Overview

In an increasingly connected and ”smart” world, demands for network bandwidth and computing re- sources grow exponentially. Massive amounts of data are generated and must be delivered, processed and integrated into real-time actions. There are major fundamental research challenges for effectively designing networking & (edge) systems and leveraging artificial intelligence to support Industry 4.0 applications.

Take autonomous vehicles (AV) as an example. Despite rapid advances in AV technologies, achieving fully autonomous driving still has a long way to go. With the promises of 5G, an alternative approach for further developing autonomous vehicle (AV) technology has gained momentum: “hybrid-autonomous” or teleoperated AVs, where a human operator remotely drives an “autonomous” vehicle when needed. Unlike many other Industry 4.0 use cases that also require extra-high bandwidth and ultra low latency, (partial) AV teleoperation is especially daunting. For example, mobility at vehicle driving speeds pose many networking challenges such as highly varying channel conditions, wildly fluctuating bandwidth, and frequent handovers between cell stations. Moreover, the often-congested road conditions and needs to interact with other humans (e.g., cars driven by human drivers, bicyclists, pedestrians) create more challenges in AV teleoperations. Addressing these challenges requires concerted interdisciplinary efforts integrating advances in networking, systems, AI/ML and human-machine interaction as well as engineering and social sciences.

In this project we aim to advance interdisciplinary, fundamental CISE research through (partially) teleoperated autonomous vehicles (AVs) as a driving use case. We lay out an interdisciplinary and transformative research agenda to develop integrated networking, systems and AI support for Industrial 4.0 application that is human-centered. The innovations include: 1) A semantics oriented and fine-grained networking framework that exploits diversity to provide high bandwidth and low latency. 2) An agile, secure-by-design edge systems architecture that is optimized for AI workloads. 3) Both the networking framework and edge system architecture are designed with a novel application-drive, cross-layer and whole-system approach which enables cooperation across end devices, networks, edge systems and human operators. 4) AI/ML is systematically integrated across layers and system components, with built-in mechanisms also to mitigate risks of inaccurate or false AI predictions. Finally, 5) A human-centered approach that combines faster-than-real-time AI simulations and integrated machine & human intelligence to seamlessly incorporate human-in/on-the-loop.

Using AV teleoperations as a compelling use case, we will apply the above innovations to build a comprehensive platform, dubbed NextMOVE. NextMOVE will provide resilient and safety-critical support for (partially) teleoperated AVs. We will take into real world constraints and application requirements in our system design. We will test and refine our systems and evaluate outcomes iteratively using real-world scenarios and datasets.

This project will help facilitate safe and incremental adoption of (teleoperated) AVs to address many societal challenges, while accelerating the AV technology towards full autonomy. In particular, this project provides a unique opportunity for testing AV teleoperations in Midwest winter and other scenarios. In tackling the fundamental CISE research challenges, the innovations advanced in the project will have broad applicability to many Industry 4.0 use cases and other application domains including smart manufacturing, precision agriculture and telehealth. All are vital to the national economy, security and well-being. This project will serve as forums for academia-government-industry collaboration and technology translation as well as nexus points for broadening participation in research, education and community outreach.

This is a collaborative research project between the University of Minnesota and the University of Michigan. The project is funded by National Science Foundation under the CISE Large Core program with NSF Awards 2321531 and 2321532.

We also acknowledge the generous support from a Cisco Research Grant and unrestricted gifts from InterDigital Inc. for research activities related to this project.