The Bandwidth Problem
A single Earth observation satellite generates 1–5 TB of raw imagery per day. A constellation of 100 satellites produces 100–500 TB/day. The downlink bandwidth available during a ground station pass is typically 1–2 Gbps for 8–12 minutes, yielding roughly 60–180 GB per pass. The math does not work: satellites generate far more data than they can transmit.
| Constellation | Satellites | Daily Data | Downlink Capacity | Gap |
|---|---|---|---|---|
| Planet (SuperDove) | ~200 | ~30 TB | ~5 TB | 6x |
| Maxar WorldView | 4 | ~2 TB | ~0.8 TB | 2.5x |
| Pixxel Fireflies | 6 (growing to 24) | ~8 TB | ~1 TB | 8x |
| Future mega-constellations | 500+ | ~500 TB | ~20 TB | 25x |
The traditional solution is selective tasking: only image what a customer has requested. But this throws away the global monitoring capability that constellations are designed for. The alternative is processing on-orbit -- running ML inference on the satellite itself and transmitting only the results.
Current State of On-Orbit Processing
Several companies are deploying or testing ML inference hardware on satellites:
| Company | Approach | Hardware | Status |
|---|---|---|---|
| OrbitsEdge | Hardened edge compute module (SatFrame) | HPE Spaceborne Computer-2 heritage | ISS-tested; commercial deployment |
| Pixxel | On-board ML for hyperspectral classification | Custom FPGA + ARM pipeline | Firefly constellation operational |
| Satellogic | On-board cloud detection, change detection | Xilinx FPGA accelerators | Operational on 30+ satellites |
| Ubotica (now Dott) | CogniSAT AI processor for ESA missions | Myriad X VPU, radiation-hardened | Flown on PhiSat-1 (ESA) |
| Xplore | Edge ML for autonomous tasking | NVIDIA Jetson (rad-tolerant packaging) | Development |
The ground segment is also evolving. AWS Ground Station provides ground station-as-a-service, and Azure Orbital offers similar capabilities. These services reduce the latency from downlink to processing but do not solve the bandwidth bottleneck -- they just move the processing from the customer's data center to the cloud provider's.
The Embedding Connection
Consider two satellite missions:
- Crop monitoring: needs to distinguish wheat from corn from soybeans, detect drought stress vs. nitrogen deficiency vs. disease, and track phenological stage. The relevant spectral features are subtle differences in red-edge reflectance, chlorophyll fluorescence, and SWIR absorption. These features occupy specific embedding dimensions that PCA would discard.
- Maritime surveillance: needs to detect ships, classify vessel types, and identify dark vessels (AIS transponder off). The relevant features are shape, wake patterns, and SAR backscatter intensity. Entirely different embedding dimensions from crop monitoring.
A single global PCA basis optimized for "variance explained" would keep the dimensions that separate land from water from cloud (the top principal components) and discard the dimensions that separate wheat stress from corn stress, or tanker from container ship. Both missions would fail at their specific tasks while "performing well" on a generic benchmark.
Task-Specific Compression
The solution is not to find the "right" global dimensionality. It is to train task-specific projections that preserve the features relevant to each mission.
| Mission | Critical Features | Min. Dims (Estimated) | PCA Would Destroy |
|---|---|---|---|
| Crop species classification | Red-edge slope, chlorophyll indices, phenology | 64–128 | Inter-species spectral differences |
| Crop stress detection | NDVI anomaly, water stress indices, fluorescence | 128–256 | Stress-type discrimination |
| Mineral mapping | SWIR absorption features (narrow, diagnostic) | 256+ | Mineral-specific absorption bands |
| Ship detection (SAR) | Backscatter intensity, shape, wake | 32–64 | Fine vessel type distinctions |
| Change detection | Temporal difference features | 32–64 | Gradual vs. sudden change |
| Cloud masking | Brightness, texture, thermal contrast | 8–16 | Nothing -- this is the easy task |
Cloud masking works at 16 dimensions because it is an inter-domain task (cloud vs. not-cloud) -- exactly the kind of task that PCA compression handles well. Mineral mapping requires 256+ dimensions because it is an intra-domain task (distinguishing kaolinite from montmorillonite from illite within the broader category of "clay minerals"). This is the same pattern documented in our dimensionality research for legal terms.
