WeRide GENESIS supports multi-model adaptation, ranging from L2+ ADAS to L4 autonomous driving systems. On January 28, WeRide officially launched its self-developed universal simulation model — WeRide GENESIS. WeRide GENESIS utilizes generative AI technology, enabling it to “generate worlds” within the virtual realm. This means it can construct highly realistic simulated urban environments in just a few minutes. Compared to traditional physical vehicle testing, WeRide GENESIS can accurately replicate extreme long-tail scenarios that are difficult to obtain frequently on real roads. This allows the autonomous driving system to undergo high-intensity training and validation in a virtual environment. This technological breakthrough stems from a common challenge faced by the global autonomous driving industry: significant differences exist between cities in terms of road networks, traffic habits, regulations, and climate. Traditional physical vehicle testing has shown limitations in scenario coverage and iteration efficiency. WeRide GENESIS features four built-in core AI modules. The first is AI Scenarios, which is responsible for constructing various critical situations. It comprehensively covers tens of billions of kilometers of natural driving data, along with the massive volume of long-tail and extreme cases collected by WeRide over more than eight years from public roads. The platform can simulate multiple scenarios that autonomous vehicles might encounter. These include everyday situations like adjacent vehicle incursions, unprotected left turns, and emergency avoidance, as well as rare events such as fires, earthquakes, road blockages, and extreme weather. AI Scenarios The AI Agents module constructs highly realistic behavioral models for traffic participants. It simulates interactive behaviors ranging from daily routines to high-risk situations, enhancing the system’s decision-making robustness within complex Operational Design Domains (ODDs). AI Agents The AI Metrics module establishes a quantitative evaluation system covering dimensions such as safety, compliance, comfort, and efficiency. It can transform driving behaviors into analyzable data performance, automatically determining the effectiveness of algorithm iterations. AI Metrics The AI Diagnosis module automates problem root cause analysis and repair suggestions. This module can automatically capture undesirable driving behaviors, analyze their fundamental causes, and provide actionable improvement plans. When an autonomous vehicle encounters perception delays or prediction errors in complex interactive scenarios, the system can rapidly diagnose, rectify, and re-validate performance, ensuring the vehicle operates as intended. AI Diagnosis Through the collaboration of these four AI modules, WeRide GENESIS builds a complete closed-loop iteration system. It automatically generates high-fidelity scenarios, automatically quantifies performance bottlenecks, automatically identifies weak points, and automatically provides optimization directions. Officially, WeRide states that GENESIS supports multi-model adaptation from L2++ advanced driver-assistance systems to L4 unmanned driving systems. It is also compatible with road elements and sensor configurations from different cities around the world. Currently, WeRide is conducting autonomous driving research, development, testing, and operations in over 40 cities across 11 countries globally.