Skip to content

TX10 — Autonomous Systems

Created: 2026-04-06 (session 31)

Portfolio Overview

Active projects: 11
Programs: STRG 6 (54.5%), SBIR/STTR 2 (18.2%), GCD 1 (9.1%), FireSense 1 (9.1%), MCO 1 (9.1%)
TRL distribution: TRL 2: 6 (54.5%), TRL 3: 3 (27.3%), TRL 4: 1 (9.1%), TRL 5: 1 (9.1%)
Query: find_projects(technology_area="TX10", status="Active"), snapshot 2026-04-04

Important distinction: TX10 = the cognitive/reasoning layer (planning, fault diagnosis, situational awareness, multi-agent coordination). TX04 = the physical robot layer (locomotion, manipulation, sensing). They co-exist and overlap in practice but are taxonimically distinct.

Portfolio character: Extremely thin (11 projects) and almost entirely early-stage research. The one MCO program project represents the only serious maturation pathway in this taxonomy area.

Taxonomy Structure

TX10 sub-areas from active project labels: - TX10.1: Situational and Self-Awareness Technologies - TX10.2: Executive Functions (planning, scheduling, execution, fault diagnosis) - TX10.2.2: Activity and Resource Planning and Scheduling - TX10.2.4: Execution and Control - TX10.2.5: Fault Diagnosis and Prognosis - TX10.3: Interaction and Collaboration - TX10.3.1: Joint Knowledge and Understanding - TX10.4: System-Level Autonomy - TX10.4.3: Operational Assurance of Autonomous Systems

The Dominant Project: Earth Independent Operations Anomaly Response

Earth Independent Operations Anomaly Response ([157864])
MCO (Mars Campaign Office), NASA ARC. TRL 3→6. 2024-10 to 2030-09. 4,068 views. Destinations: Mars, Moon.

The MCO EIOAR project (4,068 views) dwarfs all other TX10 projects by scope, duration, and visibility. It is a 6-year program with 8 explicitly-defined sub-projects:

Sub Name Function
AR1 Diagnostics Anomaly detection; "out of family" identification; fault isolation
AR2 Fault Hypothesis Generation Generate, isolate, and rank fault hypotheses
AR3 Procedure Synthesis Synthesize novel response procedures for unknown/ambiguous faults
AR4 Procedure Validation Validate novel procedures by crew safety, mission success, maintaining capabilities
AR5 Crew Diagnostic Interface Human-AI interface for crew anomaly analysis
AR6 Time to Effect Estimate duration until anomaly effect manifests
AR7 Autonomous Safing Auto-safe if time to effect too short for crew response
AR8 Large Language Models LLMs as enabling technology for AR capabilities

Rationale: Mars missions have 44-minute round-trip communication delays (plus potential communication blackouts). Historical data (Valinia et al. 2021) suggests >50% probability of high-consequence fault during Mars transit. The current model of "ground team directs crew" breaks down completely.

PM: Christopher A. Teubert + Katelyn J. Griffith (ARC)
Program Directors: Dayna S. Ise, Lindsay T. Aitchison, Lynn N. Smith (MCO)

Library item: Research paper "A Hybrid Model and Data Driven Approach for Anomaly Detection in Space Power Systems" (PDF, 1.7 MB, file ID 388413).

Cross-reference: Covered in detail in MCO/EIO page and partially in TX04 Robotics.

STRG Research Portfolio (6 projects, TRL 2→3)

Academic autonomy research at TRL 2-3, all ending 2026-2028:

Project ID Lead Sub-area End
Autonomous Fault Detection for Robotic In-Space Assembly 158558 USC TX10.2.5 2028
Autonomous Multiagent Search-and-Image Tasking 156328 CU Boulder TX10.3.1 2027
Deep RL Multi-Modal Locomotion for Wheel-on-Limb Robotics 118435 U Maryland TX10.2.2 2026
Enabling Long-term Robot Autonomy via Adaptable Fault Resilience 118425 CU Boulder TX10.2.5 2026
Long-Duration Risk-Aware Adaptive Sampling (158489) MIT TX10.2.4 2028
Low-Power Real-Time Planning for Uncertain Environments 158190 Rice TX10.2.4 2028

Notable themes: - Fault resilience dominates (3 of 6 projects target fault detection/diagnosis/recovery) - Wheel-on-limb robotics (U Maryland) and adaptive sampling (MIT) target planetary rovers - CU Boulder has 2 projects — both fault-focused, one ending 2026, one 2027

GCD Project: Certification for Autonomy (CARMEL)

Certification for Autonomy (184634)
GCD, NASA ARC. TRL 3→4. 2025-04 to 2026-11. 615 views. Multi-center + industry.

Read from live API: 2026-04-06

CARMEL (Certification for Autonomous and Reliable Mission-critical Embedded Lifecycle, per description) is the most consequential TX10 project after EIOAR. It is not a technology — it is a certification methodology framework for how autonomous systems get approved for NASA space missions.

