Study On Incentive Mechanisms For Multi-Services In Edge Networks

Edge Computing is an emerging computing paradigm that encourages resource sharing and inter-neighbor assistance by devices,which contributes surplus resources to improve the performance of others.Content Caching,Computation Offloading,and Edge Intelligence are typical service paradigms in edge networks.Existing work has fully investigated how the above paradigms and technologies can be used to enhance the experience of mobile device users.However,the willingness of massive devices to construct edge networks and contribute resources to edge services is an essential factor in determining whether edge services are well experienced.Therefore,how to design a reasonable and feasible incentive mechanism for multiple services in the edge network is a fundamental issue.Unlike the traditional environment,the huge number of devices,wireless connection,dynamic access,and the economics of devices in the edge network dictate that key factors such as heterogeneity,dynamism,and information asymmetry need to be considered when designing incentive mechanisms in the edge network.Efficient and feasible incentives for different services can motivate mobile devices to participate in the edge network composition and to put their best efforts into the services.This will not only build a virtuous circle of sharing and mutual assistance at the edge,but promote the commercialization and further development of the edge model.Therefore,the study of incentive design for multi-services in edge network has both theoretical and applied value.This paper studies the design of incentive mechanisms for content services,computing services,and training services in edge networks and main work in this paper include the following three aspects:1.We study the problem of designing incentives in heterogeneous time-varying edge content markets to maximize the benefits of content providers.First,we characterize the time sensitivity of edge content through multiple dimensions such as content popularity,content quality,and content freshness.Then we use the probability of receiving content services to characterize content audience heterogeneity by introducing information economics theory,and constructing a model of edge content and users.Second,we divide the content market into a monopolistic environment and an open environment in terms of whether the content is exclusively supplied,so as to characterize market uncertainty and discuss the design of incentives in each environment.In the monopolistic environment,we construct a two-stage Stackelberg Game model to derive a pricing scheme and content update strategy that maximizes the benefits of content suppliers.To cope with the uncertainty of the open environment,we propose a reinforcement learning-based incentive mechanism to obtain the optimal pricing strategy for content providers.Finally,the experiment results demonstrate that our designed incentive mechanism is efficient in helping content providers to maximize their revenue in different edge content markets.2.We propose the incentive mechanism for computing resource providers in heterogeneous edge computing services to maximize the benefits of resource providers while ensuring user performance.First,without considering the competitive nature of computing services,we introduce the microeconomic theory and use a market pricing model and a supply-demand relationship model to model the interests of both sides of the transaction,so as to ensure the interests of users while portraying the incentive problem as a profit maximization problem.Second,considering the competitive nature of computing services,we design an effective profit-maximizing multi-round auction mechanism to match buyers and sellers and determine the final price.By introducing 'Bid Performance Ratio','Price Performance Ratio' and the second-price auction model,we theoretically demonstrate that the designed mechanism satisfies individual rationality,incentive compatibility,and computational efficiency constraints.Finally,using a non-competitive environment as a benchmark,The simulation results show that the auction mechanism guarantees the computational effectiveness of the mechanism while maximizing the profitability of the computational resource provider and safeguarding the computational needs of users.3.We study the problem of designing incentive mechanisms in heterogeneous edge intelligence collaborative training scenarios to promote edge users to participate in edge training and put their best efforts into it.First,we characterize agent heterogeneity by introducing training contribution and principal heterogeneity through time-marginal gains,and then comprehensively discuss the impact of role heterogeneity on training gains.Second,through a comprehensive analysis of incentive design in the static and complete information environment,we reveal the importance and challenges posed by considering dynamism and information asymmetry for incentive design.Finally,in the dynamic incomplete information scenario,we propose a scalable and deployable edge intelligence incentive mechanism based on contract theory,which combines individual rationality and incentive compatibility constraints to provide sufficient conditions for contract feasibility at the theoretical level.The experiments indicate that the designed mechanism brings positive benefits to both sides of the co-training,while the training effect under the incentive is somewhat better than the existing methods.