NaviHydra: Controllable Navigation-guided End-to-end Autonomous Driving with Hydra-distillation

2025-12-11

AI Summary

End-to-end autonomous driving models often struggle to simultaneously satisfy high-level navigation command compliance and trajectory safety under dynamic environmental conditions. To address this, we propose a controllable end-to-end framework comprising four key components: (1) a novel metric for quantifying navigation compliance; (2) a Hydra distillation mechanism leveraging rule-based simulators to efficiently transfer safety priors into neural networks; (3) BEV-based trajectory representation enhanced with instruction-conditioned modeling to improve fine-grained intent understanding; and (4) a controllability-oriented evaluation protocol. Evaluated on the NAVSIM benchmark, our approach achieves state-of-the-art performance—improving navigation compliance by +12.7% and reducing collision rate by 38.5%—demonstrating effective co-optimization of command controllability and safety in complex, dynamic scenarios.

Technology Category

• Humans and AI: Human-Aware Planning and Behavior Prediction
• Natural Language Processing: Safety and Robustness
• Philosophy and Ethics of AI: Safety, Robustness & Trustworthiness

Application Category

• Responsible Web: Machine-in-the-loop, human agency and autonomy
• Security and Privacy: Browser and end-point security
• Search and Retrieval-Augmented AI: Web evaluation methodologies and metrics

Abstract

The complexity of autonomous driving scenarios requires robust models that can interpret high-level navigation commands and generate safe trajectories. While traditional rule-based systems can react to these commands, they often struggle in dynamic environments, and end-to-end methods face challenges in complying with explicit navigation commands. To address this, we present NaviHydra, a controllable navigation-guided end-to-end model distilled from an existing rule-based simulator. Our framework accepts high-level navigation commands as control signals, generating trajectories that align with specified intentions. We utilize a Bird's Eye View (BEV) based trajectory gathering method to enhance the trajectory feature extraction. Additionally, we introduce a novel navigation compliance metric to evaluate adherence to intended route, improving controllability and navigation safety. To comprehensively assess our model's controllability, we design a test that evaluates its response to various navigation commands. Our method significantly outperforms baseline models, achieving state-of-the-art results in the NAVSIM benchmark, demonstrating its effectiveness in advancing autonomous driving.

Problem

• Develops a controllable navigation-guided autonomous driving model
• Enhances trajectory feature extraction using Bird's Eye View
• Introduces a metric to evaluate navigation command compliance

Innovation

• Controllable navigation-guided end-to-end model distilled from simulator
• Bird's Eye View based trajectory gathering enhances feature extraction
• Novel navigation compliance metric improves controllability and safety