Learning to Drive from a World Model
This addresses the need for scalable and simpler self-driving systems, though it appears incremental by building on existing simulation and learning approaches.
The paper tackles the problem of self-driving by proposing an end-to-end training architecture that learns driving policies from real data without hand-coded rules, using simulation methods including a learned world model, and shows these policies can be deployed in real-world systems.
Most self-driving systems rely on hand-coded perception outputs and engineered driving rules. Learning directly from human driving data with an end-to-end method can allow for a training architecture that is simpler and scales well with compute and data. In this work, we propose an end-to-end training architecture that uses real driving data to train a driving policy in an on-policy simulator. We show two different methods of simulation, one with reprojective simulation and one with a learned world model. We show that both methods can be used to train a policy that learns driving behavior without any hand-coded driving rules. We evaluate the performance of these policies in a closed-loop simulation and when deployed in a real-world advanced driver-assistance system.