Trajeglish: Traffic Modeling as Next-Token Prediction

arXiv:2312.04535v2 Announce Type: replace
Abstract: A longstanding challenge for self-driving development is simulating dynamic driving scenarios seeded from recorded driving logs. In pursuit of this functionality, we apply tools from discrete sequence modeling to model how vehicles, pedestrians and cyclists interact in driving scenarios. Using a simple data-driven tokenization scheme, we discretize trajectories to centimeter-level resolution using a small vocabulary. We then model the multi-agent sequence of discrete motion tokens with a GPT-like encoder-decoder that is autoregressive in time and …

abstract apply arxiv challenge cs.lg cs.ro data data-driven development driving dynamic logs modeling next pedestrians prediction self-driving simple token tokenization tools traffic type vehicles