# LawBreaker: An Approach for Specifying Traffic Laws and Fuzzing Autonomous Vehicles

Source: [LawBreaker: An Approach for Specifying Traffic Laws and Fuzzing Autonomous Vehicles](<https://www.sota2.com/research/paper/68fa6e5c1e209131cea3f5c0>)

Collection index: [llms.txt](<https://www.sota2.com/research/llms.txt>)

Data updated: 2025-10-26T00:56:03.105Z

Authors: Yang Sun, Christopher M. Poskitt, Jun Sun, Yuqi Chen, Zijiang Yang

Published: 2022-08-31T07:03:03.000Z

## Abstract

Autonomous driving systems (ADSs) must be tested thoroughly before they can be deployed in autonomous vehicles. High-fidelity simulators allow them to be tested against diverse scenarios, including those that are difficult to recreate in real-world testing grounds. While previous approaches have shown that test cases can be generated automatically, they tend to focus on weak oracles (e.g. reaching the destination without collisions) without assessing whether the journey itself was undertaken safely and satisfied the law. In this work, we propose LawBreaker, an automated framework for testing ADSs against real-world traffic laws, which is designed to be compatible with different scenario description languages. LawBreaker provides a rich driver-oriented specification language for describing traffic laws, and a fuzzing engine that searches for different ways of violating them by maximising specification coverage. To evaluate our approach, we implemented it for Apollo+LGSVL and specified the traffic laws of China. LawBreaker was able to find 14 violations of these laws, including 173 test cases that caused accidents.

## Recorded benchmark results

A research result applies to its stated dataset, split, metric and evaluation configuration. Compare like-for-like settings, preserve the metric's unit and direction, and distinguish paper-reported results from independent reproduction. The best result recorded here is not a claim of exhaustive, current global coverage.

- [Traffic Law Violation Coverage on Chinese traffic law specification 82 violation goals Apollo ADS Derived from LawBreaker](<https://www.sota2.com/research/sota/traffic-law-violation-coverage-on-chinese-traffic-law-specification-82-violation-goals-apollo-ads-derived-from-lawbreaker.md>): Traffic Law Violation Coverage; Chinese traffic law specification 82 violation goals Apollo ADS Derived from LawBreaker; R1 Coverage Rate: 23.

- [Violation Scenario Generation on Scenario S1](<https://www.sota2.com/research/sota/violation-scenario-generation-on-scenario-s1.md>): Violation Scenario Generation; Scenario S1; Mean Score: 4.9.

- [Violation Scenario Generation on Scenario S2](<https://www.sota2.com/research/sota/violation-scenario-generation-on-scenario-s2.md>): Violation Scenario Generation; Scenario S2; Mean Violation: 3.44.

- [Violation Scenario Generation on Scenario S3](<https://www.sota2.com/research/sota/violation-scenario-generation-on-scenario-s3.md>): Violation Scenario Generation; Scenario S3; Mean Score: 4.11.

- [Violation Scenario Generation on Scenario S4](<https://www.sota2.com/research/sota/violation-scenario-generation-on-scenario-s4.md>): Violation Scenario Generation; Scenario S4; Mean Violation: 6.78.

- [Scenario Generation on Scenario S1](<https://www.sota2.com/research/sota/scenario-generation-on-scenario-s1.md>): Scenario Generation; Scenario S1; Inference Time (s): 37.

- [Scenario Generation on Scenario S4](<https://www.sota2.com/research/sota/scenario-generation-on-scenario-s4.md>): Scenario Generation; Scenario S4; Inference Time (s): 46.

- [Scenario Generation on Scenario S2](<https://www.sota2.com/research/sota/scenario-generation-on-scenario-s2.md>): Scenario Generation; Scenario S2; Inference Time (s): 52.

- [Scenario Generation on Scenario S3](<https://www.sota2.com/research/sota/scenario-generation-on-scenario-s3.md>): Scenario Generation; Scenario S3; Inference Time (s): 62.

## Original sources and implementation

[Original paper](<https://arxiv.org/abs/2208.14656>)

[PDF](<https://arxiv.org/pdf/2208.14656.pdf>)