A Method of Distributed Multi-Sensor Data Fusion under Privacy Protection

Multi-sensor data fusion technology has great potential to effectively solve the data compatibility problems inherent in smart devices. However, current methodologies lack high-performance delivery while neglecting key issues surrounding data privacy. To address these shortcomings, this paper presents an innovative integration of federated learning into multi-sensor data fusion. Leveraging the Gated Recurrent Unit (GRU) algorithm, the federated learning local model is designed to tackle single-sensor data fitting challenges. For the first time, a three-dimensional, time-space parallel running multi-sensor fusion method is proposed. This unique approach exhibits superior convergence performance whilst effectively ensuring the privacy of user data. The soundness and fundamental premise of the proposed methodology is supported by experimental results, demonstrating its robust superiority.