Simulation to Reality: Testbeds and Architectures for Connected and Automated Vehicles

TL;DR

The paper addresses the challenge of validating connected and automated vehicle software across simulation, small-scale, and full-scale testbeds. It maps out a taxonomy of simulators and automotive middlewares (e.g., ROS 2, AUTOSAR Adaptive) and analyzes how these tools support real-time, deterministic, and secure operation across distributed E/E architectures. It derives transitioning requirements to bridge simulation, small-scale, and full-scale testing, and offers practical guidelines for selecting appropriate testbeds and orchestrating cross-scale validation. The findings highlight the trade-offs between realism, cost, and safety, and emphasize structured approaches to experiment recording and system orchestration. The work provides actionable recommendations to researchers and industry on achieving rigorous, repeatable validation for safe CAV deployment.

Abstract

Ensuring the safe and efficient operation of CAVs relies heavily on the software framework used. A software framework needs to ensure real-time properties, reliable communication, and efficient resource utilization. Furthermore, a software framework needs to enable seamless transition between testing stages, from simulation to small-scale to full-scale experiments. In this paper, we survey prominent software frameworks used for in-vehicle and inter-vehicle communication in CAVs. We analyze these frameworks regarding opportunities and challenges, such as their real-time properties and transitioning capabilities. Additionally, we delve into the tooling requirements necessary for addressing the associated challenges. We illustrate the practical implications of these challenges through case studies focusing on critical areas such as perception, motion planning, and control. Furthermore, we identify research gaps in the field, highlighting areas where further investigation is needed to advance the development and deployment of safe and efficient CAV systems.

Figures (6)

• Figure 1: The domains of this review: Testbeds, simulators and software frameworks and their overlapping domains.
• Figure 2: The role of a middleware in the development of distributed software. The middleware functions as an intermediary between the application and the operating system. In this case AUTOSAR Adaptive makes use of DDS and the operating systems network stack to communicate across the two devices.
• Figure 3: Illustration of the classifications for simulations. We found four classifications: comprehensive simulator, traffic flow simulator, E/E simulator, and component simulator. Image of comprehensive simulator from team_carla_nodate.
• Figure 4: Illustration of the classifications for small-scale testbeds. We found six classifications: platooning, V2V communication, smart cities, motion planning & control, deep learning, aggressive driving, and cloud computation.
• Figure 5: Illustration of the classifications for full-scale testbeds. We found five classifications: standalone, V2V communication, perception, validation, and control & motion planning.