cs.NIOct 26, 2026
SaveSatellite Intelligence: Distributed Training, Collaborative Inference,and LLM Adaptation in LEO Constellation
Organizations: Interdisciplinary Centre for Security, Reliability and Trust (SnT), University of Luxembourg, L-1855 Luxembourg, Luxembourg · Department of Computer and Electronic Engineering, University of Hong Kong, Pok Fu Lam, Hong Kong, China · Institute of Space Internet, Fudan University, Shanghai, China · National Mobile Communications Research Laboratory, Southeast University, Nanjing 210096, China, and also with Purple Mountain Laboratories, Nanjing 211100, China ·
Abstract
The proliferation of orbital sensing data and the real-time demands of emerging global services are pushing low Earth orbit (LEO) satellite networks from bent-pipe relays toward intelligent onboard processing. This shift confronts a fundamental tension, where modern artificial intelligence (AI) training paradigms require intensive computation, abundant memory, and stable connectivity, all of which are scarce under the dynamic topologies, intermittent links, and severe size, weight, and power (SWaP) constraints of satellite networks. Addressing this tension demands rethinking AI workflows, from training and inference to the emerging deployment of large language models (LLMs), under a constraint-aware design philosophy. To the best of our knowledge, this paper provides the first survey that jointly covers all three, whereas prior surveys examine these aspects in isolation. We begin by identifying the deployment challenge of satellite intelligence (SI) and deriving design principles for SI. We then examine three advancing paradigms: i) distributed training, where federated learning (FL) and split learning (SL) enable satellite-ground co-training under resource limitations; ii) collaborative inference, which exploits single-satellite, multi-satellite, and satellite-ground architectures to meet real-time service demands; and iii) on-orbit LLM adaptation and serving, which tackles the frontier challenge of deploying LLMs under severe orbital constraints and resource-aware serving. Finally, we identify open challenges and future directions. By bridging the AI and satellite networks, this paper provides a research roadmap for the implementation of space-native AI technologies.
Figures & tables
| Year | Ref. | Introduction | Orbital Constraint | |||
| Analysis | Satellite Distributed | |||||
| Training | Satellite Collaborative | |||||
| Inference | On-Orbit LLM | |||||
| Adaptation | ||||||
| 2020 | [ 1 ] | Present an overview of satellite communication systems in the New Space era, and discuss their applications, enabling technologies, and future challenges | ● | ○ | ○ | ○ |
| 2021 | [ 2 ] | Introduce major AI techniques for satellite communication systems, and discuss their applications in satellite network optimization and management | ● | ○ | ○ | ○ |
TABLE I: Comparison of related surveys and this survey, where ● fully addressed, ◐ partially covered, ○ not addressed.
Fig. 1: Organization of this survey.
Fig. 2: Deployment challenges of SI and the corresponding design principles.
| Approach | Ref. | Performance Metrics | Control Variables | Key Contribution |
|---|---|---|---|---|
| Efficient Local | ||||
| Updates | [ 83 ] | Accuracy; Trainable parameters; Training memory | Adaptation rank; Adapted linear layers | Low-rank factorized updates of large linear layers on a frozen backbone |
| [ 40 ] | Test accuracy; Convergence speed | Per-satellite sub-structure combination; Staleness-based aggregation switching | Sub-structure training matched to heterogeneous satellite computing, memory, and uplink budgets, with pseudo-synchronous aggregation compensating staleness | |
| [ 84 ] | Total time and energy cost; Target accuracy | Participating clients; Local iterations | Local-epoch selection balancing model contribution against computing overhead | |
| [ 85 ] | Convergence; Accuracy under heterogeneous data and resources | Proximal coefficient; Variable local work | Proximal penalty bounding local deviation under heterogeneous epochs | |
| [ 86 ] | Convergence rate; Communication volume | Top- sparsity level; Residual memory | Sparsified updates with residual accumulation compensating discarded elements |
TABLE II: Summary of representative techniques for satellite FL.
Fig. 3: Satellite-GS collaborative architecture for distributed training.
Fig. 4: Training procedure of satellite FL with satellite-GS model exchange during contact windows.
Fig. 5: ISL-assisted update relay for satellites without GS visibility in satellite FL.
| Approach | Ref. | Performance Metrics | Control Variables | Key Contribution |
|---|---|---|---|---|
| Dynamic Split Execution | [ 122 ] | Training latency; Accuracy | Cut layer; Client-side model aggregation interval | Convergence-guided cut-layer adaptation to limited client-side resources |
| [ 123 ] | Per-round latency; Accuracy | Cut layer; Subchannel allocation; Transmit power | Joint cut-layer and resource allocation with last-layer gradient aggregation lightening client-side computation | |
| [ 124 ] | Accuracy; Computation cost | Sub-model width; Sub-model depth | Capacity-matched nested sub-models for heterogeneous client hardware | |
| [ 125 ] | Peak memory; Recomputation overhead | Retained activations; Recomputation order | Activation recomputation trading extra computation for reduced peak memory | |
| [ 126 ] | Time-to-accuracy; Communication volume | Channel grouping; Group-wise compression level | Entropy-guided channel-wise compression shrinking smashed-data volume | |
| [ 127 ] | Accuracy; Communication overhead | Feature-wise dropout probability; Quantization level | Dispersion-based feature-wise dropout and quantization shrinking smashed-data volume |
TABLE III: Summary of representative techniques for satellite SL
Fig. 6: Training procedure of satellite SL with activation and gradient exchange at the cut layer.
