Deployable Neuromorphic AI in Space: Taxonomy and Architecture

Giannis Panagiotopoulos, Luca Macchiaiolo, Priyadarshini Kannan, Evgenios Tsigkanos, Axel von Arnim

Keywords: Neuromorphic computing, On-board AI, Spiking Neural Networks (SNN), System Architecture, Hybrid AI architectures

TL;DR: A position paper introducing a taxonomy and mapping from space autonomy applications to deployable neuromorphic on-board architectures, with feasibility anchors in the power-latency space

Abstract:

Neuromorphic computing is increasingly considered for on-board artificial intelligence in space due to its potential for energy-efficient, low-latency inference and its natural fit to sparse, event-driven sensing. Yet, much of the existing literature remains focused on algorithms, device characteristics, or single-application demonstrations, offering limited guidance on how neuromorphic acceleration should be integrated and operated at spacecraft system level. Consequently, architectural decisions are often technology-driven and difficult to transfer across missions and autonomy functions.
This position paper addresses this gap by introducing an application-driven, system-level framework for neuromorphic on-board AI. We propose a two-dimensional taxonomy based on (i) criticality (safety-critical vs. mission-critical) and (ii) temporal behavior (persistent vs. event-driven), yielding four workload categories capturing key architectural drivers. We map these categories to reusable integration patterns, including execution modes, power management, and host–accelerator interaction assumptions.
We ground the framework with feasibility-level quantitative anchors from representative workloads on neuromorphic and hybrid platforms, focusing on indicative latency, energy, and duty-cycle behavior to support early system-level architectural trade studies for autonomous spacecraft.

Email Sharing: We authorize the sharing of all author emails with Program Chairs.

Data Release: We authorize the release of our submission and author names to the public in the event of acceptance.

Submission Number: 29