QIC-MOEA: A Context-Aware Quantum-Inspired Evolutionary Framework for Self-Adaptive Pervasive Systems

Abstract

Everyday computing is rapidly shifting toward pervasive systems, but real-world edge environments remain incredibly volatile. For these self-adaptive devices to function, they must constantly balance strictly conflicting goals—like saving power versus processing critical data instantly. Traditional methods struggle here: rule-based systems lack flexibility, and reinforcement learning either takes too long to compute or exploits mathematical loopholes (reward hacking). To solve this, we introduce the Quantum-Inspired Contextual Multi-Objective Evolutionary Algorithm (QIC-MOEA), a context-aware framework designed for real-time adaptation. By leveraging a probabilistic Q-bit representation, our approach explores multiple adaptation strategies simultaneously, completely bypassing the heavy computational drag of classical algorithms. We pair this with Context-Weighted Rotation Gates to dynamically scale reactions to environmental shocks, and enforce strict Quality of Service boundaries to eliminate reward hacking. Tested in a simulated healthcare scenario, QIC-MOEA significantly outperformed Q-Learning and Fuzzy Logic baselines. It reduced energy consumption by over 92% and latency by up to 36%, while improving context appropriateness by more than 313%. Crucially, QIC-MOEA executed these complex multi-objective adaptations in just 11.53 milliseconds, proving that quantum-inspired decision-making is highly feasible for resource-constrained edge hardware.