build details

Show: section status errors & todos local changes recent changes last change in-page changes feedback controls

Drone Localization

Modified 2020-09-16 by Dev Ramesh

The Drone flys over a known environment. To localize, state estimation is required on a system that changes over time.(The drone changes the view when flying over time.)

Guassian

Particle Filter

localization

Monte Carlo Localization

Modified 2020-09-12 by josetoribio

Monte Carlo Localization is localizing using particle filters.

Monte carlo Localization has 3 phases

prediction phase

Update phase

resamplying

Prediction Phase

Modified 2020-09-12 by josetoribio

During the Predicition phase the drone doesn’t know where it is in the map and can equally be anywhere in the map.

Update Phase

Modified 2020-09-12 by josetoribio

During the Update phase the drone updates the particles to be aware of the where on the map it is at.

ReSamplying Phase

Modified 2020-09-12 by josetoribio

DUring the Resamplying Phase the weighted particles with higher numbers are recycled while deleting the lower weighted particles.

Particle FIlters

Modified 2022-03-11 by josetoribio

Particle filters represents a Probability Distrubution Function. A Probability distrubution function grabs the probability of a given event.With 50 features, the PDF will give a weighted probability of those 50 features.Given the random features on the camera, the Filter can give the probability of the location by weighting out each particle feature it has. The more weight the particle has the more likely that particle has to do with the location of the drone then a particle with a lower weight.The higher weight gets recyled in the filter while the lower weights are deleted. Deleting the lower weighted particles from algorithm, helps to effeciently localize because only the higher weighted particles are keeped in the algorithm while searching for more particles.

The map used for the DuckieSky Drone is distict around the map because the more features the better! The accuracy of particles is relatived to the ammount of features used to represent the Probability Distrubution Function.

How will the map work when the marks on the map are the same?

How do weight particles effect localization?