Driving · 2026-10-14
YOUDrive: Driving Style as a Composable Direction for End-to-End Autonomous Driving
Guantian Zheng, Lidong Yu, Tianyu Gao, Jiaxing Chen, Jiashu Li, Chen Lv
AI-assisted summary based on the paper. Consult the linked source for methods and limitations. Benchmark numbers appear after editor verification.
TL;DR
YOUDrive is a vision–language–action planner that controls continuous, composable driving style at inference. Each teacher policy is distilled into a task vector that induces a style direction in a shared VLA planner; a flow-matching decoder uses this conditioned representation to sample kinematically feasible trajectory alternatives. The authors report monotonic assertiveness control on NAVSIM navtest while keeping PDMS above 0.90, reaching 0.954, and propose a Style Alignment Score (SAS).
Why it matters
The paper targets adjustable driving style in end-to-end autonomous driving without retraining the backbone. By turning teacher behaviors into reusable, scalable task-vector directions that condition a generative decoder, the approach aims to provide controllable behavior parameters at inference while maintaining planning safety and performance, with a separate metric (SAS) for human-referenced kinematic behavior.
Method
- Distill each teacher planner into a task vector (LoRA adapter update) relative to a shared VLA backbone; the resulting VLA content representation change defines a scene-dependent style direction.
- Condition a flow-matching trajectory decoder on scaled/composed style directions to sample multiple feasible control sequences integrated through a constrained motion model; select with a prediction-based PDM selector.
- Introduce Style Alignment Score (SAS) that gates trajectories by NAVSIM safety/task terms and then scores human-referenced kinematic assertiveness/proximity.
Limitation
SAS is validated along assertiveness (H). Proximity (P) is reported only as a secondary, preliminary axis; on navtest proximity atoms vary far less across policies than dynamics atoms (near-degenerate), so whether social style is independently controllable remains open and is left to future work.
Abstract (from arXiv)
End-to-end driving policies can achieve strong task performance while offering limited control over their driving style. Driving style describes how a vehicle accelerates, brakes, and maintains distance. Policies trained without explicit preference conditioning lack a direct mechanism for adjusting these behaviors at inference. We propose YOUDrive, a vision--language--action planner for continuous and composable driving-style control. Each teacher's behavior is distilled into a task vector, a learned parameter update that induces a style direction in the planner's representation. These directions can be scaled and combined at inference without retraining the backbone. Conditioned on these directions, a flow-matching decoder models a multimodal trajectory distribution and generates kinematically feasible motion alternatives. We introduce the Style Alignment Score (SAS) to describe human-referenced kinematic behavior after safety and task gating. On NAVSIM navtest, a single coefficient monotonically controls assertiveness across six teacher-derived personas, while PDMS remains above 0.90 across the tested individual and blended configurations, reaching 0.954. Further analyses reveal systematic policy-specific kinematics after controlling for scene context, whereas the evaluated discrete style labels provide limited additional explanatory power. Code, checkpoints, and the metric will be released.
Related papers
- Language2Drive: Do Vision-Language-Action Models Really Follow the Instructions in Autonomous Driving?
- Cartender: Verifiable Language-Conditioned Driving by Learning the Drives Not Taken
- A Survey on End-to-End Autonomous Driving Training from the Perspectives of Data, Strategy, and Platform
- doPlan: A Variable-Horizon Dataset for Multi-Stage Language-Conditioned Planning in Autonomous Driving
- TerraTransfer++: Demonstration-Free End-to-End Planner
Publication record
Bibliographic evidence describes where and when a record appeared. It does not validate scientific claims.
First public release: 2026-10-10 · Indexed by Pub-AI: 2026-10-14 UTC
YOUDrive: Driving Style as a Composable Direction for End-to-End Autonomous Driving
Preprint posted: 2026-10-10 · Original record ↗
- arxiv
- 2610.14802
Explicit source versions (1)
- v1 · 2026-10-10
- Preprint available
Source date: 2026-10-10 · First detected: 2026-10-25 UTC · Historical baseline
Source evidence (1)
- Official evidence ↗ · atom:published (official metadata)
Abstract · arxiv
End-to-end driving policies can achieve strong task performance while offering limited control over their driving style. Driving style describes how a vehicle accelerates, brakes, and maintains distance. Policies trained without explicit preference conditioning lack a direct mechanism for adjusting these behaviors at inference. We propose YOUDrive, a vision--language--action planner for continuous and composable driving-style control. Each teacher's behavior is distilled into a task vector, a learned parameter update that induces a style direction in the planner's representation. These directions can be scaled and combined at inference without retraining the backbone. Conditioned on these directions, a flow-matching decoder models a multimodal trajectory distribution and generates kinematically feasible motion alternatives. We introduce the Style Alignment Score (SAS) to describe human-referenced kinematic behavior after safety and task gating. On NAVSIM navtest, a single coefficient monotonically controls assertiveness across six teacher-derived personas, while PDMS remains above 0.90 across the tested individual and blended configurations, reaching 0.954. Further analyses reveal systematic policy-specific kinematics after controlling for scene context, whereas the evaluated discrete style labels provide limited additional explanatory power. Code, checkpoints, and the metric will be released.