Dynamic Coded Distributed Convolution for UAV-based Networked Airborne Computing

TL;DR

This paper proposes a dynamic coded convolution method for UAV-based networked airborne computing, enabling efficient and resilient collaborative processing of convolution tasks despite network heterogeneity and resource variability.

Contribution

It introduces a novel dynamic coded convolution strategy with privacy awareness tailored for UAV-based NAC, addressing node heterogeneity and network dynamics.

Findings

High efficiency demonstrated in simulations

Resilience to uncertain stragglers shown

Effective handling of network heterogeneity

Abstract

A single unmanned aerial vehicle (UAV) has limited computing resources and battery capacity, making it difficult to handle computationally intensive tasks such as the convolution operations in many deep learning applications. UAV-based networked airborne computing (NAC) is a promising technique to address this challenge. It allows UAVs within a range to share resources among each other via UAV-to-UAV communication links and carry out computation-intensive tasks in a collaborative manner. This paper investigates the vector convolution problem over the NAC architecture. A novel dynamic coded convolution strategy with privacy awareness is developed to address the unique features of UAV-based NAC, including node heterogeneity, frequently changing network typologies, time-varying communication and computation resources. Simulation results show its high efficiency and resilience to uncertain stragglers.