The Kalman filter as an effective tool to solve the state estimation problem for linear dynamic systems can be derived from a generalized perspective by applying the sum-product message passing over a factor graph. This viewpoint encourages us to visualize the state estimation problem over a network where all the nodes aspire to obtain consensus-based state estimates of a dynamic system by collecting sequential measurements over time. In this work, nodes process in a distributed and cooperative fashion and exchange Gaussian messages among neighbors resulting in a Gaussian belief propagation algorithm. We discuss and illustrate the performance of our proposed method under acyclic and cyclic network typologies.

中文翻译：

---

基于因子图的分布式共识卡尔曼滤波器

卡尔曼滤波器作为解决线性动态系统状态估计问题的有效工具，可以通过应用传递因子图的和积消息从广义角度推导出来。这个观点鼓励我们在网络上可视化状态估计问题，其中所有节点都希望通过随时间收集顺序测量来获得动态系统的基于共识的状态估计。在这项工作中，节点以分布式和协作方式处理并在邻居之间交换高斯消息，从而产生高斯置信传播算法。我们讨论并说明了我们提出的方法在非循环和循环网络类型下的性能。