On September 24, Interface News learned that the World Model Team, an AI research team under Qianli Technology, recently released BehaviorWorldGen, a world model framework for autonomous driving, which mainly solves the problem that surrounding vehicles in existing world simulators are difficult to respond dynamically based on vehicle behavior. Its core component, BehaviorFlow, controls surrounding vehicle behavior by injecting interpretable frame-level “meta-actions” such as maintaining lanes, changing lanes, and turning into the vehicle, and can also generate multi-vehicle interaction scenarios such as gas jams, concessions, and intersection games for closed-loop training of autonomous driving motion models.

On September 24, Interface News learned that the World Model Team, an AI research team under Qianli Technology, recently released BehaviorWorldGen, a world model framework for autonomous driving, which mainly solves the problem that surrounding vehicles in existing world simulators are difficult to respond dynamically based on vehicle behavior. Its core component, BehaviorFlow, controls surrounding vehicle behavior by injecting interpretable frame-level “meta-actions” such as maintaining lanes, changing lanes, and turning into the vehicle, and can also generate multi-vehicle interaction scenarios such as gas jams, concessions, and intersection games for closed-loop training of autonomous driving motion models.