VADv2 is an end-to-end autonomous driving model that addresses the uncertainty and non-deterministic nature of planning in autonomous driving. It uses probabilistic planning to model the planning policy as an environment-conditioned non-stationary stochastic process, formulated as \( p(a|o) \), where \( o \) is the historical and current observations of the driving environment, and \( a \) is a candidate planning action. The model takes multi-view image sequences as input, transforms sensor data into environmental token embeddings, outputs the probabilistic distribution of actions, and samples one action to control the vehicle. VADv2 achieves state-of-the-art closed-loop performance on the CARLA Town05 benchmark, significantly outperforming existing methods, even with only camera sensors. The model runs stably in a fully end-to-end manner, demonstrating its effectiveness in complex driving scenarios. The paper also presents qualitative results and ablation studies to validate the key components of VADv2.VADv2 is an end-to-end autonomous driving model that addresses the uncertainty and non-deterministic nature of planning in autonomous driving. It uses probabilistic planning to model the planning policy as an environment-conditioned non-stationary stochastic process, formulated as \( p(a|o) \), where \( o \) is the historical and current observations of the driving environment, and \( a \) is a candidate planning action. The model takes multi-view image sequences as input, transforms sensor data into environmental token embeddings, outputs the probabilistic distribution of actions, and samples one action to control the vehicle. VADv2 achieves state-of-the-art closed-loop performance on the CARLA Town05 benchmark, significantly outperforming existing methods, even with only camera sensors. The model runs stably in a fully end-to-end manner, demonstrating its effectiveness in complex driving scenarios. The paper also presents qualitative results and ablation studies to validate the key components of VADv2.