PUBLISHER: ResearchInChina | PRODUCT CODE: 2039282
PUBLISHER: ResearchInChina | PRODUCT CODE: 2039282
Research on automotive SOA and cross-domain middleware: The era of AI atomic services and AI cross-domain fusion agents is coming.
Driven by intelligence such as AI, the entire automotive software technology has gradually transformed from software-defined vehicles (SDVs) to AI-defined vehicles, reflected in EEAs shifting to centrally integrated AI native architectures, software SOA + AI native, from technology iteration cycle of 5-7 years to model iteration in days, etc.
AI atomic services, based on SOA, provide AI agents with greater scheduling flexibility and scenario adaptability through a more refined breakdown of full SOA capabilities, enabling vehicles to truly possess the intelligent capabilities to "understand requirements, predict requirements, and meet requirements".
At the end of 2025, SAIC Group released the full-stack 4.0 AI EEA, which reshapes the automotive architecture with AI as the core. This architecture proposes a "SOA+A2A/MCP Framework + Agent" tripartite ecosystem to build a new paradigm of Agentic AI collaborative intelligence. Agentic AI becomes the agent engine, covering multiple special agents such as music entertainment, schedule management, energy replenishment services, intelligent vehicle control and intelligent driving assistance, and realizes all-scenario intelligent services with core modules such as world model, intention understanding, and planning execution. This architecture is planned to be mass produced in 2027.
In April 2026, GAC Group officially released the latest next-generation smart architecture - X-Soul 4.0 with full-domain integration. Based on this architecture, 3,600 functional atoms are created, nearly doubling the number compared with the previous generation. Each atom means a basic capability. Users can freely combine exclusive automotive scenarios like building blocks. Currently, GAC car owners have created more than 100,000 user-defined scenarios based on this system, making vehicles truly exclusive mobility partners. In the future, an AI scenario agent will be launched to proactively adapt to user habits.
In April 2026, SAIC Roewe and Volcano Engine jointly launched the AI native technology architecture - CPP to build 2000+ SOA full-domain service interfaces, allowing vehicles to achieve atomic-level software and hardware information disassembly and precise scheduling. This architecture realizes the deep integration of AI and vehicle hardware through the Context-Planner-Pixel closed-loop design, so that AI can not only perceive and make decisions, but also directly control vehicle functions, forming a complete chain of "perception-decision-execution", which is the technical basis for the "Jiayue" series of Roewe's AI native vehicles.
New concept for cross-domain fusion middleware: AI foundation models + full vehicle empowerment by super agent
Driven by AI foundation models, AI agents can break through the traditionally separated technical barriers between domains and links in the automotive industry with their autonomous perception, decision collaboration, and continuous evolution capabilities. They can internally connect multi-domain technology modules such as intelligent driving, cockpits, and vehicle control to achieve cross-domain data flow and functional collaboration. They can also externally link the vehicle-road-cloud ecosystem, user needs, and supply chain resources to form a "requirement - development - implementation - feedback" closed loop.
In 2026, OEMs actively deploy AI cross-domain fusion agent products, such as IM Ultra Agent and Geely Super Eva. Agents are gradually reshaping the core value of vehicles which are no longer simply transportation tools, but have become "super intelligent partners" capable of understanding, thinking, and acting, propelling intelligent mobility into a new era. Suppliers are also actively exploring the implementation of cross-domain fusion products based on AI foundation models + agents.
In April 2026, Huawei officially released HarmonySpace 6, in which Xiaoyi was upgraded to an all-scenario cross-domain interactive cockpit agent as Xiaoyi's core MoLA architecture evolved to version 2.0. While improving the in-cockpit experience, the new Xiaoyi can also be linked with ADS to support fuzzy relative position search and fuzzy feature search, further improving usability. Xiaoyi also supports third-party agent access to realize functions such as ordering food, ordering coffee, and listening to podcasts.
In April 2026, Horizon Robotics officially released KaKaCla, a vehicle agent operating system which realizes the intelligent interaction paradigm of "task as a service". With only natural language instructions, users can schedule intelligent driving and intelligent cockpit in parallel at the same time. Physical agents, digital agents and edge-cloud foundation models respond synchronously to automatically plan and execute cross-domain operation processes. Through the combination of HSD (physical AI driving capability) + KaKaClaw (digital AI interaction capability), Horizon Robotics has completely integrated physical world driving and digital world interaction at the mass production level, allowing vehicles to have all agent capabilities that can "drive, chat, act, and grow".
Automotive software tools evolve to AI native tool flow, reconstructing a new paradigm of software development
As automotive intelligence accelerates, traditional software development models are facing challenges such as low human-machine collaboration efficiency and long delivery cycles. Driven by the accelerated development of AI agents, autonomous vehicle software development tools are transforming from a traditional modular architecture to an AI agent-driven paradigm. In this transformation, AI is no longer an efficiency improvement tool in a certain link, but a change throughout the entire life cycle of requirements, design, development, testing, review, deployment, operation and maintenance. At the same time, the collaboration model has gradually evolved from "people + tools" to the AI native tool flow model of "people + agent groups". In recent years, OEMs and suppliers have also been actively exploring new changes in AI-driven software development processes.
In April 2026, Mercedes-Benz formed a strategic partnership with Cognition to deploy Devin and Windsurf across its global engineering teams. This is the first large-scale deployment of AI software engineering in the automotive industry. Windsurf, as an AI-driven IDE, serves teams building cloud-native applications, microservices, and modern web and mobile solutions; Devin acts as the autonomous, cloud-based agent for async work. Devin for Terminal handles the most challenging engineering tasks, providing command-line level autonomous execution capabilities.
In April 2026, KOTEI officially released SDW AIKO, an automotive software development platform for the AI native era. It integrates deep human-machine collaboration with development process refactoring, driving the entire lifecycle of automotive software engineering with an AI agent. From project initiation to continuous optimization, SDW AIKO enables AI to function like a collaborative team, participating in analysis, planning, and execution, continuously accumulating experience, and self-evolving-becoming smarter with application-truly upgrading AI from an auxiliary tool to a development partner.