Rover Location

[Ted Meyers]

I'd like a better way of determining the true location of my rover during testing.  For now, I'm testing it around known landmarks and recording by hand approximately where it went (to be later compared to GPS, odometery, etc).  It seems like there must be a better way.  I'm thinking of something like audio beacons, or a camera (if I could get it up high enough).  It could be something that is not allowed for a contest like the AVC, but something that would be easy to set up for test runs.  Any ideas?

Also, last year by this time Sparkfun had announced the AVC.  I'm hoping for an announcement any day now!  And a FreeDay announcement?

[Michael Shimniok]

I have been pondering this too. I would like to at least measure run for run path consistency. By dribbling water out the back? :) and some way to compare actual path to intended path. Maybe a separate visual system to locate / triangulate a visual beacon on the robot.

Sent from my iPhone

On Dec 21, 2012, at 6:08 PM, Ted Meyers <ted.m...@gmail.com> wrote:

I'd like a better way of determining the true location of my rover during testing.  For now, I'm testing it around known landmarks and recording by hand approximately where it went (to be later compared to GPS, odometery, etc).  It seems like there must be a better way.  I'm thinking of something like audio beacons, or a camera (if I could get it up high enough).  It could be something that is not allowed for a contest like the AVC, but something that would be easy to set up for test runs.  Any ideas?

Also, last year by this time Sparkfun had announced the AVC.  I'm hoping for an announcement any day now!  And a FreeDay announcement?

--
You received this message because you are subscribed to the Google Groups "diyrovers" group.
To post to this group, send email to diyr...@googlegroups.com.
To unsubscribe from this group, send email to diyrovers+...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msg/diyrovers/-/t8Yio1SLsR4J.
For more options, visit https://groups.google.com/groups/opt_out.
 
 

[wholder]

It's an interesting problem that I've pondered, too.  So far, my leading candidates for things to try are:

1. The IR tracking camera in the Wii remote is one possible solution, as it's cheap (buy a used WiiMote on eBay for about $15 and take it apart) and easy to work with.  One could suspend the camera overhead looking down and put a bright IR source on the car (or two, front and rear, to track the car's direction.)  The camera doesn't have the greatest resolution (it interpolates to 1024x768), but at a scale of 1 pixel/inch, it could track an area 85x21 feet, which is a decent size area for a vehicle to run around in.  However, I think the camera would have to be fairly high up to get this field of view.

2. Alternately, put the camera on the car looking up at an IR source, or perhaps more than one IR source.  Since the camera can track up to 4 IR sources, it might be possible to arrange an overhead grid of lights where no more than 4, and no fewer than two lights are in view at any location, so you could get the car's position for as large of an area as you could cover with the grid.  As a bonus, you'd always get direction info, too, since the camera would always have two sources in view.  This kind of scheme would work best indoors, though, such as inside a large gymnasium, or at night, as I think the brightness of the sky would tend to drown out the IR point sources in the daytime.

It's a shame that there's not a commercial alternative (perhaps with better resolution and the ability to track more lights) to the Wii camera, as it's a very versatile unit.  But, with a good bandpass IR filter and some clever software, it should be possible to emulate what the Wii camera does, as it's just computing the centroid of some points of light.  So, it might also be possible to build something like this without too much expense using a cheap HD camera fitted with the IR filter.  The recording could be processed using a PC, then sync'd up with the car's log data to display a playback of a test run.  If I had more time to devote to side projects, this would be an interesting project to work on but, sadly, such has not been the case, lately.

Wayne

[Scott Harris]

I tacked this problem by driving Tobor on the paint lines of a tennis court and recording data. I then compared the kalman filter output to the known path. Tuning the odometry subsystem alone, for example, I used an optimizer to adjust the rms error between the known path and the estimated path by adjusting the two wheel diameters and the wheel spacing parameters.

Another good test is path closure. Drive a big loop and then make sure your start and stop points are the same.
Scott

[Scott Harris]

[Ted Meyers]

So, I had an idea: place a GPS at a known location on/near the course.  This GPS's location would not change, so it could be used as a reference to compensate for drift.  Better yet, use two fixed GPSes and you can compensate for drift, rotation, and scale.  Seems like this should work; sort of like implementing my own version of Differential GPS.

[wholder]

Unfortunately, I do not think this will work without using much more sophisticated receivers that work at the signal timing level.  For some background on this, see this article:

  http://www.oreillynet.com/pub/a/wireless/2000/12/29/two_gps.html

and this thread:

  http://www.gpspassion.com/forumsen/topic.asp?TOPIC_ID=109246

Wayne

[Ted Meyers]

Thanks Wayne,

Good to know before getting too far down that road...  Obviously this has been tried before, without any success.  It sounded like such a good idea.  Oh well.

Ted

[Ted Meyers]

Next idea: have the bot send out ultrasonic pings, listening stations pick up the pings and geolocate the bot.  This does require the listening stations and the bot to all be in radio communications.  Of course, this is just complicated enough that for me its not worth it.

[jesse brockmann]

You know, I've been thinking about another idea for location.    Have IR leds on bot, and have IR video cameras setup at each corner.  Then you could track the bot without any gps, etc.  :)     Or rather then having a camera at each corner... have one in a quadcopter flying overhead. ;)    More reasonable maybe to have a camera on board and try to do some type of pattern recognition.    So, who want's to try it?

On Sat, Jan 26, 2013 at 9:52 AM, Ted Meyers <ted.m...@gmail.com> wrote:

Next idea: have the bot send out ultrasonic pings, listening stations pick up the pings and geolocate the bot.  This does require the listening stations and the bot to all be in radio communications.  Of course, this is just complicated enough that for me its not worth it.

--

You received this message because you are subscribed to the Google Groups "diyrovers" group.
To post to this group, send email to diyr...@googlegroups.com.
To unsubscribe from this group, send email to diyrovers+...@googlegroups.com.

To view this discussion on the web visit https://groups.google.com/d/msg/diyrovers/-/me75aCdg6u4J.