Abstract
In the Internet of Vehicles (IoV), Federated Learning (FL) provides aprivacy-preserving solution by aggregating local models without sharing data.Traditional supervised learning requires image data with labels, but datalabeling involves significant manual effort. Federated Self-Supervised Learning(FSSL) utilizes Self-Supervised Learning (SSL) for local training in FL,eliminating the need for labels while protecting privacy. Compared to other SSLmethods, Momentum Contrast (MoCo) reduces the demand for computing resourcesand storage space by creating a dictionary. However, using MoCo in FSSLrequires uploading the local dictionary from vehicles to Base Station (BS),which poses a risk of privacy leakage. Simplified Contrast (SimCo) addressesthe privacy leakage issue in MoCo-based FSSL by using dual temperature insteadof a dictionary to control sample distribution. Additionally, considering thenegative impact of motion blur on model aggregation, and based on SimCo, wepropose a motion blur-resistant FSSL method, referred to as BFSSL. Furthermore,we address energy consumption and delay in the BFSSL process by proposing aDeep Reinforcement Learning (DRL)-based resource allocation scheme, calledDRL-BFSSL. In this scheme, BS allocates the Central Processing Unit (CPU)frequency and transmission power of vehicles to minimize energy consumption andlatency, while aggregating received models based on the motion blur level.Simulation results validate the effectiveness of our proposed aggregation andresource allocation methods.