PaperlayerView exact revision on arXiv

Source-linked AI summary

Tiansuan Constellation: An Open Research Platform

Shangguang Wang, Qing Li, Mengwei Xu, Xiao Ma, Ao Zhou, Qibo Sun

arXiv:2112.14957v1cs.DC

TL;DR

Real-satellite research is limited by the cost and risks of space technology, while simulations do not fully capture real constellation conditions. Tiansuan proposes an open platform for experiments on real satellites and reports two deployment case studies. The platform is intended to support research topics ranging from networking to onboard AI.

Abstract

from arXiv · show

Satellite network is the first step of interstellar voyages. It can provide global Internet connectivity everywhere on earth, where most areas cannot access the Internet by the terrestrial infrastructure due to the geographic accessibility and high cost. The space industry experiences a rise in large low-earth-orbit satellite constellations to achieve universal connectivity. The research community is also urgent to do some leading research to bridge the connectivity divide. Researchers now conduct their work by simulation, which is far from enough. However, experiments on real satellites are blocked by the high threshold of space technology, such as deployment cost and unknown risks. To solve the above dilemma, we are eager to contribute to the universal connectivity and build an open research platform, Tiansuan constellation to support experiments on real satellite networks. We discuss the potential research topics that would benefit from Tiansuan constellation. We provide two case studies that have already deployed in two experimental satellites of Tiansuan constellation.

I. INTRODUCTION

Terrestrial infrastructure leaves many areas without Internet access, while satellite research remains constrained by simulations and barriers to real-satellite experiments. Tiansuan is proposed as an open platform for real-constellation experiments that can expose physical-environment effects and deployment pitfalls.

I. INTRODUCTION

  • More than 80% of land and 95% of ocean lack network access, while terrestrial Internet is constrained by geographic accessibility and infrastructure cost.Large LEO constellations are emerging, but simulations cannot fully substitute for experiments on real satellites, which face deployment costs and unknown risks.Source: §I · p. 1
  • Tiansuan proposes a three-stage open research platform, beginning with six satellites, to support experiments on real satellite constellations.The supplied passage gives the first-stage count but truncates the later-stage counts.Source: §I · p. 1
  • Real-constellation experiments can study physical effects such as weather that simulations cannot fully capture and can uncover pitfalls for commercial deployment.The platform comprises institute-owned satellites and satellites joining the Tiansuan plan with its payloads.Source: §I · pp. 1–2

II. TIANSUAN CONSTELLATION DESIGN

Tiansuan combines a staged, multi-owner satellite platform with an open operating model for research, industry, and payload testing. It is intended to support a broad range of experiments.

A. Overview

  • The platform plans three stages of 6, 24, and 300 satellites, with three participation types spanning institute-owned satellites and other institutes’ satellites joining through payloads or a unified interface.Most satellites are planned for sun-synchronous orbit, with the first three combining edge computing and remote sensing and the last three exploring inter-satellite communication links.Source: A. Overview · p. 2

B. Operation Mechanism

  • Research institutes can join by submitting requirements or collaborating to enhance the constellation, while companies and other organizations can use it for research or device and payload testing.The paper presents the platform as a complement to satellite companies’ ecosystems and says it can reduce overall costs for companies and research institutes.Source: B. Operation Mechanism · p. 2

III. POTENTIAL SPECTRUM OF EXPERIMENTS

  • The paper introduces several potential research directions that Tiansuan could support, without specifying them in this passage.The passage frames the directions as a partial spectrum because researchers may devise experiments beyond those anticipated.Source: §III · p. 2

A. Networking

Tiansuan is positioned to support networking experiments across protocol layers and on core networks. The section highlights changing satellite connectivity, constrained onboard resources, and handovers as research challenges.

A. Networking

  • Tiansuan can support networking experiments from the physical through application layers, as well as network-control-plane experiments such as core networks.The paper describes satellite communication as an enabling technology for 6G networks and global mobile coverage.Source: A. Networking · p. 2
  • Core-network functions on LEO satellites must adapt to rapidly changing topology and connectivity while efficiently orchestrating limited power and processing resources.Verifying a cognitive service architecture for 6G core networks is one of Tiansuan’s objectives; the section also describes lightweight satellite core networks as a way to make constellation management more flexible and agile.Source: A. Networking · p. 3

1) Mobility Management:

  • Frequent handovers in large satellite constellations can affect service quality, making mobility management an important experiment for integrated satellite-terrestrial networks.The passage contrasts this with short link duration in LEO satellites.Source: 1) Mobility Management · p. 3

2) Core Network Coordination:

The proposed core-network architecture assigns most control to satellite edge cores and coordination guidance to a cloud core. Serverless and stateless design is described as a way to deploy lightweight, relocatable network functions.

