Georgia Institute Of Technology - CS 8803-O0103-ParticleFilters

Document Content and Description Below

06/02/2019 03-ParticleFilters file:///D:/WorkBench/Finder/AI4R/03-ParticleFilters/03-ParticleFilters.html 1/37 Particle Filters * easiest to program * most flexible State Space * One of the bigg... est disadvantage for Histogram filters is that it scales expo nentially when it comes to space. * Kalman Filter scales quadratically - only two things are to be stored, vecto r for mean, and co-variance matrix (which is quadratic) * Both are approximate. * Histograms are approximate because world isn't really discrete. * KF are approximate because they are exact only for linear systems. World is actually non-linear! * In some domains, PF scale exponentially. * In other domains e.g. tracking, they tend to scale much better. * Key advantage -- easy to program!06/02/2019 03-ParticleFilters [Show More]

Last updated: 1 year ago

Preview 1 out of 37 pages