Fig. 7: Heterogeneous client-side model customization with a shared base model in satellite SL.
| Dimension | Satellite FL | Satellite SL |
| Onboard computation | Full model, forward and backward | Shallow client-side layers |
| Onboard memory demand | Model, optimizer states, gradients | Client model, cached activations/gradients |
| Communication pattern | One update per round | Smashed data per batch |
| Communication frequency | Low, round-level | High, batch-level |
| Link interruption tolerance | High, rounds can be delayed | Low, stalls the training loop |
| Convergence dependency | Cross-satellite aggregation | Continuous satellite-GS interaction |
TABLE IV: Comparison of satellite FL and satellite SL
Fig. 8: Surrogate-assisted offline training in satellite SL.
Fig. 9: An illustration of local inference, multi-satellite collaborative inference, and satellite-GS collaborative inference.
| Approach | Ref. | Performance Metrics | Control Variables | Key Contribution |
|---|---|---|---|---|
| Model Compression and | ||||
| Reconstruction | [ 170 ] | Accuracy; Latency; Model size | Pruning ratio; Numerical precision | Channel pruning; INT8 quantization-aware training |
| [ 165 ] | Accuracy; Throughput; Board power | Weight and activation bit widths | Quantized CNN; FPGA mapping | |
| [ 171 ] | Segmentation IoU; Model size; Latency | Teacher/student architectures | nnU-Net-to-U-Net knowledge distillation | |
| [ 166 ] | Detection mAP; Parameter count | Student backbone; Teacher loss weights | Contextual feature transfer; Multi-teacher distillation | |
| [ 172 ] | Cloud-masking accuracy; Parameter count | Tensor structure; Bond dimensions | Low-rank tensor factorization |
TABLE V: Summary of representative techniques for local inference
Fig. 10: Early-exit inference with confidence-based exit decisions for onboard processing.
| Approach | Ref. | Performance Metrics | Control Variables | Key Contribution |
|---|---|---|---|---|
| Resource-Adaptive Model | ||||
| Deployment | [ 193 ] | Throughput; Completion time; Network traffic | Satellite group; Layer split points | Neighborhood selection; Runtime prediction; PPO-based partitioning |
| [ 194 ] | Task completion rate; Delay; Energy consumption; Resource utilization | DNN slice boundaries; Offloading destinations; Routes | Workload-balanced partitioning; SAC-based routing and offloading | |
| [ 195 ] | Successfully processed requests; Deployed microservice count | Microservice placement; Per-satellite replica counts | Dependency-aware deployment; Multi-step robust adversarial reinforcement learning | |
| [ 196 ] | Throughput; Average task response time | Task-to-swarm assignment; Partition count; Cut points; Subtask placement | Swarm-level assignment; Runtime-aware partitioning and node-level offloading | |
| [ 197 ] | Token-generation latency under top- expert activation | Layer-to-subnet mapping; Expert-to-satellite placement | Distributed MoE inference architecture; Ring-based layer placement; Activation-aware expert placement |
TABLE VI: Summary of representative techniques for multi-satellite collaborative inference
Fig. 11: Micro-batch pipelining with admission control and checkpointing in multi-satellite collaborative inference.
| Approach | Ref. | Performance Metrics | Control Variables | Key Contribution |
|---|---|---|---|---|
| Contact-Aware | ||||
| Model Partitioning | [ 207 ] | End-to-end latency; Resource estimation error | Layer execution locations; Processing schedule | Waiting between contacts enters the deadline constraint for layer placement |
| [ 208 ] | Weighted task success rate; Weighted average latency | Offloading destination; Semantic split point; Computing allocation | Visibility masks exclude unavailable GSs; Congestion penalties guide splitting and offloading | |
| Downlink-Efficient | ||||
| Compression | [ 25 ] | HSI classification accuracy; Band-selection time; Energy consumption | Selected spectral bands; Band count | Band selection balances task quality and energy while fitting processing and transmission into a contact |
| [ 209 ] | Classification accuracy; Reconstruction PSNR; First-contact latency | Latent size and precision; Receiver configuration | Channel-aware encoding limits payloads; Doppler-aware GS reception improves packet recovery |
TABLE VII: Summary of representative techniques for satellite-GS collaborative inference
| Dimension | Local | Multi-Satellite | Satellite-GS |
| Latency sources | Onboard processing | Processing + ISLs | Processing + SGLs + waiting |
| Model capacity | Single-satellite resources | Pooled orbital resources | Ground-assisted resources |
| Communication dependency | None during execution | ISLs | SGLs |
| Service continuity | Contact-independent | ISL-path dependent | Contact-window limited |
| Failure exposure | Local hardware | Participating nodes/links | Satellite/ground access |
| Accuracy factors | Model and compression | Model, depth, features | Model and received data |
TABLE VIII: Comparison of inference paradigms for SI
Fig. 12: Contact-aware model partitioning in satellite-GS collaborative inference.