The Matryoshka Approach for Space
Matryoshka Representation Learning (MRL) trains a model to produce useful embeddings at multiple dimensionalities simultaneously. The first 32 dimensions are optimized to be useful on their own. The first 64 include those 32 plus 32 more. And so on up to the full 768 or 1024 dimensions.
For space-based processing, MRL offers an elegant solution:
- Train on ground: train the Matryoshka model with domain-specific data for the satellite's mission (crop monitoring, maritime, etc.)
- Deploy on orbit: run inference using the full model, producing full-dimensional embeddings
- Truncate for transmission: transmit only the first k dimensions, where k is chosen based on the mission's precision requirement and available bandwidth
- Reconstruct on ground: use the truncated embedding for immediate queries; request full embedding for high-precision analysis during the next downlink window
The Bandwidth Math
| What's Transmitted | Size per Image Tile | Tiles per Pass | Total per Pass |
|---|---|---|---|
| Raw imagery (10 bands, 16-bit) | ~200 MB | ~300 | ~60 GB |
| Compressed imagery (JPEG2000) | ~40 MB | ~300 | ~12 GB |
| Embeddings at 768d (float32) | ~3 KB | ~300 | ~900 KB |
| Embeddings at 128d (float16) | ~256 B | ~300 | ~75 KB |
| Embeddings + metadata + alerts | ~1 KB | ~300 | ~300 KB |
The difference is five orders of magnitude. Embedding transmission fits in the margins of telemetry bandwidth that every satellite already has. The question is not whether to transmit embeddings -- the question is whether the embeddings preserve enough information to be useful.
Federated Learning Across Constellations
As satellite constellations grow, a new paradigm emerges: federated learning across orbiting nodes. Each satellite trains (or fine-tunes) a local model on its observations. Inter-satellite links (ISLs) allow gradient sharing between constellation members without transmitting raw data to ground.
The challenges are substantial:
- Non-IID data: each satellite sees different parts of Earth at different times. The data distribution is inherently non-identical across nodes.
- ISL bandwidth: inter-satellite links (laser or RF) have limited bandwidth -- sharing model updates is feasible, sharing raw imagery is not.
- Compute heterogeneity: satellites in a constellation may have different hardware generations and capabilities.
- Orbital mechanics: communication windows between satellites depend on relative position, which changes continuously.
Cross-Satellite Vector Search
A tantalizing application: maintain a distributed vector index across the constellation. Each satellite holds embeddings for its recent observations. A query from ground ("find all images of this port from the last 48 hours") is broadcast to the constellation, each satellite searches its local index, and the top results are transmitted as metadata (embedding + coordinates + timestamp). Only the relevant raw imagery for the top matches is downlinked.
This transforms the satellite constellation from a data collection system into a queryable sensor network -- a spatial search engine with global coverage and sub-daily revisit.
The Future: Compute in Orbit
Three trends are converging:
- Radiation-tolerant AI accelerators: purpose-built chips (not adapted consumer GPUs) designed for inference in the radiation environment of LEO. Ubotica's CogniSAT, Syntiant's NDP, and custom FPGA designs are leading this.
- Software-defined payloads: satellites whose processing pipelines can be updated after launch. ESA's PhiSat program and DARPA's Blackjack demonstrate the concept. The model deployed at launch is not the model running three years later.
- Optical inter-satellite links: SpaceX Starlink has demonstrated laser ISLs at multi-Gbps. When constellation members can communicate at high bandwidth, distributed processing becomes practical -- the constellation becomes a mesh computer, not a collection of independent sensors.
The constraint that does not change is physics: downlink bandwidth to ground stations remains the bottleneck, and power budgets on satellites are finite. Both constraints favor compact representations -- embeddings rather than raw data, lower dimensionality rather than higher. The research question is how to compress without destroying the distinctions that define the mission.