Research Infrastructure: Planning and Prototyping a Community-Wide Autonomous Driving Software Testing Infrastructure

Autonomous vehicles (AVs), commonly referred to as self-driving cars, have been gaining popularity in recent years, as demonstrated by over 50 companies developing autonomous driving software (ADS). AVs are expected to bring convenience and other benefits, for example for those who cannot drive. Despite well-known companies making advances in autonomous driving, the technology is not as safe as it should be -- with widely publicized accidents from major companies in the AV industry. Over the past few years, software engineering research in the domain of ADS has yielded tools that can generate virtual scenarios for testing. However, reusing these tools is challenging as it requires sufficient experience with installation and complex interoperability issues, as well as costly hardware investments. To overcome these issues, this project aims to develop CADRI (Cloud-based Autonomous Driving software Research Infrastructure), a first-of-its-kind cloud-based ADS research infrastructure to ease software research in the context of ADS testing. CADRI will be delivered as a tool suite comprising three principal components: (1) a cataloged library of cutting-edge ADS testing approaches; (2) pre-configured ADSes and simulators on the cloud that allow new approaches to be easily evaluated; and (3) a dataset repository and benchmark that is accessible through a web-based graphical user interface. CADRI will be the first software research infrastructure of its kind with the goal of improving accessibility and reusability of ADSes, relevant simulators, and ADS testing approaches. It will provide an effective, community-wide platform for ADS research and practice through the continuous integration of diverse tools. The planning activities supported through this award include prototype library development, community planning workshops, prototype software development, and pilot studies. CADRI has the potential to improve the future of ADS research and development by making AVs safer and more comfortable. It will facilitate the discovery and adoption of cutting-edge techniques and tools that are best suited to the problems at hand. CADRI will also foster more effective university-industry collaboration. The project will result in a lower barrier to entry for researchers in the domain of ADS. Through collaborations with the Autoware Foundation, Toyota, and Zoox, this project will work toward building the infrastructure in a manner that leverages ADS testing in industry. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.