TrajTok: Technical Report for 2025 Waymo Open Sim Agents Challenge
This work addresses trajectory prediction for autonomous driving systems, representing an incremental improvement in a domain-specific application.
The authors tackled trajectory prediction for autonomous vehicles by developing TrajTok, a trajectory tokenizer that combines data-driven and rule-based methods with improved coverage, symmetry, and robustness, along with a spatial-aware label smoothing technique. They applied this to the SMART model and achieved a realism score of 0.7852 on the Waymo Open Sim Agents Challenge 2025.
In this technical report, we introduce TrajTok, a trajectory tokenizer for discrete next-token-prediction based behavior generation models, which combines data-driven and rule-based methods with better coverage, symmetry and robustness, along with a spatial-aware label smoothing method for cross-entropy loss. We adopt the tokenizer and loss for the SMART model and reach a superior performance with realism score of 0.7852 on the Waymo Open Sim Agents Challenge 2025. We will open-source the code in the future.