KiGRAS: Kinematic-Driven Generative Model for Realistic Agent Simulation
Jianbo Zhao, Jiaheng Zhuang, Zhou Qibin, Taiyu Ban, Ziyao Xu, Hangning Zhou, Junhe Wang, Guoan Wang, Zhiheng Li, Bin Li
AI summary
Problem
Existing autoregressive trajectory generation models operate in a redundant state space, making it difficult to ensure physical feasibility and efficiently represent realistic driving patterns.
Approach
KiGRAS reformulates trajectory generation as predicting discrete control actions at each time step, using a kinematic model to map actions to physically valid states, creating a compact and physically constrained predictive space.
Key results
- Achieves state-of-the-art performance on Waymo’s SimAgents Challenge
- Reduces model parameters to 0.7M while outperforming larger models
- Ensures physical feasibility of all generated trajectories through kinematic constraints
- Enables driving habit customization via a unified fine-tuning approach
Why it matters
It lowers computational costs for realistic agent simulation, enabling safe and deployable autonomous driving systems in resource-constrained environments.
Abstract
Trajectory generation is a pivotal task in au- tonomous driving. Recent studies have introduced the autoregres- sive paradigm, leveraging the state transition model to approximate future trajectory distributions. This paradigm closely mirrors the real-world trajectory generation process and has achieved notable success. However, its potential is limited by the ineffective representation of realistic trajectories within the redundant state space. To address this limitation, we propose the Kinematic-Driven Generative Model for Realistic Agent Simulation (KiGRAS). In- stead of modeling in the state space, KiGRAS factorizes the driving scene into action probability distributions at each time step, providing a compact space to represent realistic driving patterns. By establishing physical causality from actions (cause) to trajectories (effect) through the kinematic model, KiGRAS eliminates massive redundant trajectories. All states derived from actions in the causal space are constrained to be physically feasible. Furthermore, redundant trajectories representing identical action sequences are mapped to the same representation, reflecting their underlying actions. This approach significantly reduces task complexity and ensures physical feasibility. KiGRAS achieves state-of-the-art performance in Waymo’s SimAgents Challenge, ranking first on the WOMD leaderboard with significantly fewer parameters than other models.