Navigation discussion
Archiver
Posts: 46,084
Hello All: I am working on an outdoor robot, and have been tossing around
numerous ideas for navigation. If possible I would like to start a small
discussion on the topic.
I myself am going to stick with some sort of basic stamp as my processor,
possibly 2 or 3 onboard if necessary. My biggest reason being the simple
programming language.
My concerns are navigating around small, and odd shaped obstacles. I believe
we all can detect a solid wall, or thick tree base pretty reliably. What
concerns me are objects such as fir trees, where their branches come down to
the ground, but are sparse, or have open gaps in the needles/branches. I can
envision my robot trying to drive right under a fir tree. I really do not
think normal IR is going to detect this situation. A 2nd problem I see are
very narrow sticks, or even thin plants in the yard. Indoors I can see many
times when my IR can not detect a thin chair leg. Bumper sensors do solve
this problem normally.
How to prevent the robot from driving inside a very cluttered area? For
example a garden may have an opening at the ends where the rows start, but
inside the garden there are many plants/clutter. Ideally the robot should
never enter such a clutterd location.
Speed is not an issue, a slow speed, around 1-2 mph is fine. A 2nd example
is the edge of woods, where there are hundreds of small saplings/twigs etc
causing random clutter. A simple IR beam would pass between the saplings,
and so the robot would try to also.
1) Compass for navigation, I am afraid it would be to coarse a measurment. I
dont think it would be possible to program all obstacles, and expect a
compass to keep the robot aligned close enough to its track to avoid them.
2) GPS,(thanks to all for replys) It appears that GPS will be from 9-20 foot
tolerance, unless using DGPS radio signal as well. This basically eliminates
GPS for in close navigation.
3) Sharp GP2D02 sensors, as well as the new Devantech SRF04 UltraSonic
Ranger. These items seem like the best method, however now it looks like,
either the speed of the cpu, and sensing will kill any benifits. Possibly 1
sensor on a servo, and pan it around? This is good for horizontal, then do
it again in a vertical direction, rescan horizontally etc? The program code
of updating the servo, reading sensor, deciding whats really out there etc
etc seems like a nightmare to my old brain.
4)Video input seems out simply becauase of the cost, and program skill
necessary.
5)Programmed in way pts (obstacles) this then requires again precise navigation.
-- I do not care if the robot KNOWS where it is, at any time, just so that
it is semi reliable in avoiding objects. I do not think a human Precisely
KNOWS where they are at, at any one time. We simply walk in between trees
(Unless on horseback), and proceed to the house. My apologys for the long
letter, however I have nobody else in my area, to bounce ideas off of.
Thank you to everyone.
Sincerely
Kerry
Admin@M...
WWW server hosting [url=Http://mntnweb.com]Http://mntnweb.com[/url]
Kerry Barlow
p.o. box 21
kirkwood ny
13795
numerous ideas for navigation. If possible I would like to start a small
discussion on the topic.
I myself am going to stick with some sort of basic stamp as my processor,
possibly 2 or 3 onboard if necessary. My biggest reason being the simple
programming language.
My concerns are navigating around small, and odd shaped obstacles. I believe
we all can detect a solid wall, or thick tree base pretty reliably. What
concerns me are objects such as fir trees, where their branches come down to
the ground, but are sparse, or have open gaps in the needles/branches. I can
envision my robot trying to drive right under a fir tree. I really do not
think normal IR is going to detect this situation. A 2nd problem I see are
very narrow sticks, or even thin plants in the yard. Indoors I can see many
times when my IR can not detect a thin chair leg. Bumper sensors do solve
this problem normally.
How to prevent the robot from driving inside a very cluttered area? For
example a garden may have an opening at the ends where the rows start, but
inside the garden there are many plants/clutter. Ideally the robot should
never enter such a clutterd location.
Speed is not an issue, a slow speed, around 1-2 mph is fine. A 2nd example
is the edge of woods, where there are hundreds of small saplings/twigs etc
causing random clutter. A simple IR beam would pass between the saplings,
and so the robot would try to also.
1) Compass for navigation, I am afraid it would be to coarse a measurment. I
dont think it would be possible to program all obstacles, and expect a
compass to keep the robot aligned close enough to its track to avoid them.
2) GPS,(thanks to all for replys) It appears that GPS will be from 9-20 foot
tolerance, unless using DGPS radio signal as well. This basically eliminates
GPS for in close navigation.
