Regresar

Distributed Consensus-based Kalman Filter under Limited Communication

Abstract:

In this work, we consider a distributed estimation problem in a communication-constrained environment. To address the limited communication challenge, we present a fully distributed Kalman filtering algorithm in which each agent shares a compressed version of its estimated state information with its neighboring nodes. In the proposed algorithm, we explicitly compute the estimation error covariances of each node in a distributed manner based on the consensus filter using the compressed estimates. An intuitive finding is that for a specific mid-tread quantization function, compared with the uncompressed distributed Kalman consensus filter, the state estimates obtained with the quantized Kalman consensus filter are significantly similar; however, the estimation error covariances are noticeably different. We validate the theoretical results using simulations.