As safety standards for autonomous driving systems continue to rise, the need for rigorous validation methods has become increasingly urgent. This paper combines three software engineering approaches for ensuring the safety and reliability of autonomous driving systems. The first approach evaluates system’s performance through scripted simulations in a co-simulation platform, mimicking real-world conditions. The second approach focuses on generating monitors to identify faults, increasing confidence in the system’s correctness and safety. The third approach includes continuous integration, enabling automatic testing throughout the development process.
To achieve autonomous driving testing and monitoring, we propose integrating the co-simulation and monitoring pipeline into a testing framework. This framework allows us to continuously test the system’s behavior in various scenarios, ensuring the safety and reliability of the autonomous driving system before deployment. We demonstrate the effectiveness of our approach through the STRIVE platform, which provides a solution for testing and monitoring autonomous driving systems. Our results show that this approach can help developers build safe and reliable autonomous driving systems, addressing a critical need in the industry from the first day of the system’s development.
Abstract
Conclusion and Future Work
STRIVE enables thorough testing of ADAS/AD systems by integrating CARLA and Artery simulators, forming a co-simulation for validation of AD behavior as well as V2X communication. Runtime monitors allows continuous observation of V2X communication and detects anomalies, while container-based deployment increases flexibility and scalability while saving time in offline deployments. It is critical for cost reduction to address inefficiencies and optimize resource consumption. As shown in Figure 5, the system takes an additional 5 seconds to complete a test run after almost finishing (approximately 10 seconds). This overhead must be minimized to improve efficiency. V2X monitors are faster than traffic rule monitors due to the intrinsic complexity of properties that combine time and space. As future work, we intend to reduce the overhead of data serialization between containers and assist runtime monitors with hardware accelerators.
Acknowledgment
This work is supported by the European Union/Next Generation EU, through Programa de Recuperacão e Resiliência (PRR) [Project Route 25 with Nr. C645463824-00000063].