3) Sharp GP2D02 sensors, as well as the new Devantech SRF04 UltraSonic
Ranger. These items seem like the best method, however now it looks like,
either the speed of the cpu, and sensing will kill any benifits. Possibly 1
sensor on a servo, and pan it around? This is good for horizontal, then do
it again in a vertical direction, rescan horizontally etc? The program code
of updating the servo, reading sensor, deciding whats really out there etc
etc seems like a nightmare to my old brain.
4)Video input seems out simply becauase of the cost, and program skill
necessary.
5)Programmed in way pts (obstacles) this then requires again precise navigation.
-- I do not care if the robot KNOWS where it is, at any time, just so that
it is semi reliable in avoiding objects. I do not think a human Precisely
KNOWS where they are at, at any one time. We simply walk in between trees
(Unless on horseback), and proceed to the house. My apologys for the long
letter, however I have nobody else in my area, to bounce ideas off of.
Thank you to everyone.
Sincerely
Kerry
Admin@M...
WWW server hosting [url=Http://mntnweb.com]Http://mntnweb.com[/url]
Kerry Barlow
p.o. box 21
kirkwood ny
13795
Comments
Not being into robotics the following $.02-worth may be pointless, but:
Why not try looking for 'venues' or 'wide enough open spaces' with 'depth'
equal to the sensor's range. It seems to me that this logic should be
simpler and more fail-safe than looking for obstacles by means of sensors
with adequate resolution and deciding if there is a passageway between them.
BTW, small relatively inexpensive compasses with 1-5 degree resolution and
repeatability exit and their interfaces with stamps have been figured out
long ago.
Regards,
Dov
Original Message
From: "Kerry Barlow" <admin@m...>
To: <basicstamps@yahoogroups.com>
Sent: Sunday, March 18, 2001 9:45 PM
Subject: [noparse][[/noparse]basicstamps] Navigation discussion
> Hello All: I am working on an outdoor robot, and have been tossing around
> numerous ideas for navigation. If possible I would like to start a small
> discussion on the topic.
> I myself am going to stick with some sort of basic stamp as my processor,
> possibly 2 or 3 onboard if necessary. My biggest reason being the simple
> programming language.
> My concerns are navigating around small, and odd shaped obstacles. I
believe
> we all can detect a solid wall, or thick tree base pretty reliably. What
> concerns me are objects such as fir trees, where their branches come down
to
> the ground, but are sparse, or have open gaps in the needles/branches. I
can
> envision my robot trying to drive right under a fir tree. I really do not
> think normal IR is going to detect this situation. A 2nd problem I see are
> very narrow sticks, or even thin plants in the yard. Indoors I can see
many
> times when my IR can not detect a thin chair leg. Bumper sensors do solve
> this problem normally.
> How to prevent the robot from driving inside a very cluttered area? For
> example a garden may have an opening at the ends where the rows start, but
> inside the garden there are many plants/clutter. Ideally the robot should
> never enter such a clutterd location.
> Speed is not an issue, a slow speed, around 1-2 mph is fine. A 2nd example
> is the edge of woods, where there are hundreds of small saplings/twigs etc
> causing random clutter. A simple IR beam would pass between the saplings,
> and so the robot would try to also.
> 1) Compass for navigation, I am afraid it would be to coarse a measurment.
I
> dont think it would be possible to program all obstacles, and expect a
> compass to keep the robot aligned close enough to its track to avoid them.
> 2) GPS,(thanks to all for replys) It appears that GPS will be from 9-20
foot
> tolerance, unless using DGPS radio signal as well. This basically
eliminates
> GPS for in close navigation.
> 3) Sharp GP2D02 sensors, as well as the new Devantech SRF04 UltraSonic
> Ranger. These items seem like the best method, however now it looks like,
> either the speed of the cpu, and sensing will kill any benifits. Possibly
1
> sensor on a servo, and pan it around? This is good for horizontal, then do
> it again in a vertical direction, rescan horizontally etc? The program
code
> of updating the servo, reading sensor, deciding whats really out there etc
> etc seems like a nightmare to my old brain.
> 4)Video input seems out simply becauase of the cost, and program skill
> necessary.
> 5)Programmed in way pts (obstacles) this then requires again precise
navigation.
> -- I do not care if the robot KNOWS where it is, at any time, just so
that
> it is semi reliable in avoiding objects. I do not think a human Precisely
> KNOWS where they are at, at any one time. We simply walk in between trees
> (Unless on horseback), and proceed to the house. My apologys for the long
> letter, however I have nobody else in my area, to bounce ideas off of.
