Guidance and Control Neural Network Acceleration using Memristors

2025-09-02

AI Summary

Small satellites face severe constraints in power budget and radiation-hardened electronics, hindering deployment of high-performance on-board AI. Method: This paper proposes a memristor-based (PCM/RRAM) compute-in-memory neural network accelerator tailored for guidance and control tasks (G&CNET), explicitly modeling device non-idealities—including conductance drift and stochastic noise—and investigating post-degradation retraining strategies. Contribution/Results: Simulation results demonstrate that the accelerator efficiently learns expert control policies. Although initial inference accuracy degrades due to hardware non-idealities, lightweight retraining fully restores nominal performance. The architecture achieves high radiation tolerance, ultra-low power consumption, and robustness against device aging—establishing a practical, hardware-aware paradigm for next-generation on-board AI systems.

Technology Category

• Machine Learning: Hardware-aware ML
• Search and Optimization: Learning to Search
• Intelligent Robots: Behavior Learning & Control

Application Category

• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Responsible Web: Machine-in-the-loop, human agency and autonomy

Abstract

In recent years, the space community has been exploring the possibilities of Artificial Intelligence (AI), specifically Artificial Neural Networks (ANNs), for a variety of on board applications. However, this development is limited by the restricted energy budget of smallsats and cubesats as well as radiation concerns plaguing modern chips. This necessitates research into neural network accelerators capable of meeting these requirements whilst satisfying the compute and performance needs of the application. This paper explores the use of Phase-Change Memory (PCM) and Resistive Random-Access Memory (RRAM) memristors for on-board in-memory computing AI acceleration in space applications. A guidance and control neural network (G&CNET) accelerated using memristors is simulated in a variety of scenarios and with both device types to evaluate the performance of memristor-based accelerators, considering device non-idealities such as noise and conductance drift. We show that the memristive accelerator is able to learn the expert actions, though challenges remain with the impact of noise on accuracy. We also show that re-training after degradation is able to restore performance to nominal levels. This study provides a foundation for future research into memristor-based AI accelerators for space, highlighting their potential and the need for further investigation.

Problem

• Accelerating neural networks for space applications
• Addressing energy and radiation constraints on small satellites
• Evaluating memristor-based AI accelerators with device non-idealities

Innovation

• Memristor-based neural network acceleration for space
• Phase-change and resistive memory in-memory computing
• Re-training restores performance after device degradation