Subject:
|
Re: POC Exploration Robot
|
Newsgroups:
|
lugnet.robotics
|
Date:
|
Thu, 16 Sep 2004 07:00:08 GMT
|
Original-From:
|
Mr S <{szinn_the1@yahoo}nomorespam{.com}>
|
Viewed:
|
2040 times
|
| |
|
|
Mathieu,
My replies are mixed in your message, hope they are
helpful.
--- Mathieu Lalonde <e0nblue@ssl-mail.com> wrote:
[[snip]]
> What my teammate and I would like to do is a proof-
> of-concept exploration robot (a la Mars Rover)
> built entirely out of Legos. I am posting to this
> list to ask the community about the feasibility of
> such a project, before we invest a significant
> amount of money into Mindstorms parts:
This is reasonable, since you are talking about two
mindstorms kits, plus vision command, though you could
use any logitech camera if you are going to use a long
and light USB cable?
> The robot would consist of two RCX microcontrollers
> coupled with a variety of sensors relevant to our
> project (ie: temperature, atmospheric pressure,
> humidity, light, etc) and a Vision Command Unit (or
> whatever Lego calls it’s QuickCam rip-off), all
> mounted on a 4 wheel drive (or maybe a synchro
> drive).
The sensor inputs of the RCX are analog, and powered,
you might want to look at this page:
http://www.plazaearth.com/usr/gasperi/lego.htm
I'm right there with you on the size thing. I have a
robot that is quite big, all four wheels are driven
using a combination of 6 9v gear motors, chain, gears,
and differential gears. 2 motors on each front wheel,
and two shared out on the rear wheels with the
differential gear. This takes three RCX outputs. The
three motors in use to drive the bot means that using
the RCX remote control is awkward. You might consider
two wheel drive for this reason. Also, much of the
information on the web concerning navigation is done
with differential steering, and that sort of counts
out any cut-n-paste help with the navigation if you
have four wheels.
I can't seem to get brickshelf to show my pics?
> It would support two modes of operation: auto-pilot
> and manual. A custom unix app would use the IR tower
> to send and receive data from the robot.
Sound idea, but the IR communications may be more
limiting than you have thought of. Its not simply a
serial interface with IR as the medium. It should
still work for you, given that the IR comms works at
all in the lighting conditions. Are you going to put
out a cable for the IR tower and make it part of the
robot?
> In manuel
> mode, the unix server would receive commands (sensor
> activation, directions, etc) from a client and would
> transmit then to the robot.
In a room not exposed to sunlight, the IR comms would
work reasonably, but in sunlight, expect many problems
with IR. You may find the RCX more capable than what
your description seems to indicate.
> The first RCX’s IR module would be used for
> communication, and the second IR module would be
> used for proximity detection (in combination with a
> light sensor).
I'm planning this for my sumo robot, but have not done
it yet. It is supposed to work well, but would still
suffer from sunlight exposure. Search this news list
for max picking LIDAR.
Also, if you need the two RCX's to talk with each
other, the RCX IR ports will have to face each other
or use some mirror configuration which might cause you
issues with talking to the server?
> In auto-pilot mode, the robot would stumble
> randomly around the room until it would detect an
> object, where it would stop and signal the discovery
> to the server, which would turn on the robot’s cam,
> take a still picture (the cam is connected to the
> server via USB) , parse it for pattern/color
> recognition and add it to a database.
This is definitely doable, though you might want to
look into using a sharp IR sensor with the
modifications that can be found on the Internet for
RCX use.
Note: stumbling around randomly is harder than trying
to stumble around in a pattern.
> To prevent
> multiple entries for the same object, the robot
> would send its position on the ground upon the
> discovery of a new object (I still haven’t found a
> proper way for the robot to calculate it’s
> position), so that similar objects with similar
> coordinates would be rejected.
This is all on the server side (more or less, so you
might look at the Open Computer Vision group)
http://groups.yahoo.com/group/OpenCV/
There is a good article on the SRS website about
navigation using an RCX:
http://www.seattlerobotics.org/encoder/200108/using_a_pid.html
I'm still working on hacking some mouse encoders, but
haven't got anything to write the list about yet.
> Given the limitations of the language and the
> software that come with the Mindstorms, we would
> like to use BrickOS/C on the robot and *nix/C++
> on the server side.
BricX command center / NCQ is not such a bad thing to
use.
> So, here come my questions: Is the whole thing
> feasible? What would you do differently and why?
> What major problems are we bound to run into? What
> are the drawbacks of the choices we made (BrickOS
> instead of NQC or Lego’s devel software, using two
> RCXs, using a proximity detector, using a 4 wheel
> drive vs a synchro drive, etc)?
I think that you need to do the pro/con list version
of math concerning the sensors that you want to
implement, what you want to achieve with them, and
measure that againsts the limits of the 2 RCX
situation. That gives you 6 outputs, and 6 inputs.
Some of the sensors can be multiplexed, like putting a
light sensor and a touch sensor in parallel. Also, you
might consider making sensors of your own, see the URL
about sensor inputs, many diodes are light sensitive,
and many parts are temperature sensative. There is
information about using the sharp IR sensors with RCX.
www.mindsensors.com (?) has an IR interface to 2
standard RC servos.. might be useful to you in
positioning the camera or sensors?
Don't forget some simple bump sensors just incase
those proximity detectors fail.
There is an example of the Mars rover built with lego
on the Internet somewhere, including the synchrodrive,
but you might find that a bit clumsy for your
purposes. You might try ackerman style steering on
both front and back, like the Denali SUV? My robot
Pivots in the middle like some construction equipment,
but the front/rear ackerman steering might be just as
good as the synchrodrive in most situations?
>
> Any feedback would be _greatly_ appreciated!
>
> PS: If some people show interest in our project, I
> will gladly keep you guys updated with pictures and
> info as we progress.
Please do keep us all informed :)
Attached is my bookmarks for LEGO 'stuff' on the web.
Cheers
SZ
__________________________________
Do you Yahoo!?
Yahoo! Mail – Now with 25x more storage than before!
http://promotions.yahoo.com/new_mail
|
|
1 Message in This Thread:
- Entire Thread on One Page:
- Nested:
All | Brief | Compact | Dots
Linear:
All | Brief | Compact
|
|
|
|
Active threads in Robotics
|
|
|
|
|
|
