First-principles Based 3D Virtual Simulation Testing for Discovering SOTIF Corner Cases of Autonomous Driving

TL;DR

The paper tackles SOTIF safety testing for autonomous driving by addressing two key gaps: high-fidelity sensor–environment modeling under adverse weather and efficient discovery of corner cases in a large, dynamic scenario space. It introduces AutoSVT, a framework that embeds first-principles LiDAR fog modeling into CARLA, couples ADS and simulator with an eBPF-based synchronization bridge (BPFRTI), and uses simulated annealing to steer scenario seeds and mutations toward safety-critical fog-related corner cases. Empirical results show the approach discovers roughly four times as many corner cases as the state-of-the-art DriveFuzz under identical settings, particularly highlighting fog-induced perception and planning failures. The work yields a practical, extensible platform for comprehensive SOTIF testing and safety assessment, with open-source tooling and case studies across Apollo OpenSource and CARLA Leaderboard scenarios.

Abstract

3D virtual simulation, which generates diversified test scenarios and tests full-stack of Autonomous Driving Systems (ADSes) modules dynamically as a whole, is a promising approach for Safety of The Intended Functionality (SOTIF) ADS testing. However, as different configurations of a test scenario will affect the sensor perceptions and environment interaction, e.g. light pulses emitted by the LiDAR sensor will undergo backscattering and attenuation, which is usually overlooked by existing works, leading to false positives or wrong results. Moreover, the input space of an ADS is extremely large, with infinite number of possible initial scenarios and mutations, along both temporal and spatial domains. This paper proposes a first-principles based sensor modeling and environment interaction scheme, and integrates it into CARLA simulator. With this scheme, a long-overlooked category of adverse weather related corner cases are discovered, along with their root causes. Moreover, a meta-heuristic algorithm is designed based on several empirical insights, which guide both seed scenarios and mutations, significantly reducing the search dimensions of scenarios and enhancing the efficiency of corner case identification. Experimental results show that under identical simulation setups, our algorithm discovers about four times as many corner cases as compared to state-of-the-art work.

Figures (10)

• Figure 1: Overview of AutoSVT.
• Figure 2: The procedure of first-principles based LiDAR simulation in foggy weather.
• Figure 3: Workflow of BPFRTI.
• Figure 4: Throttle control commands and feedback in Apollo with and without synchronization.
• Figure 5: Point clouds in junction under various fog densities.