Ground-Side Mission Plan Compilation with Policy-as-Code Guardrails for Cloud-Native Satellite Platforms
Distributed, Parallel, and Cluster Computing
Summary
The authors present the Satellite Mission Compiler, a tool that takes human-written satellite mission plans and turns them into formats that onboard cloud-native satellite software can run. Their system checks the plans for structural errors and policy issues, helps classify problems by severity, and follows strict safety principles. It produces workflows compatible with popular scheduling tools and supports advanced resource management like GPU allocation. They validated their tool through various tests and made it accessible to AI agents. The compiler is shared as open-source software under a free license.
Authors
Hsiu-Chi Tsai, Chia-Tung Chung
Abstract
Onboard cloud-native runtimes for satellites are emerging on multiple tracks (ORCHIDE, Axiom Space's AxDCU-1, Kepler's Jetson nodes), but each assumes that the workflow artifacts it executes arrive from the ground. ORCHIDE's architecture document D3.1 states explicitly that "only the Deferred Phase is part of the ORCHIDE scope," and no open-source ground-side toolchain has been released by the consortium. We present Satellite Mission Compiler, a four-stage pipeline that addresses this gap: it takes a human-authored mission plan, checks it against machine-checkable structural and policy rules, and compiles it into the container-workflow artifacts that cloud-native satellite runtimes consume. The pipeline parses the plan against a Pydantic schema derived from public ORCHIDE materials, evaluates it against an OPA/Rego policy package of ten deny rules with documented provenance, compiles it into a typed WorkflowIntent intermediate representation, and renders it as Argo Workflow DAGs and Kueue Job manifests with Dynamic Resource Allocation (DRA) support. We classify pre-uplink loss events into four severity tiers tied to specific schema and policy checks, and anchor the layered-validation design in the safety reading of defense-in-depth (NASA-STD-8739.8B) rather than the security reading (NIST SP 800-53). The implementation is validated by golden translation evaluations, argo lint, an in-process baseline that reproduces OPA's decisions, and live single-node cluster submission, including a DRA-backed GPU admission cascade on Kueue v0.17.3 (re-validated on v0.18.3) and, on v0.18.3, a unified GPU+CPU device-class quota with a scheduler-level accelerator fallback. Six Model Context Protocol (MCP) tools expose the pipeline to AI agents. The compiler is released under EUPL-1.2 (DOI 10.5281/zenodo.21228150).