> Thank you to everyone.
>
>
> Sincerely
> Kerry
> Admin@M...
> WWW server hosting [url=Http://mntnweb.com]Http://mntnweb.com[/url]
> Kerry Barlow
> p.o. box 21
> kirkwood ny
> 13795
>
>
>
>
>
>
> Your use of Yahoo! Groups is subject to http://docs.yahoo.com/info/terms/
>
>
presumptuousness.) with simple, old-fashioned optical pairs. If you want the
robot to proceed through tall grass, some slick programming will be needed,
but you should be able to determine if it has encountered a blade of grass
or a broken branch, or some clutter.
chris
>
Original Message
> From: Kerry Barlow [noparse]/noparse]mailto:[url=http://forums.parallaxinc.com/group/basicstamps/post?postID=pfTMUm2mHmLmb_FR_X60RXSFnRrFNBIg-7uUM5CrKhyzeL4Un9Hu-HCCP3EUJ7PrflbxwsH6ibH23X8]admin@m...[/url
> Sent: Sunday, March 18, 2001 2:45 PM
> To: basicstamps@yahoogroups.com
> Subject: [noparse][[/noparse]basicstamps] Navigation discussion
>
>
> Hello All: I am working on an outdoor robot, and have been
> tossing around
> numerous ideas for navigation. If possible I would like to
> start a small
> discussion on the topic.
> I myself am going to stick with some sort of basic stamp as
> my processor,
> possibly 2 or 3 onboard if necessary. My biggest reason being
> the simple
> programming language.
> My concerns are navigating around small, and odd shaped
> obstacles. I believe
> we all can detect a solid wall, or thick tree base pretty
> reliably. What
> concerns me are objects such as fir trees, where their
> branches come down to
> the ground, but are sparse, or have open gaps in the
> needles/branches. I can
> envision my robot trying to drive right under a fir tree. I
> really do not
> think normal IR is going to detect this situation. A 2nd
> problem I see are
> very narrow sticks, or even thin plants in the yard. Indoors
> I can see many
> times when my IR can not detect a thin chair leg. Bumper
> sensors do solve
> this problem normally.
> How to prevent the robot from driving inside a very
> cluttered area? For
> example a garden may have an opening at the ends where the
> rows start, but
> inside the garden there are many plants/clutter. Ideally the
> robot should
> never enter such a clutterd location.
> Speed is not an issue, a slow speed, around 1-2 mph is fine.
> A 2nd example
> is the edge of woods, where there are hundreds of small
> saplings/twigs etc
> causing random clutter. A simple IR beam would pass between
> the saplings,
> and so the robot would try to also.
> 1) Compass for navigation, I am afraid it would be to coarse
> a measurment. I
> dont think it would be possible to program all obstacles, and expect a
> compass to keep the robot aligned close enough to its track
> to avoid them.
> 2) GPS,(thanks to all for replys) It appears that GPS will be
> from 9-20 foot
> tolerance, unless using DGPS radio signal as well. This
> basically eliminates
> GPS for in close navigation.
> 3) Sharp GP2D02 sensors, as well as the new Devantech SRF04 UltraSonic
> Ranger. These items seem like the best method, however now it
> looks like,
> either the speed of the cpu, and sensing will kill any
> benifits. Possibly 1
> sensor on a servo, and pan it around? This is good for
> horizontal, then do
> it again in a vertical direction, rescan horizontally etc?
> The program code
> of updating the servo, reading sensor, deciding whats really
> out there etc
> etc seems like a nightmare to my old brain.
> 4)Video input seems out simply becauase of the cost, and program skill
> necessary.
> 5)Programmed in way pts (obstacles) this then requires again
> precise navigation.
> -- I do not care if the robot KNOWS where it is, at any
> time, just so that
> it is semi reliable in avoiding objects. I do not think a
> human Precisely
> KNOWS where they are at, at any one time. We simply walk in
> between trees
> (Unless on horseback), and proceed to the house. My apologys
> for the long
> letter, however I have nobody else in my area, to bounce ideas off of.
> Thank you to everyone.
>
>
> Sincerely
> Kerry
> Admin@M...
> WWW server hosting [url=Http://mntnweb.com]Http://mntnweb.com[/url]
> Kerry Barlow
> p.o. box 21
> kirkwood ny
> 13795
>
>
>
>
>
>
> Your use of Yahoo! Groups is subject to
> http://docs.yahoo.com/info/terms/
>
>
>
Video output onboard. I have one of those Rabbit video sendors, that will
work over a short range. Then someday send the video to my Internet website.
