Research On Resource Allocation For Parked Car Roadside Unit-Assisted Air-ground Vehicular Networks

With the promotion and application of emerging vehicular network services,unprecedented challenges have been put forward,particularly for the efficient processing of extensive delay-sensitive and computation-intensive tasks.By integrating with mobile edge computing techniques,roadside units(RSUs)are able to provide seamless coverage and edge computation capability for vehicle users(VUs)with high mobility,which can be envisioned as a promising scheme.Owing to the high deployment and maintenance cost of RSUs,it is appealing to adopt idle parked cars so as to assist RSUs,thereby achieving low-cost and resource-efficiency vehicular networks.Additionally,the inclusion of Unmanned Aerial Vehicles(UAVs)-assisted air-ground vehicular network has high flexibility,which can provide communication coverage,edge computing and content delivery services for VUs.However,both network nodes and terminals of the air-ground vehicular network possess mobility characteristics.How to jointly manage multi-dimensional resources for the highly dynamic network,thereby reducing the response delay of user requirements,is a key scientific issue.To this end,we investigate the resource allocation problem in parked car-assisted air-ground vehicular networks,and propose a multi-dimension resource allocation approach to minimize the system latency,thereby satisfying the dynamic and diverse quality of service requirements.In summary,the main contributions are as follows:(1)Firstly,to cope with the high deployment cost of edge servers(ESs)in airground vehicular networks,we elaborate on the parked car roadside units(PCRSUs)recruitment and the edge UAV deployment issue.Specifically,we take into account the parking duration metric and propose a novel judgment criterion for PCRSU recruitment to efficiently utilize the PCRSU resources.Furthermore,in order to take advantage of the flexibility of UAVs,heuristic harris hawks optimization(HHO)algorithm is applied to dynamically deploy UAVs,thereby enhancing the coverage of ground network.Extensive simulation is conducted to demonstrate that the proposed scheme can accurately select and recruit PCRSUs while optimizing the UAV deployment to improve the system overall performance.(2)Secondly,to meet the dynamic and diverse quality of service requirements of air-ground vehicular network users,we design the response delay minimization edge content caching and delivery mechanism.First of all,a context-awareness content personalization recommendation algorithm is developed,which enables ESs to predict the potential user demands based on the historical search data as well as point-ofinterest(POI)region types,and hotspot contents are cached at ESs in advance.Then,the caching update strategy is designed by jointly considering the number of content access and the recent access time.Both theoretical analysis and simulation results reveal that the proposed approach can accurately predict the request of VUs and execute efficient caching update,while ensuring the content delivery performance regardless of the fluctuations of network parameters,e.g.,the number of users and recommended content.In addition,the proposed approach outperforms the benchmark method by 50%in terms of content hit ratio.(3)Last but not least,for the sake of tackling the information uncertainty and resource constraint issue during computation offloading,we propose a Three-Tier edge server based-Vehicular Computing(T2VC)system model,and investigate the task offloading and resource allocation solution.Afterwards,a multi-armed bandit(MAB)based Queue of Energy Volatility-Aware Upper Confidence Bound(QEVA-UCB)task offloading algorithm is developed,which enables VUs to select optimal offloading candidate according to only local information.Moreover,network power control and bandwidth allocation are optimized by leveraging Lagrangian multiplier method,thereby improving system throughput and reducing transmission delay.Numerical results unveil that the proposed approach can rapidly and precisely learn the offloading policy as well as achieving queue of energy volatility awareness,while reasonably scheduling network resources to decline the processing latency of delay-sensitive and computation-intensive tasks.Particularly,the proposed solution reduces end-to-end delay by 30%compared to the baseline.