×

ADAS Large Model

By leveraging a basic model with hundreds of billions of parameters and a world model driven by reinforcement learning, a multimodal intelligent driving large model (Motovis large model) is created. This technology combines generalization ability and scen

Vision module + language module + behavior module

At the same time, by leveraging the world model, the scene is reconstructed for reinforcement learning

The MOTOVIS VLA basic large model framework is the foundation of cognitive models. It is based on visual encoders, visual language models, and action strategy models, uses the world model as the reinforcement learning environment, and fine-tune with 4D datasets, VQA datasets, regulatory datasets, and physical interaction datasets.