Excellence in Research: Advancing Scalability of Connected and Autonomous Vehicles via Bottom-up Hardware-Software Co-design

As self-driving vehicles become more common on public roads, they must sense their surroundings quickly and accurately. A single vehicle cannot always see enough on its own, because its view can be blocked by other vehicles or limited when objects are far away. Connected vehicles can help one another by sharing what their sensors detect and by relying on small roadside computers and sensors to do some of the work. This sharing improves safety, but it becomes difficult as the number of vehicles grows: the wireless links and the roadside computers become overloaded, and shared information can arrive too late to be useful. This project develops a way to keep connected self-driving vehicle systems working reliably as more vehicles join the road, without the very high cost of building large amounts of new roadside computing everywhere. The approach pairs smarter decisions on each vehicle, about when to rely on roadside help, with small low-cost hardware add-ons placed only where the network is most strained. The results can lower the cost of deploying self-driving technology, make roads safer as automated vehicles become widespread, and extend to other settings that connect many devices to nearby computing, such as smart healthcare and connected homes. The project will also support student research, curriculum development, and outreach activities for K-12 students and educators. Data, software, and simulation tools will be made publicly available. The technical work is a bottom-up co-design of the vehicles and the roadside edge infrastructure, carried out in three coordinated thrusts and evaluated on a connected-vehicle testbed in Chattanooga, Tennessee. The first thrust will design a strategy by which each vehicle decides, in coordination with nearby vehicles, when to send a perception task to the edge and when to run it locally. It will keep each decision tractable as vehicle numbers grow, estimate the edge’s unknown resource-allocation behavior, and stay robust when that estimate is uncertain. The second thrust will design a lightweight hardware extension for the edge, based on data processing units built from programmable network cards and fast storage. These units will be placed only at bottlenecks, adding computing and data transfer capacity at low cost and small physical footprint rather than upgrading the entire infrastructure. The third thrust will develop a mobility-aware framework that routes tasks and data across the varied edge hardware as vehicles move. The methods will be examined in three stages: a small-scale pilot to collect operational data, a larger simulation study driven by that data, and a field investigation with instrumented vehicles and roadside units. The expected outcome is a general design framework for reliable and cost-efficient collaboration between mobile intelligent systems and edge infrastructure. This project is funded by the U.S. National Science Foundation (NSF) Historically Black Colleges and Universities-Excellence in Research (HBCU-EIR) Program. HBCU-EIR supports research at public and private HBCUs to strengthen their research capacity and promote engagement with NSF. 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.