Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

On the Energy Consumption of UAV Edge Computing in Non-Terrestrial Networks

Alessandro Traspadini, Marco Giordani, Giovanni Giambene, Tomaso De Cola, Michele Zorzi

TL;DR

This paper tackles energy-constrained UAV edge computing in non-terrestrial networks by building an energy-centric model across the UAV, HAP, and LEO edge servers, accounting for movement, data offloading, and processing energies. It models data arrivals and processing with a set of D/M/1 queues to capture end-to-end delay and derives expressions for both onboard and edge processing delays, including NTN channel dynamics and distance-dependent propagation. Through simulations, the authors show that partial or full offloading to NTN edge servers can substantially improve UAV autonomy and reduce end-to-end delay in many configurations, with performance depending on frame rate, swarm size, antenna count, elevation angle, and offloading factor. The results offer design insights for NTN-enabled edge computing, highlighting when edge offloading is advantageous and identifying practical constraints such as stability bounds and the need for directional antennas at higher elevation angles.

Abstract

During the last few years, the use of Unmanned Aerial Vehicles (UAVs) equipped with sensors and cameras has emerged as a cutting-edge technology to provide services such as surveillance, infrastructure inspections, and target acquisition. However, this approach requires UAVs to process data onboard, mainly for person/object detection and recognition, which may pose significant energy constraints as UAVs are battery-powered. A possible solution can be the support of Non-Terrestrial Networks (NTNs) for edge computing. In particular, UAVs can partially offload data (e.g., video acquisitions from onboard sensors) to more powerful upstream High Altitude Platforms (HAPs) or satellites acting as edge computing servers to increase the battery autonomy compared to local processing, even though at the expense of some data transmission delays. Accordingly, in this study we model the energy consumption of UAVs, HAPs, and satellites considering the energy for data processing, offloading, and hovering. Then, we investigate whether data offloading can improve the system performance. Simulations demonstrate that edge computing can improve both UAV autonomy and end-to-end delay compared to onboard processing in many configurations.

On the Energy Consumption of UAV Edge Computing in Non-Terrestrial Networks

TL;DR

This paper tackles energy-constrained UAV edge computing in non-terrestrial networks by building an energy-centric model across the UAV, HAP, and LEO edge servers, accounting for movement, data offloading, and processing energies. It models data arrivals and processing with a set of D/M/1 queues to capture end-to-end delay and derives expressions for both onboard and edge processing delays, including NTN channel dynamics and distance-dependent propagation. Through simulations, the authors show that partial or full offloading to NTN edge servers can substantially improve UAV autonomy and reduce end-to-end delay in many configurations, with performance depending on frame rate, swarm size, antenna count, elevation angle, and offloading factor. The results offer design insights for NTN-enabled edge computing, highlighting when edge offloading is advantageous and identifying practical constraints such as stability bounds and the need for directional antennas at higher elevation angles.

Abstract

During the last few years, the use of Unmanned Aerial Vehicles (UAVs) equipped with sensors and cameras has emerged as a cutting-edge technology to provide services such as surveillance, infrastructure inspections, and target acquisition. However, this approach requires UAVs to process data onboard, mainly for person/object detection and recognition, which may pose significant energy constraints as UAVs are battery-powered. A possible solution can be the support of Non-Terrestrial Networks (NTNs) for edge computing. In particular, UAVs can partially offload data (e.g., video acquisitions from onboard sensors) to more powerful upstream High Altitude Platforms (HAPs) or satellites acting as edge computing servers to increase the battery autonomy compared to local processing, even though at the expense of some data transmission delays. Accordingly, in this study we model the energy consumption of UAVs, HAPs, and satellites considering the energy for data processing, offloading, and hovering. Then, we investigate whether data offloading can improve the system performance. Simulations demonstrate that edge computing can improve both UAV autonomy and end-to-end delay compared to onboard processing in many configurations.
Paper Structure (17 sections, 10 equations, 5 figures, 1 table)

This paper contains 17 sections, 10 equations, 5 figures, 1 table.

Table of Contents

  1. Introduction
  2. System Model
  3. Problem Formulation
  4. Channel Model
  5. End-to-End Delay Model
  6. Energy Model
  7. Energy for Movement ($E_{\rm M}$)
  8. Energy for Data Offloading ($E_{\rm DO}$)
  9. Energy for Data Processing ($E_{\rm P}$)
  10. Overall Energy Consumption
  11. Energy Capacity ($E_C$)
  12. Performance Evaluation
  13. Simulation Setup and Parameters
  14. HAP-Assisted Edge Computing
  15. LEO-Assisted Edge Computing
  16. ...and 2 more sections

Figures (5)

  • Figure 1: The edge computing scenario. UAVs are equipped with video cameras recording video frames of size $s_{\rm UL}$ at rate $r$, that may or may not be offloaded to a HAP or a LEO satellite for processing. The processed output is of size $s_{\rm DL}$. The computational capacity is $C_{i}$, $i\in{\{\text{UAV, HAP, LEO}\}}$.
  • Figure 2: Stability of the D/M/1 queues as a function of the frame rate $r$, the number of UAVs $n$, and the offloading factor $\eta$.
  • Figure 3: HAP-assisted edge computing. Average UAV autonomy $\bar{\kappa}$ (left) and average delay $\bar{T}$ (right) vs. $\eta$, as a function of the number of antenna elements $N$ at the UAV, the energy efficiency $\nu_{\rm UAV}$ of the GPUs at the UAV, the number of UAVs $n$, and the frame rate $r$.
  • Figure 4: LEO-assisted edge computing. Average UAV autonomy $\bar{\kappa}$ (left) and average delay $\bar{T}$ (right) vs. $\eta$, as a function of the elevation angle $\alpha$ of the LEO satellite, and the number of antenna elements $N$ of the UAV.
  • Figure 5: Energy consumption $E_{\rm HAP}$ (plain bar) and $E_{\rm LEO}$ (striped bar) of the edge server for HAP- and LEO-assisted edge computing, respectively, vs. the total flying time $t_f$, as a function of the number of UAVs $n$.