Fig. 13: On-orbit LLM adaptation for SI, spanning resource-efficient deployment and LLM-enabled cognitive capabilities.
| Approach | Ref. | Performance Metrics | Control Variables | Key Contribution |
|---|---|---|---|---|
| Hardware-Aware | ||||
| Model Compression | [ 240 ] | Perplexity; Zero-shot accuracy; Speedup | Weight bit-width; Quantization group size | Second-order weight-only quantization reducing weight memory traffic during decoding |
| [ 241 ] | Perplexity; Accuracy; On-device speedup | Per-channel scaling factors; Weight bit-width | Activation-aware weight-only quantization protecting salient channels for on-device LLMs | |
| [ 242 ] | Perplexity; Zero-shot accuracy | Layer-wise pruning ratio; Pruned heads and FFN channels | Structured pruning of attention heads and FFN channels with layer-wise non-uniform ratios | |
| [ 243 ] | Response quality; Calibration; Exposure bias | Student model size; Distillation objective | Reverse-KL distillation transferring generative capability to a compact student model | |
| [ 244 ] | Safety risk; Task utility | Quantization method and bit-width; Safety-patching updates | Quantization-aware safety patching restoring alignment degraded by quantization |
TABLE IX: Summary of representative techniques for onboard LLM serving
| Approach | Ref. | Performance Metrics | Control Variables | Key Contribution |
|---|---|---|---|---|
| Quality-Aware | ||||
| Routing | [ 265 ] | Accuracy; Inference cost | LLM cascade order; Answer-scoring thresholds | LLM cascade escalating a query to a larger model only when a scorer rejects the cheaper model’s answer |
| [ 23 ] | Total system cost of task inference; Delay; Energy consumption | Inference location; Computing and communication resource allocation | DRL-based choice of inferring each task on the device, the serving or a neighboring satellite, or the ground cloud | |
| [ 266 ] | Response quality; Share of queries served by the small model | Routing threshold | Router predicting the small-large model quality gap, sending a query to the large model only when the gap is significant | |
| [ 267 ] | Accuracy; FLOPs | Process-reward acceptance threshold | Step-level routing that invokes the target model only for draft reasoning steps with low process reward | |
| [ 268 ] | Calibration; Discrimination of correct answers | Self-evaluation format; Number of sampled answers | Self-evaluated probability that an answer is correct, usable as the confidence signal that triggers offloading |
TABLE X: Summary of representative techniques for satellite-GS collaborative LLM serving
| Dimension | Onboard Serving | Satellite-GS Collaborative Serving |
| Reasoning location | Single satellite | Satellite and GS |
| Model capacity | Compressed onboard LLM | Onboard LLM + GS-hosted LLM |
| Communication dependency | None during execution | SGLs at task/step boundaries |
| Latency sources | Prefill + decoding | Onboard + SGLs + contact waiting |
| Service continuity | Contact-independent | Contact-window limited |
| Dominant bottleneck | Memory (weights, KV cache), energy | SGL bandwidth, contact windows |
TABLE XI: Comparison of LLM serving paradigms for SI
Fig. 14: KV-cache and long-context management for onboard LLM serving.
Fig. 15: Closed-loop simulation and validation framework for orbital LLM deployment.
| Dimension | Satellite Knowledge-Augmented Reasoning | Satellite-Aware Tool Use | Satellite-GS Cognitive Collaboration | Constellation-Scale Multi-Agent Systems |
|---|---|---|---|---|
| Scope | Single LLM | Single LLM | Satellite–GS | Constellation |
| Core challenge | Time-bounded evidence validity | Physical-model compliance | Differential observability under intermittent links | Partial observability and stale information |
| LLM role | Query generation; Evidence validation | Objective interpretation; Action proposal | Satellite: Provisional planning; GS: Deep analysis | Objective decomposition; Role assignment |
| Key mechanism | Spatio-temporal knowledge graph; Hybrid retrieval | Propose–verify–authorize loop | Versioned shared mission state | Structured messaging; Event-triggered belief sharing |
| Information unit | Metadata-tagged knowledge items | Typed tool calls; Simulator feedback | Versioned state writes | Time-bounded structured messages |
| Non-LLM safeguard | Metadata filtering; Time decay | Digital-twin verification; Predefined authorization | Version checks; Directive/advisory distinction | Message expiration; Quorum-based decisions |
TABLE XII: Comparison of LLM-enabled cognitive capabilities for SI
Fig. 16: Satellite-GS cognitive collaboration via a versioned shared mission state under intermittent connectivity.
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 34 papers