Consensus Filter for Distributed Sensor Networks With Unknown Colored Noise

Consensus filtering algorithms are commonly used for distributed state estimation. In addressing the challenge of unknown colored noise in distributed sensor networks, we propose a consensus filter based on Gaussian process regression (GPR). Our method employs GPR models to learn the parameters of process and measurement noise for each node in the network. These learned noise parameters are then utilized to extend the states for precise state estimation. To improve computational efficiency, our algorithm iterates solely over the original state. Simulation results validate the effectiveness of our approach in handling colored noise systems.