Research On Key Issues In Participatory Sensing

With the rapid development of communication technology and microelectronics,mobile phones have evolved from merely communication devices to a smart terminal integrated with computing,sensing and communication capabilities.As more and more people are carrying smart devices,the latest developments and ubiquitous existence of mobile terminals has prompted researchers to propose a new sensory data collection method,called "Participatory Sensing".The basic idea of participatory sensing is to allow ordinary citizens to collect their environmental data using their smart phones,and the great power of it comes from the large number of participants involved.The incentive mechanism of participatory sensing is one of the most important research issues,for(a)it incurs real cost for participants to collect data using their energy-limited devices and upload them using their network bandwidths.(b)different from traditional sensor network,participants own their devices,and nobody but only participants can decide when and where to collect sensory data.However,incentives are frequently used to influence the behavior of them and encourage participants to collect data as the tasks require,thereby further improving the quality of data collection.On the other hand,incentive mechanism is closely related to other important research issues in participatory sensing,such as participant selection,energy optimization and improving the quality of sensing results.Therefore,this thesis mainly focuses on incentive mechanism and its related issues,and makes the following contributions:1.Propose a multi-task oriented participant selection algorithm,which is based on the basic flow of existing reversed auction incentive mechanism.We propose a QoI metric for task's data collection,and predict the data contribution estimation of participants based on their initial locations.We further model the participant selection problem as a multi-objective optimization problem,and propose a greedy-based algorithm to achieve sub-optimal solutions.The proposed algorithm can collect more satisfactory data for multiple tasks,and thus achieve better sensing results.2.Propose an two-step energy-efficient event-boundary detection method.In the first step,we require random data collection by small amount of participants,and use min-cut algorithm to find the coarse boundary of events.In the second step,we propose an energy efficient participant selection method to collect most important data for boundary detection.Our proposed method can accurately find the fine-grained boundary of events,meanwhile largely reduce the disturbance to participants for the energy consumption in data collection.3 Propose a accuracy-oriented participant selection method based on the reversed auction incentive method.Such method considers not only how to collect the maximal amount of sensory data by limited budget,but also how to make the collected data uniformly distributed,so that the accuracy of collected data after interpolation can be maximized.We also propose a entropy-based method to evaluate the distribution uniformity of collect data.4 Propose a novel incentive negotiation procedure.The proposed procedure query all participants with a price offer,instead of letting them upload their bid prices.The price offer is calculated according to the temporal-spatial distribution of participants,as well as their incentive expectation which can be updated online.Compared with the reversed auction procedure,our proposed method can not only distribute incentives among participants more fairly and reduce the heavy overload of incentive negotiation,but also improve the overall sensing quality after data interpolation.