AI-Based Electronic Control Units for Software-Defined Vehicles: A Structured Review of Architectures, Techniques, and Hardware Implementations

The automotive electrical/electronic (E/E) architecture is undergoing a fundamental transformation from loosely coupled, function-specific electronic control units (ECUs) toward domain-centralized, zonal, and fully centralized computing architectures, forming the foundation of emerging software-defined vehicles (SDVs). In parallel, artificial intelligence (AI) and machine learning (ML) are moving from predominantly cloud-based analytics toward direct deployment within automotive ECUs, enabling perception, prediction, diagnosis, cybersecurity, and adaptive control on resource-constrained embedded platforms. This paper presents a structured literature review of AI/MLenabled ECU design within the SDV paradigm. It first introduces ECU fundamentals, functional roles, and invehicle networking, including a representative zonal SDV architecture integrating controller area network (CAN), CAN with flexible data-rate (CAN-FD), local interconnect network (LIN), and automotive Ethernet. The evolution from distributed to domain, zonal, and centralized architectures is then discussed. Major ECU categories, including powertrain, transmission and energy management, battery management systems (BMSs), chassis and safety, advanced driver assistance systems (ADAS) and autonomous driving, body control, infotainment and connectivity, and gateway, zonal, and centralcompute ECUs, are systematically reviewed using functional diagrams, AI/ML techniques, datasets, and relevant literature. The paper also examines AI/ML training, optimization, compression, validation, and deployment on automotive system-on-chips (SoCs), with emphasis on the automotive open system architecture (AUTOSAR) adaptive platform, edge-AI accelerators, overthe-air (OTA) updates, field-programmable gate array (FPGA)-accelerated inference, and application-specific integrated circuit (ASIC)-based implementations. Finally, key challenges in functional safety, real-time determinism, explainability, resource constraints, cybersecurity, and reliable software updates are discussed, followed by future research directions for efficient and hardware-aware AI/ML integration in SDV-oriented ECU architectures.