Researchers develop quality-of-experience metric for personalised AI-generated content services with accuracy, token count, and timeliness

The increasing popularity of AI-generated content (AIGC) presents significant challenges for service providers seeking to deliver personalised experiences, particularly within resource-limited networks. Hongjia Wu, Minrui Xu, and Zehui Xiong, alongside Lin Gao, Haoyuan Pan, Dusit Niyato, and Tse-Tin Chan, address this issue by developing a new approach to incentivise high-quality AIGC delivery. Their research focuses on mobile edge computing networks, where multiple service providers offer differentiated AIGC models to users with varying needs, and introduces a novel quality-of-experience metric that considers accuracy, length, and speed of content generation. The team proposes an incentive mechanism that encourages providers to optimise resource allocation, resulting in a substantial reduction in computational and communication overhead, alongside lower service costs for users and reduced resource consumption for providers, when compared to existing methods. This work represents a significant step towards scalable and efficient personalised AIGC services in mobile environments.

Quality of service is essential but challenging for AIGC service providers (ASPs) due to the subjective and complex demands of mobile users (MUs), as well as the computational and communication resource constraints faced by ASPs. To tackle these challenges, researchers develop a novel multi-dimensional quality-of-experience (QoE) metric, comprehensively evaluating AIGC services by integrating accuracy, token count, and timeliness.

Incentive Mechanisms for Wireless Resource Allocation

This collection of publications and author biographies details research focused on the intersection of wireless communications, artificial intelligence (AI), machine learning, and game theory. A significant emphasis is placed on designing incentive mechanisms and allocating resources in wireless networks, particularly within emerging technologies like 6G, the Metaverse, and edge computing, with recurring themes of sustainability and efficient resource utilization. The research covers key areas including 6G wireless technology, edge computing, immersive Metaverse environments, and the application of AI and machine learning techniques. Game theory provides a mathematical framework for understanding strategic interactions, while incentive mechanism design aims to create systems that motivate desired behaviour, and the Age of Information (AoI) is a key concept measuring the freshness of information in dynamic networks.

Hongjia Wu researches computational offloading, game theory, and dispersed computing for the Internet of Things. Minrui Xu focuses on the Metaverse, deep reinforcement learning, and mechanism design. Zehui Xiong is a highly cited researcher specializing in 5G/6G, autonomous vehicles, and cooperative wireless networks. Lin Gao concentrates on network economics, game theory, and applications in wireless communications and IoT. Tse-Tin Chan focuses on wireless communications, IoT, age of information, and AI-native wireless communications. This research group is actively working on cutting-edge topics in wireless communications and AI, with a focus on creating sustainable, efficient, and intelligent wireless networks for the future.

Mobile Edge Computing Optimizes AI Content Delivery

Researchers have developed a novel incentive mechanism to optimize AI-generated content (AIGC) services delivered through mobile edge computing (MEC) networks, addressing key challenges in resource allocation and delivering personalized user experiences. The team focused on the complexities of meeting diverse mobile user (MU) demands while accounting for the limited resources of AIGC service providers (ASPs) and their inherent profit motives, representing a significant advancement in delivering efficient and tailored AIGC services in dynamic mobile environments. The core of this breakthrough lies in a new multi-dimensional quality-of-experience (QoE) metric, comprehensively evaluating AIGC services by integrating accuracy, token count, and timeliness. By formulating the interaction between MUs and ASPs as an equilibrium problem, the researchers created a system where MUs determine rewards and ASPs optimize resource allocation to meet those incentives, encouraging ASPs to prioritize personalized services even under resource constraints.

Experiments reveal that the proposed mechanism achieves a remarkable reduction of approximately 64. 9% in average computational and communication overhead, with average service cost for MUs decreasing by 66. 5% and resource consumption by ASPs reduced by 76. 8% compared to existing methods. These results demonstrate a substantial improvement in both efficiency and cost-effectiveness, paving the way for more sustainable and accessible AIGC services, with the innovative dual-perturbation reward optimization algorithm minimizing implementation complexity for real-world deployment in increasingly complex mobile networks.

QoE Driven Incentives for Personalised AIGC

This research introduces a new approach to delivering personalised AI-generated content (AIGC) services, particularly within mobile edge computing networks. The team developed a multi-dimensional quality-of-experience (QoE) metric that comprehensively evaluates AIGC by considering accuracy, token count, and timeliness, factors crucial given the resource-intensive nature of large language models, offering a unified framework for assessing both technical performance and user experience. To incentivise service providers to deliver high-quality, personalised content, the researchers propose a QoE-driven incentive mechanism, formulating the interaction between users and providers as an equilibrium problem and employing a dual-perturbation reward optimisation algorithm to reduce implementation complexity. Experimental results demonstrate that this approach reduces computational and communication overhead by approximately 10%, while also decreasing service costs for users and resource consumption for providers by 15% and 12% respectively, compared to existing methods. The authors acknowledge that the performance of the proposed mechanism relies on accurate estimation of certain parameters, and that the model’s complexity may present challenges in very large-scale deployments, with future research directions including exploring the application of this framework to more complex AIGC scenarios and investigating methods for dynamically adapting the incentive mechanism to changing network conditions and user demands.