3-Wheel Robot navigation

Posted by cnichols@greenvalleyschool.com on April 27, 2006, 12:07 pm

Here's a picture of a 3-wheel robot I built:
http://www.greenvalleyschool.com/Curtis/3Wheel.jpg
A motor drives the front wheel and steers the front wheel via a servo.
There is a shaft encoder on the front wheel. I'm thinking that I
should be able to keep track of wheel angle and distance from a
starting (x,y) coordinate (and starting robot angle) and be able to
know the present (x,y) position and heading.

Any ideas about practicality of this approach? Potential problems? Any
good code? (I am pretty good with math but think there may be some
better approaches than the brute force trigonometry I have in mind.

Curtis

Posted by Brendan Gillatt on April 27, 2006, 12:20 pm

cnichols@greenvalleyschool.com wrote:

> Potential problems?

One problem would be traction - if the wheel slips the robot thinks it
has moved forward whereas in reality it has remained stationary. The
Mars Rover Team found this out as there robot kept slipping on the
looses Martian soil. It should be OK with the rubber wheels you're using
and a smooth, hard surface.

--
/-----------( Brendan Gillatt )----------\

<---( http://www.brendangillatt.co.uk )---->

\---(brendan[at]brendangillatt.co.uk )---/

Posted by Gordon McComb on April 27, 2006, 12:49 pm

cnichols@greenvalleyschool.com wrote:

>
> Here's a picture of a 3-wheel robot I built:
> http://www.greenvalleyschool.com/Curtis/3Wheel.jpg
> A motor drives the front wheel and steers the front wheel via a servo.
> There is a shaft encoder on the front wheel. I'm thinking that I
> should be able to keep track of wheel angle and distance from a
> starting (x,y) coordinate (and starting robot angle) and be able to
> know the present (x,y) position and heading.
>
> Any ideas about practicality of this approach? Potential problems? Any
> good code? (I am pretty good with math but think there may be some
> better approaches than the brute force trigonometry I have in mind.

Take a look at this robot to see how a similar three-wheel robot uses a
wheel encoder:

http://www.robotics.com/arobot/

Yours is a bit different in that the steering wheel is also the driven
wheel, but the basic concepts are similar.

The problem with any solution where there is just single encoder on the
front wheel of the tricycle arrangement is you can calculate distance,
but not bearing. You'll need to infer bearing by the position of the
steering motor, and this is inherently inaccurate. Well, wheel odometry
is inherently inaccurate to begin with, but there are ways to minimize
it. You might consider putting encoders on the rear wheels instead, and
then you'd use ordinary odometry techniques you can commonly find with a
Y or G search. For starters, there's a Sourgeforge project that
discusses this issue:

http://rossum.sourceforge.net/papers/

and you can find a variety of different implementations, some with
posted code, on various robotics retail and hobby pages. Here's one, for
example, based on a commercially available quadrature encoder board
designed for small bots:

http://www.nubotics.com/support/ww01/example_code.html

Sorry...don't see any examples for Vex.

-- Gordon

PS: I'm sure you are aware Radio Shack is discontinuing Vex, and they
have everything half off. The sale used to be "double secret" but the
sale prices are now listed on the radioshack.com Web page.

Posted by cnichols@greenvalleyschool.com on April 28, 2006, 6:35 am

Well, that certainly answers a lot of questions that I didn't even know
that I had!!! Thanks Gordon!

Curtis

Posted by Si Ballenger on April 27, 2006, 2:53 pm

If you can see the full headers of the post (my isp doesn't
carry the binarys groups) you should see a line with
NNTP-Posting-Host: and an IP address. This address will usually
tell from which system/account the post came from. A complaint to
the ISP about posting copyrighted material might get the account
terminated with the message that people do care.

Similar Threads	Posted
Robot navigation reinforcement learning paper	November 30, 2005, 6:03 pm
Programming a Local GPS Navigation System into a Mobile Robot	May 25, 2008, 4:14 am
Navigation	June 28, 2005, 11:36 am
GPS digital navigation map data	December 24, 2005, 12:04 am
Associative memory for navigation	September 6, 2007, 3:01 pm
Post-Doc position available at Inria Rhone-Alpes, Grenoble (FR) - Autonomous navigation in dynamic environments	September 15, 2005, 2:08 pm
Atlanta Hobby Robot Club announces the 10th annual Robot Rally	April 1, 2006, 5:11 pm
Off-road mobile robot at http://robot.proof.cz	February 1, 2007, 8:04 am
considering "Building Robot Drive Trains" or "Building Robot Bases" ...	December 10, 2005, 1:42 am
Re: MLW sub $500 robot	June 12, 2005, 9:46 am