Data-Driven Approaches for Cooperative Edge Intelligence: Opportunities and Challenges

Cooperative edge intelligence is regarded as a key technology for cloud-edge-device scenarios, enabling immersive intelligence for computing, communication, and network resource management. With the increasing volume and variety of data generated by diverse devices and applications, it is critical to fully exploit this data and explore data-driven cooperative edge intelligence in cloud-edge-end scenarios. In this paper, we propose data-driven approaches for cooperative edge intelligence (DCEI) by characterizing user behavior patterns to improve system performance. Additionally, we discuss the key challenges and benefits of data-driven approaches in cloud-edge-device scenarios. To explore their potential, we elaborate on three data-driven cooperative decision-making approaches by mining user behavior patterns in different scenarios: task allocation and load balancing, mobility-aware service migration decision-making, and federated learning for distributed data training. Finally, we present a case study on data-driven approaches by designing a mapping scheme with reinforcement learning (DDRL), which leverages user behavior patterns to map RRH and BBU. Extensive data-driven experiments corroborate the efficacy of DDRL compared to state-of-the-art schemes.