Research map / Label-efficient and robust learning / Recognition and long-tail
Trajectory-guided Control Prediction for End-to-end Autonomous Driving: A Simple yet Strong Baseline
Penghao Wu, Xiaosong Jia, Li Chen, Junchi Yan, Hongyang Li, Yu Qiao
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add trajectory branch as auxiliary task to control model (ablation) | Control-Only model · CARLA 4-town validation | Driving Score: 32.45 -> 34.98 | Improved |
| add multi-step control prediction with temporal module (ablation) | Control + traj-task · CARLA 4-town validation | Driving Score: 34.98 -> 42.87 | Improved |
| add trajectory-guided attention (ablation) | Control + traj-task + temporal · CARLA 4-town validation | Driving Score: 42.87 -> 46.08 | Improved |
| add situation based fusion of trajectory and control outputs (ablation) | TCP · CARLA 4-town validation | Driving Score: 46.08 -> 57.01 | Improved |
| add trajectory-guided attention (ablation) | TCP · CARLA 4-town validation | Route Completion: 87.51 -> 84.95 | Hurt |
| use multi-task learning with shared backbone instead of ensemble of two separate models (ablation) | Control and trajectory branches · CARLA 4-town validation | Driving Score: 45.03 -> 48.27 | Improved |
| share encoder between branches instead of separate encoders (TCP vs TCP-SB) (ablation) | TCP · CARLA 4-town validation | Driving Score: 52.46 -> 57.01 | Improved |
| add trajectory-guided multi-step control and attention on top of MTL (ablation) | MTL · CARLA 4-town validation | Driving Score: 48.27 -> 57.01 | Improved |
| ensemble TCP and TCP-SB | TCP · CARLA 4-town validation | Driving Score: 57.01 -> 59.09 | Improved |
| use fusion weight alpha=0 (only specialized branch) rather than intermediate alpha (ablation) | situation based fusion · CARLA 4-town validation | Driving Score | Hurt |
| use trajectory-only output instead of control-only output | Single-branch model · CARLA 4-town validation | Driving Score: 32.45 -> 28.29 | Hurt |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- Transfuser — TransFuser: Imitation With Transformer-Based Sensor Fusion for Autonomous Driving (TPAMI 2022) · better on 1 of 1 shared result
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Future-Aware End-to-End Driving: Bidirectional Modeling of Trajectory Planning and Scene Evolution (NeurIPS 2025 · similarity 0.89)
- Bench2Drive: Towards Multi-Ability Benchmarking of Closed-Loop End-To-End Autonomous Driving (NeurIPS 2024 · similarity 0.87)
- Prioritizing Perception-Guided Self-Supervision: A New Paradigm for Causal Modeling in End-to-End Autonomous Driving (NeurIPS 2025 · similarity 0.87)
- SAH-Drive: A Scenario-Aware Hybrid Planner for Closed-Loop Vehicle Trajectory Generation (ICML 2025 · similarity 0.86)
- BAT: Behavior-Aware Human-Like Trajectory Prediction for Autonomous Driving (AAAI 2024 · similarity 0.86)
- AlignDrive: Aligned Lateral-Longitudinal Planning for End-to-End Autonomous Driving (NeurIPS 2026 · similarity 0.86)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
