Research On Key Technologies Of Crowdsensing For Internet Of Vehicles

In view of the rapid development of vehicular intelligence and networking,vehicular crowdsensing has recently become a promising perception and computing paradigm.Vehicular crowdsensing system has advantages in terms of convenient deployment and maintenance,fast large-scale coverage and scalable applications.With these preponderances,it can be employed in various applications and bring significant benefits to Intelligent Transportation Systems(ITS).To satisfy the performance requirements of ITS applications,a huge amount of sensing data are generated continuously.However,it is challenging to schedule the sensing resource as well as collect and process the massive sensing data.The emergence of edge computing and Internet of Vehicles(IoV)has brought new opportunities for the rational task allocation,rapid resource scheduling and reliable data collection in vehicular crowdsensing.Combining edge computing,vehicular intelligence and IoV technologies,vehicular crowdsensing is an important way to realize autonomous driving and intelligent transportation.In recent years,researchers have carried out some research in the field of vehicular crowdsensing.Due to the uncertainty of sensing nodes in terms of quantity and distribution and the diversified demands of related applications,there are still many factors affecting the service quality of vehicular crowdsensing that need to be further discussed.At present,the application of vehicular crowdsensing still faces the following three challenges in practical scenarios:1)considering the high randomness of sensing nodes,how to effectively select and schedule qualified participants from a large user pool to improve the sensing coverage and accuracy;2)considering the high redundancy of sensing data,how to reduce the ineffective data collection while ensuring the sensing accuracy;3)considering the low real-time performance of sensing process,how to ensure the reliability and timeliness of data transmission in the process of sensing data uploading and service data dissemination in the light of the specific characteristics of vehicular crowdsensing applications.To solve these challenges,this dissertation conducts research on the key issues such as sensing nodes selection and scheduling,vehicle sensing and data aggregation,data uploading and dissemination,aiming to provide ubiquitous,efficient and realtime vehcular crowdsensing services.The main contributions are summarized as follows:1.Research on incentive-aware non-dedicated vehicle recruitment scheme based on edge intelligenceIn order to realize rational sensing task allocation,this dissertation introduces an edge computing based non-dedicated vehicle recruitment and incentive scheme,where non-dedicated vehicles refer to widely distributed taxis and buses with low recruitment overhead.In particular,we first design an incentive mechanism to motivate cooperation among the edge server and the nondedicated vehicles,and apply the Nash bargaining theory to obtain the optimal cooperation decision.Then,we formulate the participant recruitment as an optimization problem,and prove that it is NP-hard.To address the problem,an effective heuristic algorithm based on submodular optimization is proposed.The scheme encourages and selects vehicles with higher capability and reputation value to participate in the sensing tasks,thus effectively relieving the uneven distribution of sensing resources and improving the sensing coverage and accuracy.2.Research on high-quality dedicated vehicle scheduling schemeIn order to further improve the coverage of sensing data,this dissertation designs a cooperative data collection scheme,where non-dedicated and dedicated vehicles cooperate to perform crowdsensing tasks under the guidance of the edge server.Dedicated vehicles are often directly managed by traffic authorities,which are specifically moved to the assigned locations to conduct tasks.We first propose an optimization objective based on the entropy theory to evaluate the spatiotemporal evenness of collected data accurately.Then,the offline and online scheduling mechanisms for dedicated vehicles are proposed based on the sensing results of non-dedicated vehicles.Specifically,the offline mechanism can achieve near-optimal performance with complete trajectory information,which is applicable for relatively static environment.The online scheme is robust and light-weight and do not hinge on any unpractical assumptions,exhibiting distinct advantages in highly dynamic traffic conditions.By promoting the close cooperation between the two types of vehicles,the spatiotemporal coverage evenness of sensing data is greatly enhanced,thus improving the quality of vehicular crowdsensing applications.3.Research on efficient data collection scheme with low redundancyIn order to realize efficient data collection with low redundancy,this dissertation avoids the ineffective data generation and transmission from vehicular sensing process and data aggregation process respectively.In vehicular sensing process,an online sensing parameter adjustment algorithm is proposed to guide the actions of intelligent vehicles based on the real-time traffic condition.In data aggregation process,an adaptive clustering algorithm is designed to form adaptive vehicular clusters and compress the redundant data in clusters by taking into account the cluster stability and the communication reliablility.This scheme can effectively capture the interaction between vehicular sensing process and data aggregation process,and achieve a balance between sensing accuracy and communication cost while ensuring reliable and timely data uploading.4.Research on reliable data transmission scheme based on edge computingIn order to realize real-time and reliable data transmission,this dissertation proposes sensing data uploading and service data dissemination schemes based on edge computing respectively.In terms of sensing data uploading,we jointly optimize the problems of network selection and traffic allocation in the scenario of highly dynamic heterogeneous networks based on particle swarm optimization and convex optimization theories.The scheme utilizes multi-access edge computing and fully combines the advantages of different wireless networks,effectively improving the network utilization and user satisfaction.In terms of service data dissemination,a relay-based dissemination scheme for large-volume service data is proposed based on edge computing,vehicle-to-everything(V2X)communication and erasure coding technologies.The scheme solves the challenges posed by high mobility of vehicles and limited coverage of roadside base stations,while improving the success rate of large-volume service data transmission.The above two schemes jointly guarantee the timeliness and reliability of data transmission process in vehicular crowdsensing.In summary,to solve the challenges existing in real scenarios,this dissertation proposes a series of schemes that can provide ubiquitous,efficient and real-time vehicular crowdsensing services by combining edge computing and IoV technologies.The effectiveness of the proposed schemes is verified both theoretically and experimentally,which can lay the substantial theoretical and technical foundation for the vehicular crowdsensing applications.