TinyAC: Bringing Autonomic Computing Principles to Resource-Constrained Systems

2025-09-17

AI Summary

Resource-constrained IoT devices face significant challenges in realizing autonomous computation (AC) due to severe limitations in memory, energy, and computational capacity. Method: This paper proposes TinyAC—a lightweight hybrid AC framework integrating on-device TinyML models, continual learning at the edge, and LLM-guided reasoning. Unlike conventional AC architectures, TinyAC leverages LLMs to generate high-level policies, which are then compressed into compact, edge-executable instructions—reducing communication and computational overhead. It further introduces an adaptive control mechanism for runtime anomaly detection and closed-loop response. Contribution/Results: Evaluated on typical IoT scenarios, TinyAC achieves over 92% autonomous anomaly recovery rate with memory footprint under 1.2 MB, demonstrating superior energy efficiency, real-time responsiveness, and deployability. This work bridges a critical gap in edge-AI–oriented AC system design and establishes a novel paradigm for LLM-augmented embedded intelligence.

Technology Category

• Machine Learning: Learning on the Edge & Model Compression
• Search and Optimization: Learning to Search
• Constraint Satisfaction and Optimization: Constraint Learning and Acquisition

Application Category

• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications
• Responsible Web: Machine-in-the-loop, human agency and autonomy
• Economics, Online Markets and Human Computation: Cost models of using LLMs in production systems

Abstract

Autonomic Computing (AC) is a promising approach for developing intelligent and adaptive self-management systems at the deep network edge. In this paper, we present the problems and challenges related to the use of AC for IoT devices. Our proposed hybrid approach bridges bottom-up intelligence (TinyML and on-device learning) and top-down guidance (LLMs) to achieve a scalable and explainable approach for developing intelligent and adaptive self-management tiny systems. Moreover, we argue that TinyAC systems require self-adaptive features to handle problems that may occur during their operation. Finally, we identify gaps, discuss existing challenges and future research directions.

Problem

• Applying autonomic computing principles to resource-constrained IoT devices
• Bridging bottom-up intelligence with top-down guidance for tiny systems
• Developing self-adaptive features for operational problem handling

Innovation

• Hybrid approach combines bottom-up TinyML intelligence
• Integrates top-down LLM guidance for system management
• Enables self-adaptive features for operational problem handling