CARLOS: An Open, Modular, and Scalable Simulation Framework for the Development and Testing of Software for C-ITS

Geller, Christian; Haas, Benedikt; Kloeker, Amarin; Hermens, Jona; Lampe, Bastian; Beemelmanns, Till; Eckstein, Lutz · 2024 · OpenAlex-citations

DOI: 10.1109/iv55156.2024.10588502

archive: archived pipeline: cataloged verified

Get this paper ↗ (DOI — opens at the source; we link to it, we don't host it)

Summary

This paper introduces CARLOS, an open-source, modular, and scalable simulation framework designed for the development and testing of software for Cooperative Intelligent Transport Systems (C-ITS). The research is motivated by the increasing complexity of automated driving systems and the need for efficient, automated testing procedures. Existing simulation solutions are often tailored to specific developer needs, lack modularity, or are not open-source, limiting their extensibility and community adoption. To address these gaps, the authors propose a generic architecture aligned with modern microservice and DevOps principles, specifically leveraging containerization and continuous integration to enhance maintainability, scalability, and interoperability. The proposed architecture is structured into distinct layers: a simulation layer (containing the simulator core, communication interfaces, and control actors), a storage layer for persistent data, an application layer for user interaction, and an orchestration layer for managing concurrent simulations. CARLOS serves as a reference implementation of this architecture, built upon the CARLA simulator and the Robot Operating System (ROS) ecosystem. The framework utilizes Docker for containerization, enabling the independent deployment and updating of components. Key components include a ROS bridge for data exchange between CARLA and ROS topics, and a Scenario Runner compliant with the OpenSCENARIO standard for executing scenario-based tests. The authors also developed a data generation pipeline that allows for the systematic simulation of parameter permutations via JSON configuration, facilitating large-scale data collection for machine learning and reinforcement learning applications. The paper evaluates CARLOS through three primary use cases: software prototyping, data-driven development, and automated testing. In software prototyping, the framework supports both open-loop and closed-loop tests with varying levels of fidelity. For data-driven development, the automated pipeline enables cost-effective generation of large datasets by sampling parameters across a wide space. In automated testing, CARLOS integrates with continuous integration pipelines to systematically evaluate numerous test configurations against standardized scenario databases. A qualitative comparison between CARLOS and the native CARLA ecosystem demonstrates that CARLOS offers superior usability, maintainability, scalability, and interoperability. Specifically, the modular containerized structure allows for easier customization and independent updates of components, while the orchestration layer enables efficient parallel execution of simulations, significantly improving testing throughput. The significance of this work lies in providing a robust, community-accessible foundation for simulative testing in automated driving. By making CARLOS and its example implementations publicly available on GitHub, the authors aim to foster further development driven by the CARLA and ROS communities. The framework addresses critical industry challenges by offering a flexible architecture that supports the entire software development life cycle, from initial prototyping to rigorous safety assurance. This approach reduces the reliance on expensive and risky real-world testing, enabling developers to conduct extensive, reproducible, and scalable simulations in a controlled environment. The adoption of standard interfaces like ROS and ASAM OpenX ensures compatibility with existing tools and future advancements in C-ITS software development.

Provenance

The full processing record for this entry. Every stage of this paper's journey through the pipeline is logged — what ran, with which tool and model, how many attempts it took, and when it last completed.

Summary generated by qwen3.6-27b-prismaquant on 2026-06-26; verification: verified.

Topics

Ranked by relevance to this paper. Hover a topic for its definition.

• driverless ads

Information type

What kind of knowledge this paper contributes, grouped by family — independent of topic (what it is about) and method (how it was studied).

• Methodological Resource: tool software