Recently, A research team led by Professor Zhao Jia from the School of Computer Technology and Engineering and the Artificial Intelligence Research Institute at our university has achieved a groundbreaking advance in deep reinforcement learning (DRL). Their paper, titled "Modular Deep Reinforcement Learning for Multi-Workload Offloading in Edge Networks," has been accepted for presentation at IJCAI (International Joint Conference on Artificial Intelligence), a premier global AI conference. The first author of this paper is Associate Professor Ke Hongchang, followed by Associate Professor Ding Yan, Assistant Researcher Pan Lin, and Experimental Engineer Chen Yang, with Professor Zhao Jia serving as the corresponding author.
The paper introduces an innovative DRL-MWF (Deep Reinforcement Learning with Modular Weighted Fusion) algorithm designed to tackle complex task offloading challenges in edge computing networks. Addressing the demands of dynamic environments, DRL-MWF employs a modular weighted fusion architecture to efficiently adapt to diverse computational workloads. The research team made innovative contributions by integrating reinforcement learning state representation, weighted policy function correction, and importance sampling-based prioritized experience replay, which significantly enhanced the algorithm's performance. Experimental results demonstrate that DRL-MWF outperforms existing methods across multiple metrics, showcasing its strong potential for future edge computing systems. This research not only advances AI applications in complex network environments but also provides new insights for improving edge computing efficiency.
Problem Scenario Demonstration
Key Structure of the Proposed Method
The International Joint Conference on Artificial Intelligence (IJCAI) is one of the most influential and prestigious conferences in the field of AI. Since its inception in 1969, it has successfully hosted 33 iterations, playing an irreplaceable role in the dissemination, development, and advancement of AI concepts. Recognized as a CCF-A (Class A) conference by the China Computer Federation (CCF)—a designation marking the highest tier of international academic venues—IJCAI represents the pinnacle of AI research and future trends. This is the first time that CIT has secured acceptance in the IJCAI Main Track, signifying a groundbreaking breakthrough in AI research and earning recognition from the international academic community.