2) Core Network Coordination:

  • The cognitive service architecture divides the core network into satellite edge cores that provide most network control and cloud cores that guide coordination among edge cores.The authors plan to deploy edge core networks using Tiansuan and evaluate their performance.Source: 2) Core Network Coordination · p. 3

3) Serverless and Stateless Design:

  • Serverless design splits network elements into independently developed, tested, deployed, and evolved functions, while stateless design separates state from functions.Remote interfaces can let satellite network functions access ground resources, and functions can be placed, cancelled, or redeployed according to business needs and resource availability.Source: 3) Serverless and Stateless Design · p. 3

4) Knowledge Graph:

A knowledge map can represent the real-time status and relationships of users, networks, and services in dynamic satellite networks. Knowledge-graph extraction and reasoning can support node selection, traffic planning, signaling, and migration strategies that help maintain service continuity.

4) Knowledge Graph:

  • A knowledge map of users, networks, and services can support real-time user-plane node selection and dynamic traffic-path planning despite uncertain links and network dynamics.Its overview of real-time status and logical relationships also supports network decision-making.Source: 4) Knowledge Graph · pp. 3–4
  • Knowledge-graph extraction and reasoning can support network decision-making and simplify signaling interactions.Source: 4) Knowledge Graph · p. 4
  • Temporal node attributes and satellite mobility can inform migration strategies for network-element functions, helping ensure service continuity.Source: 4) Knowledge Graph · p. 4

B. Computing

Tiansuan is proposed as a platform for satellite computing research, including space-native data processing and general computing services. The section outlines Earth-observation data and ground-station challenges, alongside research directions for federated learning, AI testing, and onboard inference.

B. Computing

  • Tiansuan would support satellite-borne computing, resource management, general computing services, and heterogeneous satellite-data integration as technical reference for space data centers.Source: B. Computing · p. 4

1) Real-Time Earth Observation:

  • About 45% of LEO satellites are used for Earth observation, where growing space-native data volume and limited downlink capacity bottleneck real-time applications.Proposed approaches include downloading the freshest information and scheduling downloads across distributed ground stations.Source: 1) Real-Time Earth Observation · p. 4
  • Satellite waiting time before ground-station links, limited contact time, bandwidth bottlenecks, and contention make fresh, high-quality image downloads difficult.The section motivates AoI-aware sampling, transmission, and processing research for Earth-observation systems.Source: 1) Real-Time Earth Observation · p. 4
  • Ground-station deployment costs around a million dollars, while dynamic link scheduling must account for orbits, link quality, and weather.The section proposes scalable ground stations and spatial-temporal link prediction using historical data and environmental information.Source: 1) Real-Time Earth Observation · p. 4

2) AI in Space:

  • Tiansuan would support federated-learning and AI-acceleration experiments to verify space-borne model training, satellite application algorithms, and a constellation-based FL platform.Federated learning can process satellite-collected information onboard while protecting data privacy.Source: 2) AI in Space · p. 5
  • The section proposes automatically testing FL-driven satellites with cases reflecting real-world space-environment changes, with generated inputs also usable to retrain models for robustness.Source: 2) AI in Space · p. 5
  • Onboard inference could filter useless images before downlink, while heterogeneous satellite SoCs and speculative inference offer research directions for AI and multimodal applications.Source: 2) AI in Space · p. 5

3) Space Service Computing:

Time-varying ground-satellite and inter-satellite links leave service routing, deployment, offloading, migration, and coordination open for research. The section presents micro-services and function-as-a-service as lightweight architectures, while identifying unresolved partitioning, deployment, management, and resource-allocation problems.

3) Space Service Computing:

  • Time-varying ground-satellite and inter-satellite connectivity leaves path selection for reliable service and dynamic deployment, offloading, migration, and coordination to be explored.Source: 3) Space Service Computing · p. 5
  • Micro-services and function-as-a-service decompose applications into fine-grained components that can be deployed and executed quickly and independently.Source: 3) Space Service Computing · p. 5
  • Open research problems include application partitioning, dynamic deployment of dependent components, distributed registration and management, and resource allocation.Source: 3) Space Service Computing · p. 5

4) Cluster Orchestration:

Cluster orchestration could manage satellite and onboard computing resources, but LEO mobility, limited per-satellite compute, and load imbalance complicate task placement. The section also calls for a satellite operating system that supports real-time applications while addressing software security, memory safety, and reliability in space.

