Research map / Large language models / AI and society
OnAIR: Applications of the NASA On-Board Artificial Intelligence Research Platform
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's full text. Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Presents OnAIR, an open-source plugin-based software pipeline and cognitive architecture for full life-cycle AI research on aerospace systems, and describes its research and flight deployments (NAMASTE, SCENIC).
Problem
Integrating AI algorithms into production aerospace systems is hard due to costs, timelines, and a risk-averse industry, and existing flight frameworks (cFS, F Prime, SpaceROS) are not built for AI research.
Approach
A plugin-based cognitive architecture with a user-chosen data adapter and four plugin interfaces of increasing knowledge abstraction (knowledge representation, learners, planners, complex reasoners) that follows a sense-perceive-act structure, enabling AI modules to be developed in parallel and deployed onboard.
Key results
- OnAIR was successfully launched and ran without error for 4 days alongside cFS on the SCENIC payload on STP-H9 (FPGA at 100 MHz), with a Kalman filter plugin processing live telemetry.
- NAMASTE used OnAIR for drone methane-search autonomy, with OnAIR enabling interoperability in about two months of a three-year project and supporting the Summer 2024 field campaign.
- OnAIR has reached roughly 30 users and open-source contributors across internal and external domains.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| Use OnAIR to integrate the NAMASTE systems | NAMASTE · NAMASTE | interoperability development time: two months | Helped |
| Run OnAIR alongside cFS for onboard telemetry processing | SCENIC · SCENIC | continuous operation duration: 4 days | Helped |
| Use OnAIR to develop and integrate autonomous-function plugins | SCENIC · SCENIC | development time | Helped |
Surprising findings
- Plugin development was the easiest part of the SCENIC project; more time was spent adapting to the FPGA instruction set and hardware integration.
Limitations
- No quantitative benchmark or controlled comparison against other frameworks; benefits are reported qualitatively through use cases.
- SCENIC experiment was minimal (a Kalman filter) and short (4 days), and NAMASTE was still ongoing at publication.
- Relies on user-designed plugins; the paper does not evaluate algorithmic performance of the AI modules.
Experimental setup
Datasets: simulated Enceladus mission telemetry and mass spectra NAMASTE methane sensor data SCENIC cFS telemetry (40 points at 0.25 Hz) soil images for contaminant detection
Builds on: Russell and Norvig sense-perceive-act agent structure Core Flight System (cFS) Redis
Closest papers by meaning
- Defining Operational Conditions for Safety-Critical AI-Based Systems from Data (NeurIPS 2026 · similarity 0.81)
- AstaBench: Rigorous Benchmarking of AI Agents with a Scientific Research Suite (ICLR 2026 · similarity 0.80)
- Position: Levels of AGI for Operationalizing Progress on the Path to AGI (ICML 2024 · similarity 0.80)
- ScienceBoard: Evaluating Multimodal Autonomous Agents in Realistic Scientific Workflows (ICLR 2026 · similarity 0.80)
- AI-Researcher: Autonomous Scientific Innovation (NeurIPS 2025 · similarity 0.80)
- Terminal-Bench: Benchmarking Agents on Hard, Realistic Tasks in Command Line Interfaces (ICLR 2026 · similarity 0.80)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.