No driver, No Regulation? --Online Legal Driving Behavior Monitoring for Self-driving Vehicles

TL;DR

This work tackles the challenge of converting fuzzy, human-oriented traffic laws into precise, machine-usable specifications for autonomous vehicles by introducing an ego-vehicle online monitor built on a trigger-domain layered architecture and metric temporal logic ($MTL$). It decomposes law articles into atomic propositions, expands regulatory coverage to over 90 clauses, and derives data-driven thresholds for online monitoring, validating the approach on CH highway (AD4CHE) and intersection (SIND) datasets. The key contributions are a practical ego-vehicle online monitoring framework, a detailed library of atomic propositions with $MTL$ expressions for highway and intersection laws, and a data-driven method for threshold computation with empirical verification. The approach yields concrete insight into common highway and intersection violations and provides a principled basis for compliance decision-making in autonomous driving, informing manufacturers and regulators about formalized traffic-law specifications. Overall, the paper provides a systematic, machine-readable pathway to enforce traffic-law compliance in AVs and to underpin law-compliance decision support in real-world deployments.

Abstract

Defined traffic laws must be respected by all vehicles. However, it is essential to know which behaviors violate the current laws, especially when a responsibility issue is involved in an accident. This brings challenges of digitizing human-driver-oriented traffic laws and monitoring vehicles' behaviors continuously. To address these challenges, this paper aims to digitize traffic law comprehensively and provide an application for online monitoring of legal driving behavior for autonomous vehicles. This paper introduces a layered trigger domain-based traffic law digitization architecture with digitization-classified discussions and detailed atomic propositions for online monitoring. The principal laws on a highway and at an intersection are taken as examples, and the corresponding logic and atomic propositions are introduced in detail. Finally, the digitized traffic laws are verified on the Chinese highway and intersection datasets, and defined thresholds are further discussed according to the driving behaviors in the considered dataset. This study can help manufacturers and the government in defining specifications and laws and can also be used as a useful reference in traffic laws compliance decision-making. Source code is available on https://github.com/SOTIF-AVLab/DOTL.

Figures (11)

• Figure 1: Online traffic law violation monitoring of autonomous vehicle.This monitor can be deployed on the ego vehicle and monitoring the traffic law violation behaviors of the ego vehicle in both highway and intersection scenarios. When deployed on every vehicle in the dataset, the statistical chart of monitoring results are shown as the two figure at the left lower corner.
• Figure 2: The classifications of Chinese traffic laws restrictions. Each circle means a different class of traffic law constraint. The size of the circle represents the proportion of this class, and the overlap means both constraint classes appear in the same article. The solid edge linetype of the circle meanings there is no ambiguous expression involved, while dash dot line means in certain articles fuzzy parameter is involved and the dot line means the corresponding expressions in each article need to clarify.
• Figure 3: The architecture of the proposed trigger domain-based layered traffic law digitization method
• Figure 4: The overall process of traffic law digitization
• Figure 5: Digital complexity of regulations