Mowing would be great, but not my lawn, to many huge tree roots. I just want
the robot to wander around, it doesnt have to DO anything, except not get
stuck every 2 seconds :-)
I did have an idea of using a Dog radio fence to keep the robot out of the
road. I already have a buried fence. Thats another discussion however, how
to get a valid radio signal off the buried wire. Buy a 2nd dog collar, and
hack into the shock circuit somehow? Seems like a waste of money, if a
simple radio receiver could do the same job.($39-60 for a collar) Maybe tie
up my dog, and rob his collar he heee
I also considerd a degree of pitch feedback, to keep the robot from trying
to climb up any steep hills. I am in the country, so have all manor of
outdoor obstacles. I thought of a simple ball rolling against contacts, if
it gets to steep.
Thank you to everyone for advice and comments.
Sincerely
Kerry
Admin@M...
WWW server hosting [url=Http://mntnweb.com]Http://mntnweb.com[/url]
Kerry Barlow
p.o. box 21
kirkwood ny
13795
obstacles (trees, gardens etc) what about walking around with it and getting
it to remember where it went? Don't know how but a thought!
Bree
>===== Original Message From Kerry Barlow <admin@m...> =====
>Hi Everyone: I did not plan to mow grass with my robot. I did plan to have a
>Video output onboard. I have one of those Rabbit video sendors, that will
>work over a short range. Then someday send the video to my Internet website.
>Mowing would be great, but not my lawn, to many huge tree roots. I just want
>the robot to wander around, it doesnt have to DO anything, except not get
>stuck every 2 seconds :-)
> I did have an idea of using a Dog radio fence to keep the robot out of the
>road. I already have a buried fence. Thats another discussion however, how
>to get a valid radio signal off the buried wire. Buy a 2nd dog collar, and
>hack into the shock circuit somehow? Seems like a waste of money, if a
>simple radio receiver could do the same job.($39-60 for a collar) Maybe tie
>up my dog, and rob his collar he heee
>I also considerd a degree of pitch feedback, to keep the robot from trying
>to climb up any steep hills. I am in the country, so have all manor of
>outdoor obstacles. I thought of a simple ball rolling against contacts, if
>it gets to steep.
>
>Thank you to everyone for advice and comments.
>
>Sincerely
>Kerry
>Admin@M...
>WWW server hosting [url=Http://mntnweb.com]Http://mntnweb.com[/url]
>Kerry Barlow
>p.o. box 21
>kirkwood ny
>13795
>
>
>
>
>
>
>Your use of Yahoo! Groups is subject to http://docs.yahoo.com/info/terms/
know it is exactly in that same location a 2nd time?
I considerd a database in memory, say 10 rows x 10 columns. If a 0 is in a
cell then its open terrain, if a 1 is there, then there is an obstacle.
Similar to the following table
0 1 2 3 4 5 6 7 8 9
1 0 0 0 0 0 0 0 0 0
2 0 0 0 0 0 0 0 0 0
3 0 0 0 1 1 0 0 0 0
4 0 0 0 1 1 0 0 0 0
5 0 0 0 0 0 0 0 0 0
6 0 0 0 0 0 0 0 1 1
7 0 0 0 0 0 0 0 0 0
8 0 0 0 0 0 0 0 0 0
9 0 0 0 0 0 0 0 0 0
The first time through I could program the robot manually, but how to have
the robot stay precise for navigation? This led me to all the GPS querys
previously <S> Again the compass is a great idea, and cheap in price. Up
could be North in the database, however is the robot going north in column
2, or is it going north in column 5? This is what I dont know how to program
around.
Wheel counters, do not seem precise enough outdoors, with wheel slippage in
mud/grass etc
I hope I am not boring people with my questions.
At 09:35 AM 3/19/2001 -0700, you wrote:
>Something that might require a little human intervention... For the big
>obstacles (trees, gardens etc) what about walking around with it and getting
>it to remember where it went? Don't know how but a thought!
>
>Bree
>
>>===== Original Message From Kerry Barlow <admin@m...> =====
>>Hi Everyone: I did not plan to mow grass with my robot. I did plan to have a
>>Video output onboard. I have one of those Rabbit video sendors, that will
>>work over a short range. Then someday send the video to my Internet website.
