2019/07/10 by Stefano V. Albrecht, Subramanian Ramamoorthy, Albrecht, Stefano V. +1
Computer Science · Decision Sciences · #Bayesian Modeling and Causal Inference #Data Quality and Management
paper · pdf · doi:10.48550/arxiv.1907.05850
Dynamic Bayesian networks (DBNs) are a general model for stochastic processes\nwith partially observed states. Belief filtering in DBNs is the task of\ninferring the belief state (i.e. the probability distribution over process\nstates) based on incomplete and uncertain observations. In this article, we\nexplore the idea of accelerating the filtering task by automatically exploiting\ncausality in the process. We consider a specific type of causal relation,\ncalled passivity, which pertains to how state variables cause changes in other\nvariables. We present the Passivity-based Selective Belief Filtering (PSBF)\nmethod, which maintains a factored belief representation and exploits passivity\nto perform selective updates over the belief factors. PSBF is evaluated in both\nsynthetic processes and a simulated multi-robot warehouse, where it\noutperformed alternative filtering methods by exploiting passivity.\n