SAE J3321_202603

This document intends to focus on AI/ML systems deployed on board and/or off board the vehicles, including the methods and tools used anywhere in the life cycle of automotive system development. This includes, for instance, AI/ML systems used in the identification of parameters, calibration of classical or neural network-based control algorithms, system level modeling and validation, data process automation pipeline such as automated labeling and annotation, data quality checks, etc. Furthermore, with the continued success of Generative AI (GenAI) capabilities, it is becoming a common practice to leverage these capabilities beyond just generating synthetic data. Multiple artifacts such as code, testing scenarios, and benchmarks could be developed using GenAI tools, necessitating the additional verification of these “generative” development steps. The fundamental challenge with generative methods is grounding, especially if data, scenes, and tests are used for the development of AI-enabled products and features. AI and ML systems have been the subject of several national and international standards and guideline documents. A brief selection of these are given in Section 2. One, ISO PAS 8800:2024, focuses on safety over the entire life cycle of AI development and deployment. This document is intended to serve the following purpose: The document is an information report with no mandatory requirements. The focus of the document is only on V&V of the development life cycle of AI-based components or systems. The V&V methods are general in nature. The methods discussed include, but are not restricted to, safety requirements or properties. Software and hardware components realizing Artificial Intelligence (AI) and Machine Learning (ML) algorithms are becoming commonplace in next-generation ground vehicle applications and services. Many of these applications will have a bearing on the safety of vehicles. The development life cycle of AI/ML applications is atypical in the sense that it is highly data driven, performs high-level, complex human-like tasks (e.g., perception, recognition, comprehension, decision, etc.), and is opaque by design. Typical use cases involve human-centric concepts and domains (e.g., roadside objects, lanes, traffic signals, pedestrians etc.), each of which have different levels of abstractions. Robustness and certainty of the performance of AI-based systems are some of the major issues facing such systems. The systematic verification and validation (V&V) of AI-based systems is therefore critical when these systems are used in automotive applications and features impacting the trustworthiness, including safety, of the vehicle functions. As such, the V&V of AI-enabled components and systems requires new, enhanced approaches as existing methods are often narrowly focused and relatively immature. Additionally, end users and practitioners face challenges due to insufficient experience in developing new AI-based systems, the absence of uniform industry standards, and a lack of focused and reliable commercial tools. Given the complex network of OEMs and suppliers in the automotive domain, this lack of a standardized approach can very often lead to the use of ad hoc techniques for V&V with varying degrees of component/system performance and reliability. Fortunately, recent and ongoing developments by both academic and industrial researchers in this area have led to new insights into the formal and standardized approaches for V&V of AI/ML applications. The purpose of this information report is to: Review the current state-of-the-art (SOTA) in the verification and validation (V&V) of AI/ML-based systems and to provide a comprehensive guide for incorporating these methods into the systematic V&V of automotive AI-based systems. Identify gaps and deficiencies in the existing V&V approaches with respect to their application to AI-based SW and define requirements for their enhancements. Provide guidelines on the application of existing and new approaches for the V&V of AI-based components/systems.