Open-Source Autonomous Driving Software Platforms: Comparison of Autoware and Apollo

TL;DR

The paper analyzes open-source autonomous driving platforms Autoware and Apollo, addressing the challenge of evaluating full-stack systems by focusing on core modules (localization, perception, planning, control) and middleware performance. It provides a structured, quantitative comparison of architectures and subcomponents, and compares two middleware approaches—ROS2/FAST-DDS versus CyberRT/shared memory—across latency, CPU, memory, and reliability. Key findings show Autoware offers broader localization options while Apollo emphasizes camera-based perception, with distinct trade-offs in planning and control architectures. The results offer practical guidance for researchers and engineers to select and configure platforms for scalable, industrial-grade autonomous driving development.

Abstract

Full-stack autonomous driving system spans diverse technological domains-including perception, planning, and control-that each require in-depth research. Moreover, validating such technologies of the system necessitates extensive supporting infrastructure, from simulators and sensors to high-definition maps. These complexities with barrier to entry pose substantial limitations for individual developers and research groups. Recently, open-source autonomous driving software platforms have emerged to address this challenge by providing autonomous driving technologies and practical supporting infrastructure for implementing and evaluating autonomous driving functionalities. Among the prominent open-source platforms, Autoware and Apollo are frequently adopted in both academia and industry. While previous studies have assessed each platform independently, few have offered a quantitative and detailed head-to-head comparison of their capabilities. In this paper, we systematically examine the core modules of Autoware and Apollo and evaluate their middleware performance to highlight key differences. These insights serve as a practical reference for researchers and engineers, guiding them in selecting the most suitable platform for their specific development environments and advancing the field of full-stack autonomous driving system.

Figures (9)

• Figure 1: Classification of autonomous driving software platforms.
• Figure 2: Architecture of autonomous driving software platforms. Core functionality includes middleware for data communication between modules. The core modules perform essential functions required for full-stack autonomous driving, such as localization, perception, planning, and control. The infrastructure consists of sensors, maps, and simulators necessary for operating these modules.
• Figure 3: Architecture of software platforms.
• Figure 4: Data flow of the middleware used in Autoware and Apollo. Raw data is transmitted from the core perception module. In Autoware, data is fragmented through serialization between the publisher socket and the subscriber socket. In contrast, Apollo transmits raw data directly to shared memory without serialization, allowing the core planning modules to access the raw data directly from shared memory.
• Figure 5: Internal structure of the core localization module used in Autoware and Apollo. The module estimates position using IMU, GNSS, and LiDAR. In LiDAR localization, a pre-built point cloud map is utilized to determine the vehicle's position. The position estimates obtained from each sensor are integrated using the Kalman Filter technique to enhance accuracy.