CARE-Drive: Coupled Alternating Refinement for Ego-trajectories
Organizations: NEXTAI Research Institute, Chery Group
Abstract
Vision-language-action (VLA) models bring multimodal understanding and world knowledge to end-to-end autonomous driving. To further improve the quality of predicted trajectories, existing studies introduce candidate refinement, residual denoising, and token correction. However, these methods typically focus on unified refinement of the entire trajectory, paying limited attention to the distinct physical quantities and their interactions. We further investigate physical specialization and coordination in trajectory refinement, centering on position and heading for simpler and more intuitive correction modeling. To this end, we propose CARE-Drive (Coupled Alternating Refinement for Ego-trajectories), which decomposes trajectory refinement into waypoint residual correction and position-conditioned heading generation. Through imitation learning and a future-token auxiliary task, the backbone provides predictive representations and a base trajectory. The position expert predicts bounded metric waypoint residuals, while the heading expert reconstructs vehicle orientation from adjacent angular increments conditioned on the corrected path. With the backbone frozen, the two experts are optimized alternately through supervised initialization and reinforcement fine-tuning. This design simplifies refinement through clear physical specialization while preserving conditional coupling during prediction and optimization. On NAVSIM v1, CARE-Drive achieves 88.8 PDMS, improving by 2.4 points over the unrefined base trajectories generated by the same backbone. Ablation studies further support the effectiveness of specialized experts and staged training.
Figures & tables
| Family | Method | Sensors | Frames | NC | DAC | TTC | Comf. | EP | PDMS |
|---|---|---|---|---|---|---|---|---|---|
| End-to-end | UniAD [ 4 ] | C | 97.8 | 91.9 | 92.9 | 100.0 | 78.8 | 83.4 | |
| Hydra-MDP [ 10 ] | C+L | 98.3 | 96.0 | 94.6 | 100.0 | 78.7 | 86.5 | ||
| DiffusionDrive [ 11 ] | C+L | 98.2 | 96.2 | 94.7 | 100.0 | 82.2 | 88.1 | ||
| World models | LAW [ 6 ] | C | 96.4 | 95.4 | 88.7 | 99.9 | 81.7 | 84.6 | |
| Epona [ 18 ] | C | 97.9 | 95.1 | 93.8 | 99.9 | 80.4 | 86.2 | ||
| PWM [ 19 ] | C | 98.6 | 95.9 | 95.4 | 100.0 | 81.8 | 88.1 |
| Configuration | Full-test PDMS | Gain over backbone |
|---|---|---|
| Backbone only | 86.462 | — |
| Add heading expert | 87.562 | |
| Add position expert; freeze heading expert | 88.672 | |
| Freeze position expert; adapt heading expert | 88.837 |
| Variant | Future queries | Supervision | PDMS |
|---|---|---|---|
| No future branch | 0 | No | 74.310 |
| Queries only | 96 | No | 74.680 |
| Auxiliary task | 96 | Yes | 76.800 |
| Student vision tower | Test PDMS |
|---|---|
| Frozen | 74.198 |
| Unfrozen (including merger) | 69.097 |
| Training | Full-test PDMS |
|---|---|
| Uniform weighting | 85.863 |
| Dynamic weighting | 86.462 |
| Configuration | PDMS |
|---|---|
| Backbone only | 86.462 |
| Staged expert training (CARE-Drive) | 88.837 |
| Joint SFT | 85.651 |
| Joint SFT + GRPO | 83.671 |
| Configuration | Added parameters (M) | PDMS |
|---|---|---|
| Backbone only | 0 | 86.462 |
| MLP decoder expansion | 156.84 | 86.510 |
| Separate experts (CARE-Drive) | 156.84 | 88.837 |
| Single expert, final RL | 156.84 | 84.227 |
| Single expert, after SFT | 156.84 | 85.147 |
| Reward coefficient | Test PDMS | PDMS |
|---|---|---|
| 0 (original reward) | 88.738 | — |
| 0.5 | 88.749 | |
| 1.0 | 88.840 | |
| 2.0 | 88.720 |
Appendix figures & tables9 assets
Supplementary material from the paper’s appendix.
Appendix
| Stage | Updated component | Objective / role | Fixed context | Updates | Val497 PDMS |
|---|---|---|---|---|---|
| A0 | Backbone | Dynamic weighted L1 + future features | Frozen visual tower | 3,625 | 87.611 |
| B1 | Heading | Token SFT | Backbone; base XY | 3,620 | 87.527 |
| B2 | Heading | First official-reward RL stage | Backbone; base XY | 3,620 | 89.581 |
| B3 | Heading | Second official-reward RL stage | Backbone; base XY | 3,620 | 90.126 |
| C1 | Position | Residual-token SFT | Backbone; B3 heading helper | 3,620 | 90.128 |
| C2 | Position | Official reward; global RMS | Backbone; B3 heading helper | 3,620 | 90.924 |
| System | Test PDMS |
|---|---|
| L1 backbone, epoch 4 | 85.988 |
| Historical experts (H) | 88.810 |
| Cross-attention | 88.726 |
| Dynamic / R0 | 88.837 |
| Variant | Val497 | Full PDMS | vs. R0 |
| R0: original recipe | 90.675 | 88.837 | — |
| R1: position penalty | 90.819 | 88.682 | |
| R2: heading penalty | 90.949 | 88.655 | |
| R1 interval: | |||
| R2 interval: | |||
| Variant | NC | DA | Mean cost |
|---|---|---|---|
| R0 | 204 | 497 | 0.0567 |
| R1 | 208 | 502 | 0.0575 |
| R2 | 201 | 522 | 0.0585 |
Vehicle Trajectory Prediction
Vehicle trajectory prediction forecasts how vehicles will move from their histories, road maps, traffic interactions, and driving intentions.
See all 69 papers