FedLoc: Federated Learning Framework for Data-Driven Cooperative Localization and Location Data Processing

This SPS webinar will introduce a novel data-driven cooperative localization and location data processing framework, called FedLoc, in line with the emerging machine learning and optimization techniques. We first review two widely used learning models, namely the deep neural network model and the Gaussian process model, show their connections, and introduce various distributed model hyper-parameter optimization schemes that can be adopted to implement the federated learning. To give a complete picture, we then introduce some other vital ingredients of the FedLoc, including privacy protection schemes, wireless network infrastructures, etc. Lastly, we demonstrate various popular use cases covering a broad range of location services, including collaborative static localization/fingerprinting, indoor target tracking, outdoor navigation using low-sampling GPS, Spatio-temporal wireless traffic data prediction, etc. All the use cases aim at collaboratively building more accurate location services without sacrificing user privacy, particularly sensitive information related to their geographical trajectories. Future research directions will be discussed at the end of this webinar.