An Adaptive Unloading Algorithm of Computing Tasks Based on Edge Cloud Collaboration Scenario for Internet of Things

In recent years, the analysis of big data in the realm of the Internet of Things (IoT) has garnered increasing attention. Several cloud platforms offer pre-trained machine learning models to comprehend IoT data. Nonetheless, to utilize these cloud services, the transmission of personal data is necessitated, and network issues may impede clients from obtaining timely analytical results. To address these challenges, the migration of data and analytical tasks to edge platforms is gaining momentum. However, the majority of edge devices lack the capacity required to process and train on vast datasets. In response to the substantial computational demands in IoT scenarios, we propose an adaptive task offloading algorithm within an edge-cloud collaborative system. Leveraging Deep Belief Networks (DBN) technology, data is hierarchically processed, and the performance of both computing devices and the edge-cloud collaborative system’s network is meticulously assessed. Our proposed adaptive task offloading algorithm is underpinned by a meticulously designed processing time model. Experimental results unequivocally demonstrate the algorithm’s pronounced advantages in enhancing the overall response time within the context of edge-cloud collaboration.