The automated driving ecosystem is considered as a highly multi-disciplinary realm requiring a major design shift, from a single-body car-based platform which encompasses entire competencies within a car, to a networked multi-object system where competencies interact among cars and surrounding resources via the state-of-the-art computing and communication technologies. Such an evolution is stymied by complexity challenges in hardware, system and service models while supporting safety, bandwidth, and service differentiation. This project develops a new service model and a hardware/software approach, Omni, that spans cars and edge computing platforms to provide end-to-end fine-grained coordination of communication and computation, and thus meets the stringent safety and performance requirement of future connected driving automation systems. The outcomes can benefit many other application domains, such as Augmented Reality/Virtual Reality and industrial Internet of Things. This project designs new curricula and contributes to society by engaging under-represented groups, developing research tools/benchmarks dissemination for education and training, as well as technology transfer to industries.
This project develops a synergistic research program that rethinks the system hardware/software stacks and service models to enable the connected driving automation system to allocate communication and computation resources seamlessly. First of all, this project develops a cross-platform uniform computing framework called Omni that addresses the cross-platform interoperability between the communication functions and computation functions, and enables the platform independence of application deployment. Based on the uniform computing framework, an intelligent service slice-provisioning scheme enables Omni to automatically provision service slices according to users’ performance requests. The service model is responsible for inferring the performance implications and organizing the specific integrated hardware-software stacks across the car/edge platforms for the requested service slice. The uniform computing framework and service model is developed upon a series of hardware and system support techniques, such as hardware accelerator designs with service slice implication, a heterogeneity-/resources-aware runtime, and a middleware layer to enable the vehicle-edge cooperation for flexible service deployment and task processing.
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.
