DAP: A Discrete-token Autoregressive Planner for Autonomous Driving

TL;DR

DAP introduces a discrete-token autoregressive planning model for autonomous driving that jointly predicts scene semantics and ego trajectories, enhancing representation learning and performance with a compact model size.

Contribution

The paper presents DAP, a novel discrete-token autoregressive planner that jointly forecasts BEV semantics and ego trajectories, enabling better scene understanding and planning in autonomous driving.

Findings

Achieves state-of-the-art open-loop metrics.

Delivers competitive closed-loop results on NAVSIM.

Operates effectively with only 160M parameters.

Abstract

Gaining sustainable performance improvement with scaling data and model budget remains a pivotal yet unresolved challenge in autonomous driving. While autoregressive models exhibited promising data-scaling efficiency in planning tasks, predicting ego trajectories alone suffers sparse supervision and weakly constrains how scene evolution should shape ego motion. Therefore, we introduce DAP, a discrete-token autoregressive planner that jointly forecasts BEV semantics and ego trajectories, thereby enforcing comprehensive representation learning and allowing predicted dynamics to directly condition ego motion. In addition, we incorporate a reinforcement-learning-based fine-tuning, which preserves supervised behavior cloning priors while injecting reward-guided improvements. Despite a compact 160M parameter budget, DAP achieves state-of-the-art performance on open-loop metrics and delivers competitive closed-loop results on the NAVSIM benchmark. Overall, the fully discrete-token autoregressive formulation operating on both rasterized BEV and ego actions provides a compact yet scalable planning paradigm for autonomous driving.