Performance Evaluation of Satellite-Based Data Offloading on Starlink Constellations

TL;DR

This paper tackles the challenge of providing real-time data processing for autonomous vehicles in rural areas by offloading compute tasks to Starlink's LEO satellites. It presents a framework that dynamically decides between onboard processing, satellite offloading, and dropping, with two strategies BOO and LDBOO to manage satellite queue congestion, backed by a realistic evaluation using Starlink orbital traces and the 3GPP TR 38.811 channel model. The key contributions are the novel offloading framework, the back-off and dropping policies, and a comprehensive performance study showing that LDBOO with SR can meet tight latency under certain capacity and constellation density conditions. The findings highlight the practical feasibility and the need for dense satellite deployments and careful parameterization (e.g., $C_{ m LEO}$, $oldsymbol{r}$, $oldsymbol{n}$) to enable real-time space-edge computing for ITS applications.

Abstract

Vehicular Edge Computing (VEC) is a key research area in autonomous driving. As Intelligent Transportation Systems (ITSs) continue to expand, ground vehicles (GVs) face the challenge of handling huge amounts of sensor data to drive safely. Specifically, due to energy and capacity limitations, GVs will need to offload resource-hungry tasks to external (cloud) computing units for faster processing. In 6th generation (6G) wireless systems, the research community is exploring the concept of Non-Terrestrial Networks (NTNs), where satellites can serve as space edge computing nodes to aggregate, store, and process data from GVs. In this paper we propose new data offloading strategies between a cluster of GVs and satellites in the Low Earth Orbits (LEOs), to optimize the trade-off between coverage and end-to-end delay. For the accuracy of the simulations, we consider real data and orbits from the Starlink constellation, one of the most representative and popular examples of commercial satellite deployments for communication. Our results demonstrate that Starlink satellites can support real-time offloading under certain conditions that depend on the onboard computational capacity of the satellites, the frame rate of the sensors, and the number of GVs.

Figures (6)

• Figure 1: Illustration of the scenario. We deploy $n$ GVs in a rural/unserved area under the coverage of a constellation of Starlink LEO satellites. GVs (satellites) are equipped with a computing platform with capacity $C_{\rm GV}$ ($C_{\rm LEO}$) for processing data.
• Figure 2: Elevation angle over time for some Starlink satellites (with IDs).
• Figure 3: Real-time probability (solid bars) and data drop probability (striped bars) for different offloading strategies vs. $\sigma$ and $t^m_o$, with $r = 30$ fps, $C_{\rm LEO} = 20$ TFLOPS, and $n = 100$ GVs.
• Figure 4: Delay vs. $n$ for different satellite selection policies vs. $r$, with $C_{\rm LEO}=10$ TFLOPS. We consider LDBOO offloading.
• Figure 5: Real-time probability as a function of $r$ and $C_{\rm LEO}$, vs. the number of Starlink satellites $s$. We consider LDBOO offloading with SR.