�鶹��Ƶ

Investigators: Eric Frew (PI), Nisar Ahmed (CU), Tim McLain (BYU), and Randy Beard (BYU)
Sponsor: NSF IUCRC Fundamental Research Program

This project will establish fundamental connections between autonomous tasks (like perception, planning, and learning) and dispersed computation accessed through network-enabled cloud-computing infrastructure. Stakeholder institutions cite autonomy as a key enabling technology for future aviation systems, including for unmanned aircraft and their integration into the U.S. National Airspace System. This vision is driven by the expectation that autonomous systems will bring benefits in terms of safety, reliability, efficiency, affordability, and previously unattainable mission capabilities. The move from traditional forms of automation to novel autonomy architectures requires research to overcome major barriers that include collection and perception of large amounts of data, coordinated decision-making in the presence of uncertainty, and learning system performance and transferring learned models between different unmanned aircraft on the shared network.

This effort will investigate the value of utilizing cloud robotics techniques for network-enabled airborne autonomy. We envision the cloud as a “hub of autonomy” for coping with uncertainties, looking beyond simply having autonomous vehicles talk to a massively centralized command center with brittle artificial intelligence and multiple databases. Rather, this work considers the cloud as a “shared offboard brain” that is itself autonomously adapting/evolving in the face of uncertainty, in order to enhance local or network-wide autonomous vehicle planning, learning, and perception at any required scale.

Algorithms will be designed and assessed that allow cooperating unmanned aircraft to access dispersed cloud computing resources. This infrastructure will be used to perform complex computation in support of autonomous perception, planning, and learning tasks. A major theme of the project will be understanding the best allocation of limited on-board payload capacity between local computing for distributed decision-making and improved wireless communication for increased bandwidth to dispersed cloud computing nodes. Research questions addressed in this work will provide a fundamental understanding of aspects of network-enabled autonomy architectures.