The global deployment of the 5G infrastructure is underway. This upcoming technology is envisioned to support billions of mobile users today and trillions of connected things tomorrow. Robotic applications are a new class of emerging applications for 5G, yet posing stringent requirements on low latency, high reliability, and high speed. In this project, we propose to innovate the 5G technology in its software stack (in the form of protocols) from the device side rather than inside the infrastructure. We use showcase robotic applications to drive our design effort. We challenge the conventional wisdom that wireless access is the sole roadblock for low latency and high reliability. Instead, we take the device perspective on 5G software innovation, which poses as an equally critical pathway from our initial results. We seek to achieve three goals. First, we refine the software stack of 5G for low latency within the standards from the device side. Second, we ensure high reliability through device-centric software techniques on the 5G system. Third, we provision 5G devices with new capabilities of both learning and reasoning for both common usage settings and failure-prone scenarios. If successful, it may help to advance robotic applications in our daily life and the society. The proposal further helps to train and prepare the next-generation workforce for 5G technology.
Part II:
The proposed research has three key areas of technical contributions. First, we use the robotic applications to drive our device-centric 5G innovations. It quantifies the requirements on low latency and high reliability, makes a case for software-defined reliability and latency tailored to applications, and offers benchmarks for assessing the design. Second, we take the device-centric approach to high reliability and low latency. Our solution leverages the state replica at the design and optimizes 5G latency on both control and data planes. To improve reliability, we exploit both concurrent network access (via connectivity to two base stations and multiple concurrent wireless channels to each) and detect-and-react for higher reliability. We devise novel proactive failure masking and reactive recovery components, by leveraging existing 5G mechanisms and recent checkpointing schemes. We further adjust various tuning knobs to the application's reliability requirement. Third, we provide new learning and reasoning functions at the device to open up the closed 5G network operations and enable the applications for further adaptations.
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.
