Incremental Distributed Robust Inference from Arbitrary Robot Poses via EM and Model SelectionReport as inadecuate


Incremental Distributed Robust Inference from Arbitrary Robot Poses via EM and Model Selection


Incremental Distributed Robust Inference from Arbitrary Robot Poses via EM and Model Selection - Download this document for free, or read online. Document in PDF available to download.

We present a novel approach for multi-robot distributedand incremental inference over variables of interest,such as robot trajectories, considering the initial relative posesbetween the robots and multi-robot data association are bothunknown. Assuming robots share with each other informativeobservations, this inference problem is formulated within anExpectation-Maximization EM optimization, performed by eachrobot separately, alternating between inference over variablesof interest and multi-robot data association. To facilitate thisprocess, a common reference frame between the robots shouldfirst be established. We show the latter is coupled with determiningmulti-robot data association, and therefore concurrentlyinfer both using a separate EM optimization. This optimization isperformed by each robot starting from several promising initialsolutions, converging to locally-optimal hypotheses regardingdata association and reference frame transformation. Choosingthe best hypothesis in an incremental problem setting is in particularchallenging due to high sensitivity to perceptual aliasingand possibly insufficient amount of data. Selecting an incorrecthypothesis introduces outliers and can lead to catastrophicresults. To address these challenges we develop a model-selectionbased approach to choose the most probable hypothesis and usethe Chinese restaurant process to disambiguate the hypothesesprior probabilities over time.



Computational Perception and Robotics - Computational Perception and Robotics Publications -



Author: Indelman, Vadim - Michael, Nathan - Dellaert, Frank - -

Source: https://smartech.gatech.edu/







Related documents