UCLA

Automated driving technology has made significant strides, aiming to revolutionize transportation through enhanced safety and efficiency. However, significant gaps still exist in Cooperative Driving Automation (CDA) development and testing methods. Due to the scarcity of large-scale real-world deployments, CDA research often combines simulated and physical environments to achieve the realism that simulations alone cannot provide. Experiments require complex traffic scenarios and emphasize cooperation among multiple agents, which are challenging to manage with conventional methods. As a result, typical CDA research stops after initial benchmarking improvements without fully addressing system completeness, leaving a gap between prototypes and practical implementation.To bridge this gap, this dissertation introduces a parallel development and testing framework inspired by Transportation 5.0 and the scenario engineering concepts \cite{wang2023transportation, li2022features}, designed to address these unique challenges, accelerating prototyping and validation to ensure prototypes are robust, reliable, and ready for deployment. Building upon the proposed framework, three key research projects were conducted to achieve real-world deployment of CDA systems. First, a cooperative platooning algorithm was developed and tested, enabling multi-lane platooning functions like cruising, lane changing, and adjusting platoon members across two lanes. Validated in simulation, it was later deployed on up to five Level 3-capable vehicles. Second, distributed tests were performed using the VOICES platform with four participants across the country, each equipped with various CDA tools such as traffic simulators and Level 3-capable vehicles. This allowed heterogeneous, real-time interaction to jointly enhance system performance. Lastly, an ADS regulation-aware path planning algorithm was developed and deployed on Level 3-capable vehicles. It utilizes a machine-readable regulation database to extract real-world California vehicle codes and employs a vision-language model to interpret the environment from camera inputs, integrating regulations into the planning process to select optimal future actions. In summary, this dissertation introduces a comprehensive, parallel development and testing framework for CDA, bridging the gap between prototyping and real-world deployment. By incrementally introducing risk factors and gradually validating systems through simulation and real-world testing, the framework ensures that CDA systems are robust, reliable, and ready for implementation. The successful application of this framework across key projects demonstrates its effectiveness in advancing the field of automated transportation.