SCOUT: A Lightweight Framework for Scenario Coverage Assessment in Autonomous Driving

TL;DR

SCOUT is a lightweight, scalable framework that accurately predicts autonomous driving scenario coverage using sensor representations, reducing reliance on costly human annotations and large models.

Contribution

We introduce SCOUT, a surrogate model that efficiently estimates scenario coverage from sensor data, trained via distillation to replicate large vision-language model labels.

Findings

Maintains high accuracy in scenario coverage prediction.

Reduces computational cost significantly compared to LVLM-based methods.

Enables scalable analysis for large autonomous driving datasets.

Abstract

Assessing scenario coverage is crucial for evaluating the robustness of autonomous agents, yet existing methods rely on expensive human annotations or computationally intensive Large Vision-Language Models (LVLMs). These approaches are impractical for large-scale deployment due to cost and efficiency constraints. To address these shortcomings, we propose SCOUT (Scenario Coverage Oversight and Understanding Tool), a lightweight surrogate model designed to predict scenario coverage labels directly from an agent's latent sensor representations. SCOUT is trained through a distillation process, learning to approximate LVLM-generated coverage labels while eliminating the need for continuous LVLM inference or human annotation. By leveraging precomputed perception features, SCOUT avoids redundant computations and enables fast, scalable scenario coverage estimation. We evaluate our method across a large dataset of real-life autonomous navigation scenarios, demonstrating that it maintains high accuracy while significantly reducing computational cost. Our results show that SCOUT provides an effective and practical alternative for large-scale coverage analysis. While its performance depends on the quality of LVLM-generated training labels, SCOUT represents a major step toward efficient scenario coverage oversight in autonomous systems.