China Advances Intelligent Driving Simulation with AI and Open Data, Report Highlights World Model Innovations

A new report titled “China and Global Intelligent Driving Simulation and World Model Research Report 2025” has been released by ResearchAndMarkets.com, highlighting advancements in simulation technologies and world models for autonomous driving. The study underscores how these tools are critical to addressing challenges in developing safer, more reliable self-driving systems. As the industry moves toward Level 3 and higher autonomy, the demand for high-quality data has intensified. End-to-end AI training for autonomous vehicles requires vast, diverse datasets, including multimodal information like images, text, and physical rules. However, real-world road data often lacks sufficient coverage of rare or complex scenarios—known as “long-tail” situations—which limits the effectiveness of algorithm training. To bridge this gap, automated simulation testing is gaining traction. It allows original equipment manufacturers (OEMs) and suppliers to reduce development costs, accelerate timelines, and address high-risk conditions that are difficult to replicate in real-world testing. World models, a key innovation in this space, use AI to create internal representations of the physical world. These models enable environmental state prediction, counterfactual reasoning, and the generation of unknown scenarios. They rely on multimodal data and reinforcement learning to simulate realistic conditions. While offering strong causal reasoning and scenario prediction capabilities, world models face challenges such as high computational demands, limited interpretability, and potential data biases. The report outlines four primary sources of scenario data for intelligent driving training: 1. Replay-based simulation: Uses real-world road test data to recreate specific situations, ensuring authenticity for verifying algorithm fixes. 2. Parametric scenarios: Defined through standardized formats like OpenScenario, allowing controlled testing of edge cases and boundary conditions. 3. Data conversion: Transforms real-world data into virtual scenarios (e.g., logsim to worldsim), supporting scalable and generalized testing. 4. World models: Leverage AI to generate and predict scenarios, improving coverage and efficiency but requiring robust computational infrastructure. Full-chain safety validation is reshaping simulation practices, shifting from isolated function testing to comprehensive, closed-loop verification. This evolution drives cross-domain integration, where simulation technologies merge software and hardware tools to address complex interactions between systems like intelligent cockpits, chassis, and thermal management. Industry efforts now focus on joint testing across domains, including vehicle-road-cloud integration and global digital twins. Improving simulation credibility remains a priority. The report notes that the sector is addressing gaps in scenario fidelity, sensor accuracy, and dynamic model reliability. Key strategies include: - AI integration: Enhancing automation in testing and validation processes, streamlining development workflows. - Open-source datasets: Initiatives like the China Association of Automobile Manufacturers’ (CAAM) efforts have released nearly 20 datasets, including public autonomous driving data and world model training resources, to reduce redundancy and foster collaboration. - Standardization: The ASAM OpenMATERIAL 3D 1.0.0 standard, launched in April 2025, establishes a framework for physical material properties and 3D object descriptions, improving realism in perception sensor simulations (e.g., LiDAR, radar, and cameras). Simulation tool providers are also upgrading platforms to meet industry needs. Recent updates include PreScan 2503, HEXAGON VTD/MSC/ADAMS/KISSoft, CarMaker 14.0, AURELION 24.3, MATLAB/Simulink R2025a, Ansys 2025R1, Oasis SIM 3.0, aiSim UE5.5, Qianxing V3.0, and PanoCarV1.7/PanoSim V33. These tools aim to enhance realism, scalability, and interoperability. The report covers critical topics, such as the application of world models in autonomous driving, the integration of AI with simulation technologies, and case studies of Chinese and global providers. It also profiles key companies in the field, emphasizing their role in shaping the future of intelligent driving. As the race for autonomous vehicle dominance intensifies, simulation and world model technologies are becoming essential for overcoming technical and logistical hurdles. The report highlights how collaboration, standardization, and AI-driven innovation are driving progress in this rapidly evolving sector.