Issue Categorization and Analysis of an Open-Source Driving Assistant System

Autonomous driving system (ADS) has attracted great much attention from both academia and industry in recent years. Since these systems are safety-critical, assurance of their safety and reliability is of great significance. Research efforts have been paid to Level-4 ADS systems to understand their safety concerns and vulnerabilities; however, no progress has been made in Level-2 systems, though they have been deployed more widely. In this work, we focus on an open-source Level-2 driver assistant system, namely, OPENPILOT, and perform an empirical study on the issues raised by developers and users in the developers' communities. We first overview and introduce the logical architecture of OPENPILOT; then, we present our methodologies of collecting pull requests and issues from two developers' communities; as a result, we collect 1293 pull requests, 694 issues, and then we classify them into 5 categories; lastly, we discuss on the strengths and weaknesses of OPENPILOT and the future directions, based on the collected issues. Our work is the first attempt to perform a comprehensive study on the issue analysis for OPENPILOT, and it also motivates more future studies on the systematic testing and analysis of these systems.