Guiding End-to-End Driving Models with Endpoint-Constrained Trajectory Optimization

📄 arXiv: 2609.31383v1 📥 PDF

作者: Brayden Zhang, Mahsa Golchoubian, Igor Gilitschenski, Boris Ivanovic, Kashyap Chitta

分类: cs.RO, cs.AI, cs.CV, cs.LG

发布日期: 2026-09-25

💡 一句话要点

提出端点约束优化以解决驾驶模型闭环性能问题

🎯 匹配领域: 支柱一：机器人控制 (Robot Control) 支柱二：RL算法与架构 (RL & Architecture)

关键词: 端到端驾驶 闭环控制 轨迹优化 自动驾驶 机器学习

📋 核心要点

  1. 现有的端到端驾驶模型在训练时采用开放式行为克隆，但在实际应用中需要闭环控制，导致性能不一致。
  2. 本文提出端点约束优化（ECO），通过锚定轨迹到历史执行，重塑中间航点以提高可行性，解决了中间轨迹不一致的问题。
  3. 在HUGSIM和AlpaSim的实验中，ECO显著提升了多种驾驶策略的闭环评分，VaVAM的HD-Score从18.1提升至31.0，提升幅度达71%。

📝 摘要（中文）

端到端驾驶策略通常通过开放式行为克隆进行训练，但在实际部署时必须在闭环中操作，这导致训练与执行之间存在根本不匹配。除了常见的协变量偏移和因果混淆外，本文还识别出一个补充因素：基于航点的监督和位移度量并不能确保中间轨迹在物理上是一致的，或易于控制器跟踪。基于这一观察，本文提出了一种轻量级的后处理层——端点约束优化（ECO），它将轨迹锚定到车辆的执行历史，保持策略的预测端点，并重塑中间航点以提高可行性。ECO不需要地图、特权模拟器状态或额外训练，可以插入到各种航点发射策略和其控制器之间。实验结果表明，ECO显著提高了闭环性能。

🔬 方法详解

问题定义：本文旨在解决端到端驾驶模型在闭环控制中表现不佳的问题，现有方法在中间轨迹的物理一致性和可跟踪性上存在不足。

核心思路：提出端点约束优化（ECO），通过保持策略的预测端点并重塑中间航点，确保轨迹在物理上可行且易于控制器跟踪。

技术框架：ECO作为一个轻量级的后处理层，可以插入到任何生成或回归基础的驾驶策略与其控制器之间，主要模块包括轨迹锚定和中间航点重塑。

关键创新：ECO的创新在于不依赖地图或额外训练，通过简单的后处理显著改善了闭环性能，解决了中间轨迹不一致的问题。

关键设计：ECO的设计中，关键参数包括航点的重塑算法和与历史执行的锚定策略，确保中间轨迹在物理上可行且符合控制器的要求。

🖼️ 关键图片

fig_0
img_1
img_2

📊 实验亮点

在HUGSIM中，ECO将VaVAM的HD-Score从18.1提升至31.0，提升幅度达71%，并在HUGSIM闭环驾驶挑战中获得第一名。在AlpaSim中，ECO分别提升了VaVAM和DiffusionDrive的场景评分，增幅为123%和22%。

🎯 应用场景

该研究具有广泛的应用潜力，尤其在自动驾驶和智能交通系统中。通过提高端到端驾驶模型的闭环性能，ECO可以帮助实现更安全和高效的自动驾驶解决方案，推动智能交通技术的发展。

📄 摘要（原文）

End-to-end driving policies are commonly trained through open-loop behavior cloning, yet they must ultimately operate in closed-loop when deployed on a vehicle, creating a fundamental mismatch between training and execution. Beyond the commonly studied effects of covariate shift and causal confusion, we identify a complementary factor for this open-loop/closed-loop gap: waypoint-based supervision and displacement metrics do not ensure that the intermediate trajectory is physically coherent or easy for the controller to track. We observe that these inconsistencies concentrate primarily at intermediate waypoints, while the predicted endpoint remains comparatively reliable. Based on this observation, we introduce Endpoint-Constrained Optimization (ECO), a lightweight postprocessing layer that anchors the trajectory to the vehicle's executed history, preserves the policy's predicted endpoint, and reshapes the intermediate waypoints to improve feasibility. ECO requires no map, privileged simulator state, or additional training, and can be inserted between a broad range of waypoint-emitting policies and their controllers. Across two closed-loop simulators, it improves the aggregate closed-loop score of all six evaluated generative and regression-based driving policies, and the gains tend to increase with how often the base plans violate motion limits. On HUGSIM, ECO improves VaVAM from 18.1 to 31.0 HD-Score (+71%), achieving 1st place on the HUGSIM Closed-Loop Driving Challenge. Similarly, on AlpaSim, ECO increases the scene scores of VaVAM and DiffusionDrive by 123% and 22%, respectively. These results show that for a broad collection of end-to-end driving models, repairing the intermediate geometry of predicted trajectories without changing the policy's predicted endpoint can substantially improve closed-loop performance.