A Framework for Autonomous Vehicle Testing Using Semantic Models

Publication Date

1-1-2023

Document Type

Conference Proceeding

Publication Title

Proceedings - 5th IEEE International Conference on Artificial Intelligence Testing, AITest 2023

DOI

10.1109/AITest58265.2023.00020

First Page

66

Last Page

73

Abstract

With the rapid development of big data and artificial intelligence, autonomous vehicles (AV) have achieved great success in diverse application domains. Although the current development of autonomous driving technology has been greatly improved, the accident rate of driverless vehicles still rises. Therefore, how to ensure the quality of AVs using different testing technologies has become an important issue. Traditional testing approaches such as decision tables and state charts struggle with diversified test inputs. Hence, there is a strong demand for new test models for AV testing. In this paper, we present a model-based testing framework that utilizes semantic trees and 3-dimensional (3D) test tables to model driving scenarios into individual test cases. Using the proposed framework, we generate fine-grained examples corresponding to the real world. Also, we evaluate the behavior of the Apollo AV system in 30 scenarios in SVL simulator. The results show that the scenarios generated by our framework can cover most conditions occurring, thus addressing the challenge of testing AVs for safety and reliability.

Keywords

autonomous vehicle testing, black-box test, model-based testing, test scenario generation

Department

Computer Engineering

Link to Full Text

Share

COinS