ROS 2-Based Architecture for Autonomous Driving Systems: Design and Implementation.

Area of Science:

• Robotics
• Software Engineering
• Autonomous Systems

Background:

• Growing interest in autonomous vehicle (AV) adoption necessitates advanced software architectures.
• Current AV development faces challenges in meeting real-time, safety, and scalability demands with heterogeneous systems.

Purpose of the Study:

• To present a lightweight, modular, and scalable software architecture for autonomous vehicles.
• To leverage Service-Oriented Architecture (SOA) principles and ROS 2 for robust AV systems.

Main Methods:

• Implemented a Service-Oriented Architecture (SOA) using ROS 2, focusing on a modular Perception, Planning, and Control layered design.
• Utilized ROS 2's Data Distribution System (DDS) Quality-of-Service (QoS) for reliable communication and fault containment.
• Employed an event-driven decision-making core with zero-order hold for efficient, non-synchronized updates.

Main Results:

• Validated the architecture on a 1:10 scale autonomous vehicle in diverse urban scenarios.
• The Perception-Planning pipeline consistently met worst-case real-time deadlines, demonstrating deterministic behavior under stress.
• Achieved compliance with real-time application standards for the tested AV use case.

Conclusions:

• The proposed ROS 2-based architecture provides a robust foundation for developing and deploying real-time autonomous vehicle systems.
• The modular and event-driven design effectively addresses the complexities of integrating heterogeneous hardware and software in AVs.
• This architecture offers a scalable solution for future advancements in autonomous driving technology.