HYDRA - Resilient Computation for Heterogeneous Autonomous Drone sYstems
Abstract
The objective of the proposed work is the development of an advanced technology to reliably distribute computation tasks in autonomous Unmanned Aerial Vehicle (UAV) systems. The ability to observe and analyze the surrounding environment is the key to autonomy. However, advanced data processing is an intense and resource consuming task, especially when contextualized in airborne platforms such as UAVs, where weight constraints may limit the availability of vital resources such as energy and processing power. Running complex algorithms on-board may result in limited analysis capabilities, long data capture-to-control time, and reduced mission lifetime. Traditional approaches in mobile computing differentiate the capabilities of the devices, with the extreme points being energy and processing constrained mobile devices and powerful, but static, cloud datacenters. A similar approach can be replicated in UAV systems, integrating in the swarm agile UAVs with limited processing resources, and heavier, more capable, UAVs operating as “flying” datacenters. However, such approach extends the vulnerable surface of the system, exposing it both to adverse environmental conditions and cyber-attacks. In fact, a congested radio spectrum or intentional jamming can disrupt the wireless links connecting the UAVs, thus impairing the ability of the swarm to collaboratively process information. Additionally, a successful attack could compromise the on-board platforms of some of the UAVs, leaving them with reduced or no processing capabilities. This project will develop HYDRA, a technology capable of dynamically rerouting sensing- processing-control pipelines in response to changes in the perceived state of the environment and UAV system. HYDRA will harness the performance improvement granted by distributed computing in heterogeneous UAV systems, while eliminating the additional vulnerability introduced by this approach. The core of HYDRA is a flexible middleware architecture controlled by an intelligent layer observing the system state and making informed decisions.
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- Oct 24, 2018
- Source ID
- HR00111910001
Entities
People
- Marco Levorato
Organizations
- Defense Advanced Research Projects Agency
- University of California Regents