperability testing?
• What is the ethical responsibility of companies to share data (telemetry, accident, etc…) with competitors?
How do you validate and verify the autonomous driving ecosystem? Autonomous vehicles will be required to operate independently (reactive) but the opportunity exists for enhanced functional safety with connectivity to infrastructure and cloud (proactive). The ISO26262 standard recommends safety analysis and testing to be performed as close as possible to the final product implementation. How does EDA enable verification in a V2X world?
• Is testing close to the final implementation (gate level) a reasonable and sustainable request given the trajectory of automotive ICs and systems?
• Are further levels of abstraction acceptable and to what degree? Can functional safety be closed in a digital representation (digital twin) or must physical testing be performed to supplement and validate the digital results?
• To support rapid development and deployment of machine learning algorithms, High Level Synthesis technologies are gaining traction. Can safety be “correct by construction” via HLS? Can fault testing be executed at that higher level of abstraction? How much testing is enough for IC Functional Safety? Functional correctness is about testing the positive space while verifying functional safety is about testing the larger negative space.
Today, functional safety verification continues to challenge even the most advanced verification teams.
• Current fault simulators leverage functional regressions and the completeness of a fault campaign is reliant on the quality of functional regression stimulus. How does a user know how much testing is enough and is current industry practice acceptable? How can formal and functional coverage metrics be used to aid and/or supplement fault simulations?
• Are traditional approaches to safety sufficient and if not, what new approaches are required to ensure safety across increasingly complex ICs? What techniques can be applied to make the safety workflow more efficient?