What CARMEL is Building

Three outputs by November 2026: 1. Exemplar assurance case — a worked example for a reference mission (single spacecraft in cislunar space, Level 3 autonomy: autonomous navigation without Earth dependence during nominal operations; ground contact only for off-nominal) 2. Certification process — metrics, measures, and V&V strategies for certifying autonomy using an assurance case approach 3. Assurance technology — software toolset to help projects build and manage their assurance cases

The Assurance Case Approach

The core methodology is the "assurance case": a structured argument that a system meets its objectives and operates safely, with documented evidence at each claim. This is objectives-driven (clear goals through structured framework), risk-informed (risk analysis for decisions), and case-assured (evidence-based argument).

Precedents cited in the project: FAA (UAS performance-based approval), NRC (nuclear waste disposal), FDA (infusion pumps).

Why CARMEL Exists

The motivation is structural: NASA's NPR 7120.5 program/project management requirements were designed for NASA-directed development where NASA oversees the full design/development/test lifecycle. With commercial acquisition (SpaceX Crew Dragon, Astrobotic CLPS, Blue Origin Lunar), NASA no longer manages that process. A commercial provider flying Level 3 autonomous navigation cannot be expected to satisfy NPR 7120.5.

CARMEL provides an alternative framework that commercial providers can use to certify autonomy for NASA missions. This is certification infrastructure for the commercial spaceflight era.

Team

Multi-center/industry collaboration: - Lead: Ames Research Center - NASA centers: GSFC, LaRC, JSC
- Industry: Blue Origin, Red Canyon Software Systems

The presence of Blue Origin is notable — they are a direct beneficiary of certification frameworks that work for commercial providers.

Program Director/Manager: Werkheiser + Thornblom (same GCD leadership as Lunar Night Survival, HI-RATE, and other 2025 GCD starts)

TX Mismatch

Human classifier: TX10.4.3 (Operational Assurance of Autonomous Systems). ML prediction: TX17.2.1 (Onboard Navigation Algorithms). The ML reads the reference mission description (autonomous spacecraft navigation) and predicts a navigation TX; the human correctly classifies the broader certification methodology goal. This is a case where the ML over-anchors on mission description keywords vs. project purpose.

SBIR Projects (2)

Project ID Lead Focus
Curating Uncertainty for Reliable Exploitation and Collaboration (CURE-C) 182934 XAnalytix Systems Uncertainty management
Uncertainty Management Framework for Space-based Autonomy 182893 CFD Research Corp Uncertainty quantification

Both are uncertainty management frameworks for space autonomy. Both start in 2025, end 2027. At TRL 4 and 5 respectively — more mature than the STRG portfolio.

FireSense Project

Distributed Spacecraft with Heuristic Intelligence for Wildfire Spread Monitoring ([157516])
FireSense Technology program. NASA ARC. TRL 3. 2023-07 to 2026-07. 664 views.

TX10.1 (Situational Awareness). Constellation of smallsats with heuristic intelligence to monitor wildfire spread. Application: responsive Earth observation, not space exploration. The FireSense program is a new program type to this KB — appears to be a NASA Earth-application technology development program funded via STMD.

Key Findings

  1. TX10 is a near-empty area. Only 11 active projects across an area that's theoretically central to all future autonomous space operations. By comparison, TX04 (Robotics) has 37 active projects.

  2. The taxonomy split is real but blurry. TX04 and TX10 both cover "autonomous systems" — the distinction is physical robots (TX04) vs cognitive autonomy software (TX10). In practice, projects span both (e.g., wheel-on-limb locomotion in TX10 vs VIPER software in TX04).

  3. MCO EIOAR dominates the field (4,068 views, 6 years, 8 sub-programs). It represents the most mature and mission-linked autonomy program in TechPort. The LLM sub-area (AR8) is notable — MCO is explicitly incorporating LLMs into safety-critical fault response for Mars missions.

  4. Fault resilience is the dominant research theme in the academic portfolio — more projects target fault detection/diagnosis than any other sub-area. This is consistent with the EIOAR's emphasis on anomaly response.

  5. TRL progression pathway is missing. STRG projects max at TRL 3, SBIR projects at TRL 4-5, and then there's a gap before the MCO programs at TRL 3-6. No FO flight-test layer exists for cognitive autonomy (unlike sensors or materials). How does software autonomy get validated?

Open Threads

  1. AR8 (LLM for fault response): Which LLMs/architectures? What safety validation approach? Read the ARC paper (file ID 388413).
  2. EIOAR documents: Paper covers "Hybrid Model and Data-Driven Anomaly Detection for Space Power Systems" — this is specifically AR1 (diagnostics). Check for more papers.
  3. TX10 vs TX11: TX11 covers AI/ML software. Some TX10 projects (especially AR8, adaptive planning) could plausibly be TX11. The TX10/TX11 boundary may be another taxonomy mismatch zone.
  4. FireSense program scope: One TX10 project. Are there other FireSense projects in other TX areas?

Cross-References