基于强化学习的车联网隐私保护和资源优化策略

Privacy Protection and Resource Optimization Strategies for Internet of Vehicles Based on Reinforcement Learning

车联网使人们的生活更加智能便捷,但是数据隐私问题和车辆有限的本地资源状况严重阻遏了其未来发展。为了解决上述问题,提出了基于联邦学习和强化学习的车联网隐私保护和资源优化策略。系统中的隐私保护模块通过全局模型下发和上传的方式代替传统的原始数据交互,解决了传统算法的数据和隐私泄露问题;资源优化模块通过建立强化学习决策模块对系统资源进行评估,选择具有最长远效益的决策,优化有限的系统资源。仿真结果表明:本文中提出的基于强化学习选择算法相较于联邦学习交互算法,系统总能耗降低77%以上;相较于传统算法,基于强化学习选择算法传输数据量降低了98.49%。

The applications of Internet of Vehicles make live more intelligent and convenient,while data privacy and limited resources of vehicles severely hinder future development.Privacy protection and resource optimization strategies for Internet of Vehicles is proposed based on federated learning and reinforcement learning.The privacy protection module solves the problem of data and privacy leakage of traditional algorithm by downloading and uploading global model instead of raw data.The resource optimization module evaluates system resources by establishing an agent based on reinforcement learning which could make the best long-term benefit decision to optimize system resources.The simulation results show that the reinforcement learning-based selection algorithm can reduce the system energy consumption by up to 77%compared with the federated learning interactive algorithm.Compared with the traditional algorithm,it can reduce the amount of transmission data by up to 98.49%.