RE: Location Detection

There are a lot of ways that you can fine tune the thresh holds that each of 
your positions sensors would have.  
A robot that I am currently working on uses an accelerometer/digital compass, 
gps and digital encoders for its positioning.  However not all these are 
extremely accurate, but using a filter (kalman filter for example) you can get 
much more precise coordinates. the robot uses stereo vision and sonar for 
object detection (along with infrared for close up) and is also put through 
kalman filters to increase the accuracy of the sensors.
another method that you can use is the same technology in a mouse. kinda built 
a mouse sensor into the bottom of your robot chassis, when you move, the mouse 
picks up the your x/y coordinates.  this would seem to be pretty acurate if you 
were moving on a smooth surface, but wouldnt work at all outdoors.
hope this helps
Dave
 

 
________________________________

From: gumstix-users-admin@xxxxxxxxxxxxxxxxxxxxx on behalf of Athanasios 
Anastasiou
Sent: Sat 7/9/2005 7:12 AM
To: jeremy@xxxxxxxxxxxxxx
Cc: gumstix-users@xxxxxxxxxxxxxxxxxxxxx
Subject: [Gumstix-users] Location Detection



Hello Jeremy

I also had the idea of using several ultrasound fixed beacons and
deriving the position from these BUT there is a much easier solution to
this which is used as an alternative navigational aid.

It is called Inertial Navigation and it breaks down to basic principles
of Kinematics.

You just mount an accelerometer on your robot and calibrate it before it
sets off moving around. The calibration is just setting the starting
point. Lets suppose that this is A(0,0). When the robot starts to move,
the accelerometer records the dx, dy and dz of the robot chassis and
updates the starting point. So after integrating the accelerations you
get a new point B(x,y).

Ofcourse there are a few caveats in this method, one of which is that if
your robot moves towards a direction, under the threshold of detection
of the accelerometer, its movement will not be recorded and its position
not updated. But this is not so serious in robots that have 4 wheels and
move in interior spaces (probably more of a problem for marine robots
using this method.).

The second problem is that you have to make sure that the accelerometer
is always vertical in respect to the chassis of the robot (otherwise you
need to correct the accelerations for the angle of movement). If you are
absolutely sure that the robot is not going to get tilted you can use a
single accelerometer...If not, you need an extra accelerometer just for
measuring the angles and thus adding three more degrees of freedom
(movement in three directions plus roll pitch and yaw angle).

STMicroelectronics sells some very interesting accelerometers that can
give you the three basic accelerations (dx,dy,dz) over serial or I2C
buses. In an application note i noticed that a single accelerometer can
also be used to provide angular position as well. I am not so sure about
that...

Could there be any other alternatives to correcting the values that an
accelerometer returns except mounting it on super slick gimbals and
gyroscopic mechanisms??????

I hope this helps

All the best
thanOS


-------------------------------------------------------
This SF.Net email is sponsored by the 'Do More With Dual!' webinar happening
July 14 at 8am PDT/11am EDT. We invite you to explore the latest in dual
core and dual graphics technology at this free one hour event hosted by HP,
AMD, and NVIDIA.  To register visit http://www.hp.com/go/dualwebinar
_______________________________________________
gumstix-users mailing list
gumstix-users@xxxxxxxxxxxxxxxxxxxxx
https://lists.sourceforge.net/lists/listinfo/gumstix-users

<<winmail.dat>>