Gasgoo Munich – On September 16, during the 6th Automotive Semiconductor Ecosystem Conference hosted by Gasgoo, Jiang Shan, a senior basic software expert at the State Key Laboratory of High-end Automotive Integration and Control under FAW’s R&D Institute, delivered a keynote address discussing the evolving role of automotive basic software amid the rise of AI.
Jiang noted that the industry is transitioning from “software-defined vehicles” to “AI-defined vehicles,” presenting OEMs with challenges in selecting the right hardware and software stack. Rather than focusing solely on metrics like performance and cost, there’s a growing need for a cohesive system where CPUs, GPUs, NPUs, and MCUs work together. To provide the necessary computing power, automakers must integrate compilers, runtime libraries, and toolchains.
Historically, basic software adapted to chips using standard middleware, but the complexity of AI-era controllers requires it to transition from a simple adaptation layer to a resource management layer. Jiang emphasized the importance of efficiently organizing diverse hardware capabilities to achieve real software synergy.
An AIOS (AI Operating System) cannot merely support AI tasks; it must navigate resource conflicts and maintain functional safety as AI and control tasks coexist on the same chip. With Level 3 autonomy, the responsibility for AI becomes greater for automakers, necessitating that AI components are manageable and verifiable.
Jiang addressed the challenge of extending safety verification throughout the AI lifecycle. AI is a continuously evolving entity, and safety-related AI functions for autonomous driving require ongoing observation and management.
She highlighted that the automotive industry’s existing engineering frameworks, such as AUTOSAR and ASPICE, could help create effective boundaries for AI, which is critical as the technology evolves rapidly.
Finally, Jiang called for collaboration across the industry. Achieving synergy between chips and basic software involves a collective effort from OEMs, chip vendors, software developers, and various stakeholders, stressing that foundational capabilities must be developed upfront to address common challenges.
