Subject:
|
Re: localization (was re: GPS)
|
Newsgroups:
|
lugnet.robotics
|
Date:
|
Wed, 11 Jan 2006 21:24:09 GMT
|
Viewed:
|
2046 times
|
| |
|
|
In lugnet.robotics, Chris 'Xenon' Hanson <xenon@3dnature.com> wrote:
> I've always been partial to ultrasonic rangefinding. It's not too difficult indoors
> under controlled circumstances. The old Nintendo PowerGlove used this technique. For robot
> localization, you'd probably want a minimum of three non-moving marker beacons, possibly
> four. These beacons give a (mostly) omnidirectional ultrasonic chirp when commanded by the
> bot, and the bot times how long before it receives the chirp. With fast enough processing,
> some pretty decent accuracy can be achieved.
>
> You need to have each beacon located far enough from the others and in a different
> enough direction that triangulation is possible.
>
> It's sort of like GPS, only indoors and sonic instead of radio.
>
> You'd probably need to use a radio broadcast from the bot to the marker to tell it when
> to chirp.
>
> It's something that would have to be assembled into a complete tracking system, it
> would be nearly impossible for most hobbyists to assemble the hardware and software.
>
> I know John Barnes has done a lot of ultrasonic work, perhaps he could comment on the
> practicality of making a system like this available. I imagine due to the
> density/complexity of the positioning data, it would need to talk bluetooth or some other
> fast high-level protocol to the bot.
>
> This would be a useful system for much more than just Mindstorms, obviously.
>
> The one downside is that, like GPS, it doesn't tell you heading, just position and
> direction of travel.
I have tried all sorts of tricks with ultrasonic stuff, you're right :)
There are a couple of different schemes for the ultrasonic beacon type things.
You can have ones which are triggered from the 'bot via radio or IR. I've tried
the IR one. A really bright IR pulse can get around the same kinds of corners
that ulrasonic can, although once the sounds has started going round corners,
all bets are off anyhow. The other thing you have to do is be able to address
each beacon so you can request which one transmits. Then you pick the earliest
arriving pulse, 'cos the later ones will be due to multipath. And if they all
transmit at once, the mixture of coincidentally equal distances to two or more
beacons and the chance of multipath effects from closer beacons interferring
with signals from others further away makes individual arrival times of each
beacon's pulse hard to do. It is more difficult for a beacon to encode it's id
in its transmitted pulse (due to limited modulation bandwidth) than it is for
the requesting radio or IR pulse to have id information in it.
I've also played around with multi-sensor element phased arrays too to get
signal arrival angle, but you need a lot of sensors to get a decent result and
that gets expensive quickly.
With the new digital interfaces (whatever they finally turn out to be), I am
sure there will be lots of opportunity to play with positioning schemes, since
the narrow analog pipeline into the RCX is a restriction which I assume will
disappear.
I had been working on a project to build a cheap warehouse navigation system for
the company I used to work for until it closed its doors on Dec 23rd.
The idea I had had, and intend on persuing is a cheap(ish) optical scheme
whereby you hang little barcode placards around the environment. Then a cheap
line scan camera built around a TAOS linear array and a cheap 12.5mm lens can be
used to build a barcode reader which not only grabs the placard id but also gets
placard range based on perceived size of the barcode. (Unless someone out there
knows of a cheap barcode scanner which returns size as well as code value.) In
the warehouse application, trucks were going to drive by such things and get a
local position update each time the saw one. The NXT application may have to put
such a device on a scanning platform.
If I can make it work, cheaply enough, you'll hear about it here later in the
year.
JB
|
|
Message has 3 Replies:
 | | Re: localization (was re: GPS)
|
| Hi John, A localization scheme I have used with the RCX is uses the angles between light beacons in known locations, measured with a rotating light sensor. Two angles (3 beacons) are enough to calculate, with a bit of trig, the robot coordinates and (...) (19 years ago, 12-Jan-06, to lugnet.robotics)
| | | Re: localization (was re: GPS)
|
| (...) Have you checked out the ARTag software? (Agumented Reality Tagging) I downloaded the Rev1 demo (2 is out now) and it was really amazing, even using a super-cheap web-cam: (URL) sdk provides the means to identify their graphical tags, as well (...) (19 years ago, 12-Jan-06, to lugnet.robotics)
| | | Re: localization (was re: GPS)
|
| (...) ;) (...) Right. I'm assuming something that basically assumes Line of Sight between bot and 3 or more beacons, just like GPS. GPS doesn't deal well with multipath either. ;) > The other thing you have to do is be able to address (...) Right. (...) (19 years ago, 12-Jan-06, to lugnet.robotics)
|
Message is in Reply To:
| | Re: localization (was re: GPS)
|
| I've always been partial to ultrasonic rangefinding. It's not too difficult indoors under controlled circumstances. The old Nintendo PowerGlove used this technique. For robot localization, you'd probably want a minimum of three non-moving marker (...) (19 years ago, 11-Jan-06, to lugnet.robotics)
|
12 Messages in This Thread:
 
- Entire Thread on One Page:
- Nested:
All | Brief | Compact | Dots
Linear:
All | Brief | Compact
This Message and its Replies on One Page:
- Nested:
All | Brief | Compact | Dots
Linear:
All | Brief | Compact
|
|
|
|