4) Cluster Orchestration:

  • Satellite task allocation must balance onboard energy efficiency, computing delay, and communication delay while adapting to limited compute, high mobility, and load imbalance.The section identifies satellite clustering and cooperation strategies as research needs.Source: 4) Cluster Orchestration · p. 5
  • Treating LEO satellites as edge nodes could reuse existing applications and cloud-native tools for unified scheduling and deployment, avoiding platform-specific redevelopment.Source: 4) Cluster Orchestration · p. 5
  • Multi-level orchestration could integrate computing units within satellites as well as coordinate across satellites, with the aim of improving computing efficiency.Source: 4) Cluster Orchestration · pp. 5–6

C. Satellite Operation System

  • A satellite operating system should support user-defined, quantifiable multi-task real-time requirements on embedded hardware while addressing third-party software security, runtime memory safety, and space-environment reliability.The section names sandboxing and Rust as possible approaches, alongside multimachine backup and fault tolerance.Source: C. Satellite Operation System · p. 6

D. Security and Reliability

Tiansuan could support research on constellation security and service reliability under the constraints of the LEO environment. The section highlights open blockchain issues and the difficulty of protecting satellite computation with limited resources.

D. Security and Reliability

  • Wide satellite coverage may let everyone participate in the network, but it also raises security concerns; temperature and radiation make reliable LEO service computation challenging.Tiansuan would support solutions for security and reliability.Source: D. Security and Reliability · p. 6

1) Security of Constellation:

  • Satellite data storage, traceable digital signatures, and satellite-ground blockchain transmission remain research issues that ground simulations cannot fully address.Tiansuan offers an open-source platform for real experiments on blockchain deployment.Source: 1) Security of Constellation · p. 6

2) Reliability of Constellation:

  • Temperature swings, fluctuating computing power, and radiation-related component failures challenge service computation on resource-constrained LEO satellites.Temperatures can range from 400K in illuminated areas to 150K in non-illuminated areas; traditional redundancy and checkpoint strategies are difficult to apply.Source: 2) Reliability of Constellation · p. 6

E. Hardware Testing

Tiansuan supports in-orbit hardware testing and reports a satellite-borne B5G core network deployment. A TY20 test exercised signaling with a terrestrial private 5G network, while the two experimental satellites carry a lightweight core network.

E. Hardware Testing

  • Tiansuan supports in-orbit performance and reliability tests of sensors, cameras, antennas, acceleration boards, and distributed computing platforms.These tests evaluate hardware and computing platforms in the satellite environment.Source: E. Hardware Testing · p. 7

IV. CASE STUDY

  • The case studies comprise two use cases deployed in experimental satellites Baonyun and Chuangxingleishen, scheduled for launch on December 2, 2021.Source: §IV · p. 7

A. Satellite-borne B5G Core Network

  • The TY20 test successfully completed user registration, session establishment, and satellite-equipment control through B5G-core signaling with a terrestrial private 5G network.Downlink monitoring showed the three main core components operated normally and generated correct control data, which was downloaded to the terrestrial network.Source: A. Satellite-borne B5G Core Network · p. 7
  • A lightweight B5G core network was deployed on both Baonyun and Chuangxingleishen; functionality and performance tests were planned after launch.The updated core enhances signaling interactions and can support video calls using Session Initiation Protocol.Source: A. Satellite-borne B5G Core Network · p. 7

B. Image Inference Based on KubeEdge

The KubeEdge-based image-inference system splits detection between a satellite and a ground controller. The conclusion presents Tiansuan as a platform for experiments on real satellite constellations.

B. Image Inference Based on KubeEdge

  • Low-confidence satellite detections trigger transfer of the image to a ground controller for inference with a high-precision model; high-confidence results are used onboard.A lightweight model runs on the satellite, and the high-precision model runs on the ground controller, which connects intermittently according to satellite position.Source: B. Image Inference Based on KubeEdge · p. 7
  • The collaborative inference system uses satellite computing resources and reduces end-to-end latency in the AI pipeline.Source: B. Image Inference Based on KubeEdge · p. 7

V. CONCLUSION

  • The paper presents Tiansuan as a global research platform for evaluating experiments on real satellite constellations.Source: §V · p. 7
Loading 2112.14957v1…