Incremental Non-Gaussian Inference for SLAM Using Normalizing Flows

作     者：Huang, Qiangqiang Pu, Can Khosoussi, Kasra Rosen, David M. Fourie, Dehann How, Jonathan P. Leonard, John J. 

作者机构：MIT Comp Sci & Artificial Intelligence Lab Cambridge MA 02139 USA MIT Dept Nucl Sci & Engn Cambridge MA 02139 USA CSIRO Robot & Autonomous Syst Grp DATA61 Brisbane Qld 4069 Australia Northeastern Univ Dept Elect & Comp Engn Boston MA 02115 USA Northeastern Univ Dept Math Boston MA 02115 USA NavAbility Boston MA 02110 USA MIT Dept Aeronaut & Astronaut Engn Cambridge MA 02139 USA 

出 版 物：《IEEE TRANSACTIONS ON ROBOTICS》 (IEEE Trans. Rob.)

年 卷 期：2023年第39卷第2期

页      面：1458-1475页

核心收录：

学科分类：0808[工学-电气工程] 08[工学] 0811[工学-控制科学与工程] 

基　　金：ONR ONR MURI 

主　　题：Simultaneous localization and mapping Inference algorithms Approximation algorithms Task analysis Particle separators Belief propagation Random variables Bayes tree distribution estimation non-Gaussian normalizing flows SLAM 

摘      要：This paper presents normalizing flows for incremental smoothing and mapping (NF-iSAM), a novel algorithm for inferring the full posterior distribution in SLAM problems with nonlinear measurement models and non-Gaussian factors. NF-iSAM exploits the expressive power of neural networks, and trains normalizing flows to model and sample the full posterior. By leveraging the Bayes tree, NF-iSAM enables efficient incremental updates similar to iSAM2, albeit in the more challenging non-Gaussian setting. We demonstrate the advantages of NF-iSAM over state-of-the-art point and distribution estimation algorithms using range-only SLAM problems with data association ambiguity. NF-iSAM presents superior accuracy in describing the posterior beliefs of continuous variables (e.g., position) and discrete variables (e.g., data association).