Modeling And Analysis Of The Dynamic Behavior In Crowd Sensing Systems

The marriage of crowdsourcing and a large population of smartphones creates mobile crowdsourcing,an emerging paradigm for recruiting a large number of smartphones for performing various tasks.Both industry and academic have drawn extensive attention on it.With powerful capability on data computation and communication,the mobile devices make the tasks easier to be taken.Especially,they can sense the environment with various sensors.A variety of crowdsourcing applications have revolutionized many sectors of our life,such as environment monitoring,intelligent transportation.A single mobile crowdsourcer(i.e.,a crowdsourcing system)typically consists of two parts: a crowdsourcing platform residing on the cloud and a population of smartphone contributors.The crowdsourcer purchases sensing services from smartphone contributors,who consume their own resources to accomplish the sensing task.Smartphone contributors will join a crowdsourcer unless they can receive satisfying rewards.When mobile crowdsourcing becomes the mainstream,there will be more applications and systems.We envision that there will be a mobile crowdsourcing market which consists of many crowdsourcers and a large population of smartphone contributors.Smartphone contributors have free choices on crowdsourcers and share the limited budget offered by them.Crowdsourcers will share the limited sensing service provided by smartphone contributors.In order to better fit the gap between needs of crowdsourcers and capabilities of smartphone contributors,the interactive behavior of multiple crowdsourcers and smartphone contributors needs to be understood.First,we model the competitive behavior(i.e.,adjust the price paid)of multiple crowdsourcers.Each crowdsourcer has to attack sufficient smartphone contributions to accomplish its functionality,achieving the highest profit.However,the sensing service is limited.The issue of competition arises.We use a dynamic non-cooperative game to formulate the competition among crowdsourcers and Nash equilibrium is the solution of the game.We propose a distributed learning algorithm converging to the Nash equilibrium.Second,we model the collaborative behavior(i.e.,adjust the price paid)of multiple crowdsourcers.When the game is played multiple time,crowdsourcers may cooperate with each other considering the optimal long term profit.However,a crowdsourcer has incentive to deviate from the collusion to increase its profit.We formulate the interaction among multiple crowdsourcers as a repeated game and analyze the condition to maintain collusion.Third,we model the bilateral competition among crowdsourcers and smartphone contributors.Crowdsourcers compete for the limited sensing service and smartphone contributors compete for the limited reward.To be more practical,we consider the bounded rationality of smartphone contributors.We formulate the dynamic behavior of smartphone contributors as an evolutionary game and present an algorithm for the implementation of evolution process.To model the competition among crowdsourcers,we use a non-cooperative game.We prove the existence of Nash equilibrium and propose an iterative algorithm to achieve the Nash equilibrium.Combining the above three research points and the current work which models the behavior of smartphone contributors competing for the reward offered by a single crowdsoucer,we construct a framework to study the dynamics of multiple crowdsourcers and smartphone contributors in mobile crowdsourcing.