Fuzzy chance constraint-based task offloading for MEC under uncertainties with joint multi-network collaboration and privacy preservation

Abstract While multi-access edge computing (MEC) enables flexible task offloading, the inherent dynamism and complexity of real-world environments, accompanied by various uncertainties, exacerbate the challenges in offloading optimization. However, the limitation of existing approaches generally lies in their reliance on idealized assumptions or their inability to balance efficiency with uncertainty adaptability, which restricts their practical applicability. In this paper, we study the task offloading problem for MEC under uncertainties with joint consideration of multi-network collaboration and privacy preservation. Firstly, we establish a fuzzy optimization model to formulate the task offloading problem under uncertainties and prove its NP-hardness. The established model leverages multiple wireless networks for collaborative task data transmission and incorporates fuzzy privacy entropy to guarantee task privacy preservation. To address the inherent uncertainties in the task offloading model, we transform it into a fuzzy chance-constrained optimization formulation with predefined confidence levels. Secondly, to derive the offloading and allocation strategies from the transformed optimization model, we propose a hybrid task offloading algorithm that combines an improved adaptive genetic algorithm, Monte Carlo simulation, and neural networks. Our algorithm adopts a hybrid optimization architecture in which the adaptive genetic algorithm enhances strategy exploration through improved genetic operations, whereby the neural networks are embedded as fast proxies within evolutionary iterations to replace computationally intensive Monte Carlo simulations for solution evaluation. Finally, experimental results demonstrate that the proposed hybrid offloading algorithm surpasses existing algorithms and attains an average reduction of at least 23.24% in the objective value.