Distributed Mean-Field-Type Filters for Traffic Networks

Abstract

Traffic surveillance plays an important role in the development of a smart city, and it is a fundamental part in many applications such as security monitoring and traffic analysis. People are thrilled by the abundant data generated from the huge traffic networks but have difficulty in using them. In this paper, we propose a distributed mean-field-type filtering (DMF) framework to handle those noisy, partial-observed, and high-dimensional data. The filter incorporates a mean-field term into the system model and decomposes the state space into highly independent parts; filtering is performed in each part and then integrated. Our approach iterates through four operations: sampling, prediction, decomposition, and correction. Theoretical analysis provides a linear bound for the global error, which is independent of the network's cardinality. We implemented DMF in aircraft and vehicle tracking scenarios. Performance evaluation on synthetic and real-world data demonstrates the advantage of our approach over traditional mean-field free filters.