Data Sharing Strategy For Autonomous Drving In Internet Of Vehicles

The Internet of Vehicles(IoV)is driving continuous growth in the automotive industry.The IoV can empower efficient data transmission between vehicles,supporting the sharing of sensing data.Data sharing can improve the quality of intelligent model training and inference data,ensuring the reliability of decisionmaking in autonomous driving system.This paper focuses on how to design an data sharing strategy in IoV that achieves efficiency and privacy.Efficient and secure data sharing strategy in IoV remain challenging,which can be categorized into two aspects:1)The competition for limited wireless communication resource may result in decision coupling between vehicles and the inefficiency of distributed data sharing;2)Data sharing between vehicles can cause privacy leakage.Existing distributed learning methods and privacy protection techniques may incur significant overhead and are difficult to ensure security in highly mobile vehicular networks.This paper focuses on data sharing efficiency and security in IoV.The main works and innovations are as follows.To address the challenges of coupling and inefficiency for data sharing in IoV,this paper proposes a distributed graph-based optimization scheme to efficiently share local sensing data between vehicles.We show that maximizing system throughput is a submodular optimization problem on link establishment decisions.Then a three-layer network model is formulated to describe the problem of data transmission and processing.By transforming the network model,a minimum cost and maximum flow problem in graph theory is obtained.Simulation results show that this method improves data sharing efficiency compared with existing methods.To address the challenges of data privacy leakage for federated learning in IoV,this paper proposes a secure and efficient hierarchical decentralized learning scheme.The network-level masking mechanism and consensus matrix optimization for signaling-efficient implementations in IoV are proposed.The network-level masking can eliminate the masking repairing requirements for the inter-fog vehicle handover and is proved to be canceled via distributed consensus.The proposed scheme is evaluated for model accuracy,communication overhead,and defense effectiveness.In conclusion,in view of the efficiency and security issues of data sharing in IoV,this paper proposed a distributed graph-based efficient data sharing optimization scheme and a secure and efficient hierarchical decentralized learning scheme.The proposed schemes can achieve efficient and secure data sharing decisions in vehicular networks.