DPRG
DPRG List  



DPRG: Diffrent terrains for navigation

Subject: DPRG: Diffrent terrains for navigation
From: Bryan Andersen bryan at visi.com
Date: Fri Aug 20 15:25:44 CDT 1999

First thing I'll mention is the best thing is to have three plus 
sources of reference.  That way you can compare them against 
each other, and discard/average their outputs.  Different types 
of references do better under certain circumstances than others.  

An accelerometer (linear or rotational) is great when things are 
changing, like the robot is accelerating, decelerating, or turning.  
However they fall apart when stationary or over long periods of 
time.  They all have slight drift problems that accumulate over 
time.  Hence they need some other reference to correct for these 
inaccuracies.  

Wheel encoders also have their problems.  Wheels slip while under 
motion, and this changes depending on the surface one is traveling 
over.  This gets really bad when the robot is turning, or has the 
wheels on different types of surfaces.  Its gets especially bad 
when the robot is turning and the wheels are fat.  You don't know 
which part of the wheel gripped, was it the inside, outside or 
somewhere in-between.  For that matter, if the wheel is linked to 
another wheel via a fixed axle, which wheel gripped, and where on 
that wheel, and in what portions of the turn?  

Magnetic compasses also have their problems.  They are prone to 
interference from surrounding magnetic fields.  These might be 
based on the robot, or part of the world around the robot.  They 
will all effect the angle given by small amount, but unfortunately 
unless you've mapped out the area magnetically, and calibrated 
your robot magnetically you won't know by how much.  The nice 
thing is the errors are bounded and quite repeatable.  This means 
that if calibration is done, reasonable accuracy is available in 
the calibrated area.  Spin calibrating your robot can also help 
improve accuracy by allowing you to map out variances particular 
to your robot, but it needs to be done in an area free of other 
magnetic disturbances.  They do this on boats to get their 
compasses as accurate as possible.

Radar type sensors (IR Distance, Sonar, LiDAR, etc.) all are good 
for telling the distance and orientations to objects, but did the 
object move, or was it it's own motion that caused the distance 
change?  They are also prone to errors caused by the type of 
material being sensed.  They also require a much greater level of 
computation to use.

Video inputs.  Buh, what am I looking at?  Require a huge amount 
of computation to get useful navigation information out of, but 
they also have the greatest ability to provide authoritative fixes 
too, provided the information can be interpreted correctly, but 
they also require a reference map to provide a fix.  

Fortunatly each form has different weaknesses under different 
circumstances.  By averaging their inputs, and possibly discarding 
inputs that go wildly astray we can get much better positioning 
information than if we only used one refference.



Matt Minnis wrote:
> I use a six wheeled design.
> They are arranged in more of a hexagon than a rectangle.

Which wheels have encoder(s)?  All, one pair, one axel?  I don't 
know enough about your design to be able to figure out where it's 
gaining all the slippage.  

>From the images it seams like the encoder would be more suited to 
telling the distance traveled.  On a car style chasis with car 
style drive system, the stearing wheels will also slip a little 
sideways to the outside of the turn.  This increases with speed 
and acceleration.  This could also be a part of your accuracy 
losses.

- -- 
|  Bryan Andersen   |   bryan at visi.com   |   http://softail.visi.com   |
| Buzzwords are like annoying little flies that deserve to be swatted. |
|   -Bryan Andersen                                                    |

------------------------------

More information about the DPRG mailing list