>>Mowing would be great, but not my lawn, to many huge tree roots. I just want
>>the robot to wander around, it doesnt have to DO anything, except not get
>>stuck every 2 seconds :-)
>> I did have an idea of using a Dog radio fence to keep the robot out of the
>>road. I already have a buried fence. Thats another discussion however, how
>>to get a valid radio signal off the buried wire. Buy a 2nd dog collar, and
>>hack into the shock circuit somehow? Seems like a waste of money, if a
>>simple radio receiver could do the same job.($39-60 for a collar) Maybe tie
>>up my dog, and rob his collar he heee
>>I also considerd a degree of pitch feedback, to keep the robot from trying
>>to climb up any steep hills. I am in the country, so have all manor of
>>outdoor obstacles. I thought of a simple ball rolling against contacts, if
>>it gets to steep.
>>
>>Thank you to everyone for advice and comments.
>>
>>Sincerely
>>Kerry
>>Admin@M...
>>WWW server hosting [url=Http://mntnweb.com]Http://mntnweb.com[/url]
>>Kerry Barlow
>>p.o. box 21
>>kirkwood ny
>>13795
>>
>>
>>
>>
>>
>>
>>Your use of Yahoo! Groups is subject to http://docs.yahoo.com/info/terms/
>
>
>
>
>Your use of Yahoo! Groups is subject to http://docs.yahoo.com/info/terms/
>
>
>
Sincerely
Kerry
Admin@M...
WWW server hosting [url=Http://mntnweb.com]Http://mntnweb.com[/url]
Kerry Barlow
p.o. box 21
kirkwood ny
13795
> Bree: I had thought of a remembering scheme also, but how to have the robot
> know it is exactly in that same location a 2nd time?
> I considerd a database in memory, say 10 rows x 10 columns. If a 0 is in a
> cell then its open terrain, if a 1 is there, then there is an obstacle.
> Similar to the following table
> 0 1 2 3 4 5 6 7 8 9
> 1 0 0 0 0 0 0 0 0 0
> 2 0 0 0 0 0 0 0 0 0
> 3 0 0 0 1 1 0 0 0 0
> 4 0 0 0 1 1 0 0 0 0
> 5 0 0 0 0 0 0 0 0 0
> 6 0 0 0 0 0 0 0 1 1
> 7 0 0 0 0 0 0 0 0 0
> 8 0 0 0 0 0 0 0 0 0
> 9 0 0 0 0 0 0 0 0 0
> The first time through I could program the robot manually, but how to have
> the robot stay precise for navigation? This led me to all the GPS querys
> previously <S> Again the compass is a great idea, and cheap in price. Up
> could be North in the database, however is the robot going north in column
> 2, or is it going north in column 5? This is what I dont know how to program
> around.
> Wheel counters, do not seem precise enough outdoors, with wheel slippage in
> mud/grass etc
> I hope I am not boring people with my questions.
>
>
>
> At 09:35 AM 3/19/2001 -0700, you wrote:
> >Something that might require a little human intervention... For the big
> >obstacles (trees, gardens etc) what about walking around with it and getting
> >it to remember where it went? Don't know how but a thought!
> >
> >Bree
> >
> >>===== Original Message From Kerry Barlow <admin@m...> =====
> >>Hi Everyone: I did not plan to mow grass with my robot. I did plan to have a
> >>Video output onboard. I have one of those Rabbit video sendors, that will
> >>work over a short range. Then someday send the video to my Internet website.
> >>Mowing would be great, but not my lawn, to many huge tree roots. I just want
> >>the robot to wander around, it doesnt have to DO anything, except not get
> >>stuck every 2 seconds :-)
> >> I did have an idea of using a Dog radio fence to keep the robot out of the
> >>road. I already have a buried fence. Thats another discussion however, how
> >>to get a valid radio signal off the buried wire. Buy a 2nd dog collar, and
> >>hack into the shock circuit somehow? Seems like a waste of money, if a
> >>simple radio receiver could do the same job.($39-60 for a collar) Maybe tie
> >>up my dog, and rob his collar he heee
> >>I also considerd a degree of pitch feedback, to keep the robot from trying
> >>to climb up any steep hills. I am in the country, so have all manor of
> >>outdoor obstacles. I thought of a simple ball rolling against contacts, if
> >>it gets to steep.
What about using triangulation with 3+ ir emitters using different frequencies
or different messages? The bot could be programmed to know where it is using
the geometry of the angles where the emissions are coming from.
Mik