ADDT -- A Digital Twin Framework for Proactive Safety Validation in Autonomous Driving Systems

TL;DR

The paper tackles safety-critical failures in autonomous driving caused by rare corner cases and opaque end-to-end models. It presents ADDT, a high-fidelity digital twin platform that combines realistic environments, vehicle dynamics, sensor behavior, and fault injection to proactively identify hidden faults and validate safety before deployment. Key contributions include a digital twin–driven design automation paradigm, comprehensive sensor and compute fault injection with real-time latency profiling, hardware-in-the-loop validation, and an open-source release to accelerate industry adoption. Results demonstrate fidelity to real-world latency distributions (KL divergence $D_{KL}=0.1$) and quantify fault impacts across sensor and compute domains, while delivering substantial cost and time savings over physical road testing. The framework thus enables scalable, proactive safety validation for autonomous driving and other embodied AI systems, with broad practical impact for safer deployment and development efficiency.

Abstract

Autonomous driving systems continue to face safety-critical failures, often triggered by rare and unpredictable corner cases that evade conventional testing. We present the Autonomous Driving Digital Twin (ADDT) framework, a high-fidelity simulation platform designed to proactively identify hidden faults, evaluate real-time performance, and validate safety before deployment. ADDT combines realistic digital models of driving environments, vehicle dynamics, sensor behavior, and fault conditions to enable scalable, scenario-rich stress-testing under diverse and adverse conditions. It supports adaptive exploration of edge cases using reinforcement-driven techniques, uncovering failure modes that physical road testing often misses. By shifting from reactive debugging to proactive simulation-driven validation, ADDT enables a more rigorous and transparent approach to autonomous vehicle safety engineering. To accelerate adoption and facilitate industry-wide safety improvements, the entire ADDT framework has been released as open-source software, providing developers with an accessible and extensible tool for comprehensive safety testing at scale.

Figures (11)

• Figure 1: Key Challenges in Autonomous Driving System Development.
• Figure 2: Driving scenarios for evaluation.
• Figure 3: Latency Distribution Comparison between Real-World and Simulated Data.
• Figure 4: Latency variation under different object counts in driving scenarios.
• Figure 5: The relationship between camera orientation shift and the positional and orientational errors of detected objects.