cs.NIMar 19, 2026

iSatCR: Graph-Empowered Joint Onboard Computing and Routing for LEO Data Delivery

Authors: Jiangtao Luo, Bingbing Xu, Shaohua Xia, Yongyi Ran

Organizations: the School of Communications and Information Engineering, Chongqing University of Posts and Telecommunications, Chongqing, 400065 China.

Read PDF
Ask AI
Cite

Copy citation

DOINot available for this paper

Abstract

Sending massive Earth observation data produced by low Earth orbit (LEO) satellites back to the ground for processing consumes a large amount of on-orbit bandwidth and exacerbates the space-to-ground link bottleneck. Most prior work has concentrated on optimizing the routing of raw data within the constellation, yet cannot cope with the surge in data volume. Recently, advances in onboard computing have made it possible to process data in situ, thus significantly reducing the data volume to be transmitted. In this paper, we present iSatCR, a distributed graph-based approach that jointly optimizes onboard computing and routing to boost transmission efficiency. Within iSatCR, we design a novel graph embedding utilizing shifted feature aggregation and distributed message passing to capture satellite states, and then propose a distributed graph-based deep reinforcement learning algorithm that derives joint computing-routing strategies under constrained on-board storage to handle the complexity and dynamics of LEO networks. Extensive experiments show iSatCR outperforms baselines, particularly under high load.

Figures & tables

Fig. 1: Scenario of an observation task to be computed onboard and results transmitted back to the ground.
Fig. 2: Distributed Framework of joint decision for computing and routing.
Fig. 3: Proposed distributed graph embedding method. (a) Different regions of the graph representation correspond to the current satellite feature, neighboring satellites, and 2-hop neighboring satellites, respectively. (b) An example of the distributed resource awareness mechanism, in which satellites can perceive resource information within a 3-hop range by integrating the graph representations of neighboring satellites. (c) Message passing and feature aggregating of graph representations in distributed graph embedding mechanism.
Fig. 4: D3QN-based joint optimization of computing and routing for LEO satellites in training.
PARAMETERSVALUES
Environment Parameters 
Number of orbital planes12
Number of satellites per plane24
Orbital altitude500 km
Orbital inclination60°
Beam coverage angle45°
TABLE I: Experimental Parameter Settings
PARAMETERSConstellation Name
IridiumTelesatOneWebStarlink
Orbital Altitude780 km1015 km1200 km570 km
Orbital Inclination86.4°98.98°87.9°70°
Orbital Planes6271836
Satellites per Plane11133620
TABLE II: Satellite Constellation Parameters
CityLatitude (°)Longitude (°)
Dubai25.25255.280
Harbin45.750126.650
Istanbul41.01928.965
Jakarta-6.174106.829
Karachi24.86767.050
Moscow55.75237.616
TABLE III: Locations of Ground Stations in Simulation
Fig. 5: Average reward changing curve during DRL training.
Fig. 6: Simulation results of the satellite constellation under various algorithms and traffic loads. (a) Comparison of average task delay, including transmission, propagation, computation, and queuing delays. (b) Comparison of packet loss rates. (c) Comparison of average transmission hop count.
Fig. 7: Simulation results of the satellite constellation under various algorithms and link failure rate. (a) Comparison of average task delay. (b) Comparison of packet loss rates. (c) Comparison of average transmission hop count.
Fig. 8: Cumulative distribution of satellite computing time under different algorithms
Fig. 9: Simulation results of different algorithms under different-scale satellite constellations. (a) Comparison of average task delay. (b) Comparison of packet loss rates. (c) Comparison of average transmission hop count.
More in this topic

LEO Satellite Networks

Low Earth Orbit (LEO) satellite networks connect moving satellites, ground stations, and users to deliver communications, sensing, and onboard computing.

See all 29 papers

Explore similar work

CardsList

  1. Spatial-Temporal Learning-Based Distributed Routing for Dynamic LEO Satellite Networks

    May 4, 2026Po-Heng Chou, Chiapin Wang, Shou-Yu Chen +1Deep Q-LearningPartially Observable RL

    Research Center for Information Technology Innovation (CITI), Academia Sinica (AS), Taipei 11529, Taiwan · Department of Electrical Engineering, National Taiwan Normal University (NTNU), Taipei 106308 Taiwan

  2. ORBITALIF: An Efficient Spiking Federated Learning Framework for Onboard Cloud Removal

    Aug 25, 2026Bohan Zhang, Chenyu Xu, Yijie Mao +1Cloud RemovalRemote Sensing

    School of Information Science and Technology, ShanghaiTech University, Shanghai, China

  3. Constraint-Aware Execution Planning for Hybrid Space-Ground Compute Workloads

    Mar 4, 2026Subhadip MitraComputation OffloadingLEO Satellite Networks

    RotaStellar