1 / 9

The following slides show the information gain (y axis) vs. heading of the target relative

The following slides show the information gain (y axis) vs. heading of the target relative to the UAV (x axis). Each graph shows the gain from a second measurement after an initial measurement is taken at varying angles listed on each slide. All measurements

mikasi
Download Presentation

The following slides show the information gain (y axis) vs. heading of the target relative

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript

  1. The following slides show the information gain (y axis) vs. heading of the target relative to the UAV (x axis). Each graph shows the gain from a second measurement after an initial measurement is taken at varying angles listed on each slide. All measurements are taken at a constant range. Variance for heading and range are set at .00001 for all slides (roughly a standard deviation of .06 degrees in regard to heading) except slide 5 where variance is raised to .001

  2. Fisher info for initial measurement at 0 degrees

  3. KL divergence info for single Guassian update vice Fisher info

  4. Log of KL divergence again for single Guassian, note similarity to slide 2

  5. Variational approximation to KL divergence using log of KL divergence internally for mixtures of 2 Guassians with means at <0 0 0 0 0 0> and <.01 .01 .01 .01 .01 .01>

  6. Variational approximation to KL divergence using log of KL divergence internally for mixtures of 2 Guassians with means at <0 0 0 0 0 0> and <1 1 1 1 1 1>

  7. Variational approximation to KL divergence using scaled KL divergence internally for mixtures of 2 Guassians with means at <0 0 0 0 0 0> and <.01 .01 .01 .01 .01 .01>

  8. Variational approximation to KL divergence using scaled KL divergence internally for mixtures of 2 Guassians with means at <0 0 0 0 0 0> and <1 1 1 1 1 1>

  9. Note that in all experiments with mixed Gaussians the information values decrease when the means of the two Gaussians are spaced further apart. To confirm that this behavior is acceptable a comparison against the monte carlo sampling (Dmc(f||g)) was performed.Dmc gave the following decreasing values with 0, 3, 6, but then increased at 6, 9, 12: >> kl_mc_sample ans = 0.9164 >> kl_mc_sample ans = 0.1949 >> kl_mc_sample ans = -0.1433 >> kl_mc_sample ans = 0.2514 >> kl_mc_sample ans = 1.3295 Values continued to increase after 12. At a mean difference of 100 the value had risen to 333. This seems a programming bug since the max value obtainable By either GMM is 1.

More Related