BridgeSim: Unveiling the OL-CL Gap in End-to-End Autonomous Driving

2026-04-12

AI Summary

This work addresses the performance degradation commonly observed when open-loop pre-trained policies are deployed in closed-loop settings, primarily due to observation domain shift and objective mismatch. Through systematic analysis, the study identifies objective mismatch as the dominant factor underlying the open-loop–closed-loop performance gap and highlights critical blind spots in standard open-loop evaluation protocols. To mitigate this issue, the authors propose the first test-time adaptation (TTA) framework specifically designed to alleviate planning bias, incorporating observation calibration, Q-value debiasing, and temporal consistency constraints. Notably, this approach enables effective closed-loop adaptation without requiring additional training. Experimental results demonstrate that the method substantially improves closed-loop transfer performance and dynamic generalization capabilities, outperforming existing baselines.

Technology Category

• Multiagent Systems: Adversarial Agents
• Planning, Routing, and Scheduling: Replanning and Plan Repair
• Machine Learning: Transfer, Domain Adaptation, Multi-Task Learning

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Economics, Online Markets and Human Computation: Cost models of using LLMs in production systems
• Responsible Web: Machine-in-the-loop, human agency and autonomy

Abstract

Open-loop (OL) to closed-loop (CL) gap (OL-CL gap) exists when OL-pretrained policies scoring high in OL evaluations fail to transfer effectively in closed-loop (CL) deployment. In this paper, we unveil the root causes of this systemic failure and propose a practical remedy. Specifically, we demonstrate that OL policies suffer from Observational Domain Shift and Objective Mismatch. We show that while the former is largely recoverable with adaptation techniques, the latter creates a structural inability to model complex reactive behaviors, which forms the primary OL-CL gap. We find that a wide range of OL policies learn a biased Q-value estimator that neglects both the reactive nature of CL simulations and the temporal awareness needed to reduce compounding errors. To this end, we propose a Test-Time Adaptation (TTA) framework that calibrates observational shift, reduces state-action biases, and enforces temporal consistency. Extensive experiments show that TTA effectively mitigates planning biases and yields superior scaling dynamics than its baseline counterparts. Furthermore, our analysis highlights the existence of blind spots in standard OL evaluation protocols that fail to capture the realities of closed-loop deployment.

Problem

• OL-CL gap
• autonomous driving
• open-loop evaluation
• closed-loop deployment
• domain shift

Innovation

• Test-Time Adaptation
• OL-CL Gap
• Objective Mismatch
• Temporal Consistency
• Autonomous Driving