hi my name is william attacker gonzalez and i'm doing a project on a time
to miss navigation and control
evaluating methods to simultaneous localisation and mapping or slam
as you may have heard and then use recently
there was a recent
playing flight
motion plane flight crash
number three seventy
where hundreds of people are gone missing
and right now sciences are using a time to miss underwater vehicles in a huge
search operation
using the technology known as slam to map
and visualize the motion for to look for displaying plate
so what is slam
after getting little background so i am it is simultaneous localisation mapping as i said
before
and it uses complex geometric form as in sure gonna metric functions for robot to
scan visualise environment around it as well as position itself so it can know where
it is relative to all the objects surrounding it
so some approaches to slam better being used today
our sensor integration where robot will have not only just a camera to scan
by the infrared sensor all just sighing sensor et cetera et cetera
and uses hierarchical methods
to help the robot's can
environment using probability theory
so let's say the robot is gaining a glass door obviously the camera is not
going to return
the best
quality data so the way the ultrasonic sensor over the camera
so getting in a my product might pieces was as a tandem since robot such
as drivers cars are use a novel environments
scanning images of an environment using fast scanning algorithms
we improve upon the efficiency of mapping as compared with the current standard methods
and an overview of my project
i did a simulation of slam algorithms and scanning speed
using different sensors
that were working together
and i would evaluate these maps using mahalanobis distance equation
which is as shown below
so the initial goals and objectives of my product work to take a mobile robot
such as nine robot remember
and they connect the x box sensor
and i would but there was together
couple them together you know these are two things you can get
really well priced at best buy
local best buy
to navigate environment and map it
and i whatever analyze these physical environment s
so my experimental process involved using online open source code from
something on is "'cause" e bow
i would then start of virtual world
and one minute i missed mode in that virtual world
and start mapping and visualising
now after that was done i would analyse my results
so here's my project set up this is all on my laptop i did a
few test with the physical environment but unfortunately i wasn't able to show those here
but issue here on the left side you can see an odometer a compiler
where the robot
is very ago ready to start scanning
it's ready to localise itself
from its initial position by keeping track of
how far as wheels have spun
and on the right is the complete scans of a simple q the bottom when
action a cylinder and top right
as you can see is the speed increases
for a relatively
the
quality of the maps goes down
as it approaches the right picture
and my results here's a graph
on the bottom right bar graph
showing them how nervous
value
verses the speed of the scanning
mahalanobis value when it's a value of zero means is a perfect scan so my
very low speeds can almost return a value of zeros about twenty one i was
terry surprise how well that when i'm
as you can see the mahalanobis five gets higher and higher
and efficiency goes down with the bowser scanning speed
which unfortunately i did
did not think would happen
so significance of my project it's a very cheap alternative to vary
complex technology being used today the google car actually
which
self navigate to self use the technology notice light are
which is hunters the thousand dollars
so this was you know open source project that i did
it's something simple
that high school students like me can do for
a very small amounts of money
very cheap
and you know i was very surprise of the quality of maps
that my programs generated
from doing this project
and that's it thank you