Finding Positions of Multiple Robots using Sensors.

[roots_of_culture]

Hi all,
i am trying to experiment on building a smart navigational platform
for finding positions of multiple robots instead of using the
conventional camera - image processing.
( use - similar to Robo cup -
http://www.cs.cmu.edu/~jbruce/robocup.html)

So far i have only come been able to think of using infra-red sensors
mounted on the front and back of my robot
(http://www.segyung.com/shop/proc/vi...l_part_no=MARO-10 BODY&vendor=D099&maker=M272)

On scale the robot measures 12 cm by 12cm. So for experiment purposes
i will place a 3x3 grid (each grid box = 12cm x 12cm) so the most basic
design i have thought of is an infra red transmitters placed at the
centre of each grid, trasmitting their unique grid location.

The infra-red recievers on the robot will recieve these signals(grid
location) and transmit.

There are many loop holes in this design.
Any suggestions of other possible methods or even maybe other sensors
that can be implemented would be much appreciated.

 

[Chris]

Hi all,
i am trying to experiment on building a smart navigational platform
for finding positions of multiple robots instead of using the
conventional camera - image processing.
( use - similar to Robo cup -
http://www.cs.cmu.edu/~jbruce/robocup.html)

So far i have only come been able to think of using infra-red sensors
mounted on the front and back of my robot
(http://www.segyung.com/shop/proc/vi...l_part_no=MARO-10 BODY&vendor=D099&maker=M272)

On scale the robot measures 12 cm by 12cm. So for experiment purposes
i will place a 3x3 grid (each grid box = 12cm x 12cm) so the most basic
design i have thought of is an infra red transmitters placed at the
centre of each grid, trasmitting their unique grid location.

The infra-red recievers on the robot will recieve these signals(grid
location) and transmit.

There are many loop holes in this design.
Any suggestions of other possible methods or even maybe other sensors
that can be implemented would be much appreciated.

Hi, Roots O' Culture. It looks like these robots (at least the
CMDragons, small division) use above-field cameras to get a positional
fix. Very heavy on the central intelligence, perhaps?

There aren't any really easy ways to do this, if you need better
resolution than you can get from a GPS. Your 10,000+ LED idea (3m X 3m
field) actually has its advantages if you want truly decentralized
positioning. If you have a robot with directional control, you can
periodically get updates on position, and then just keep your position
current by working from your last navigational fix. You don't have to
be exactly over an LED every second.

But since you're talking navigation and you want to use light, how
about using triangulation, like a ship? Have two or more LEDs in
corners of the room being modulated at different frequencies. Then
have a rotating periscope, which can get a fix on the direction of at
least two LEDs. You can then calculate pretty clearly where you are,
if you have good control of the angle of the periscope, and use a lens
as well as an analog sensor which would allow you to find the angle at
which you have maximum light input. Again, having periodically gotten
a positional fix, you can calculate where you are by figuring wheel
rotation. The negative of this is that at least two LEDs have to be
visible from anywhere on the field you're trying to get a fix. Also,
your robots' periscopes might interfere with each other, so you might
want them to come up one at a time, or position the LEDs high enough so
they can all see them.

Just an idea. The concept is pretty easy in theory, but gets complex
when you put it into practice. Either you pay for a lot of LEDs and
have a requirement for a transparent field surface that allows you to
see the LEDs, or you burden the robot with a periscope/lens, driven by
a geared stepper motor to give you good angular positioning. Also,
you'll need a couple of analog filters to maximize the appropriate
frequencies for sensing, and additional computational power on the
robot.

This is only one idea. There are many other ways to accomplish this.
A lot depends on the requirements of your problem.

One other thing. Good soccer players have situational awareness, but
they don't spend a lot of their computational time getting a positional
fix. There are lines and goals they will orient toward, but they
always have a good idea of where they are on the field while they're
running, only occasionally updating their position with a glance at
landmarks. Most robots should be able to calculate where they are from
wheel rotation.

Possibly you could take some time and explain exactly what you're
trying to do, or if this is just blue sky stuff. Robotics can be an
expensive hobby, and it's trivially easy to make expensive mistakes.

Good luck
Chris

 

[roots_of_culture]

Thanks yet again Chris, u have given me more information to consider.

Possibly you could take some time and explain exactly what you're
trying to do, or if this is just blue sky stuff.

I am basically trying to go away from the on-field camera's - it will
still be a centralised system, will implement a nordic chip. At the
moment I am only trying to achieve positioning of the robots, will give
thought to control of movement of robots at a later stage.

Light sensors are the only ones i could think of. Are there any other
types of sensors that i can implement to achieve positioning?

For experimentation i will go with a 3 X 3 grid, each 12cm X 12cm, so
in effect my total area 36 cm X 36 cm.
- Will develop a GUI which will be a representation of the field grids.
- want to be able to know positions of minibots on the grids as they
move so as to display them on the GUI

But since you're talking navigation and you want to use light, how
about using triangulation, like a ship?

I did think about using triangulation, but it seems an expensive and
complicated affair.

Thought of placing position recievers in the centre of each grip and
transmitters on the front and back of the robots.
Implementing the same methods used in keyboards read the sensors.
The other option is to have transmitters in the centre of each grid,
transmitting grid location and recieved on the robots to be then
transmitted wirelessly..

which seems better?
also any other ways of achieving finding grid locations of robots
moving randomly???
Any and all help much appreciated.

 

[Chris]

Thanks yet again Chris, u have given me more information to consider.


I am basically trying to go away from the on-field camera's - it will
still be a centralised system, will implement a nordic chip. At the
moment I am only trying to achieve positioning of the robots, will give
thought to control of movement of robots at a later stage.

Light sensors are the only ones i could think of. Are there any other
types of sensors that i can implement to achieve positioning?

For experimentation i will go with a 3 X 3 grid, each 12cm X 12cm, so
in effect my total area 36 cm X 36 cm.
- Will develop a GUI which will be a representation of the field grids.
- want to be able to know positions of minibots on the grids as they
move so as to display them on the GUI

I did think about using triangulation, but it seems an expensive and
complicated affair.

Thought of placing position recievers in the centre of each grip and
transmitters on the front and back of the robots.
Implementing the same methods used in keyboards read the sensors.
The other option is to have transmitters in the centre of each grid,
transmitting grid location and recieved on the robots to be then
transmitted wirelessly..

which seems better?
also any other ways of achieving finding grid locations of robots
moving randomly???
Any and all help much appreciated.

Hi, Roots O' Culture. If "The Dance of the Roombas" requirement has
any kind of centralized control, cameras would be by far the best
method of determining location, and unless you're willing to provide
further information (like a fairly complete description of what you're
actually trying to accomplish with these robots, the size of the arena
they're working in, and how many zeros you're willing to put behind the
project cost estimate), there's not much point in just talking at
length.

You have to remember the number of LEDs is a square function. I
suggested 10,000+ LEDs for a square the size of a small room (3m X 3m).
If you've got a soccer field, placing LEDs on a 3cm X 3cm grid will
consume much of the world's stock of surplus LEDs. Remember also that
you've got to power and wire these things, and also that each LED has
to be able to provide some information your robot can use to determine
position. If you wanted to do this, each LED might have a serial
bitstream output containing the positional information. Of course
you'd need localized control for this (possibly a few thousand PICs?
;-). You'd also need coordination between PICs to prevent adjacent
LEDs from stepping on each others' transmissions. This just isn't real
for a soccer field-sized arena.

There's a reason why the graduate students at CM used cameras here. To
paraphrase Don Lancaster (the Hardware Hacker Guru), the immediate
solution is almost always inferior to the one you can think out. There
are times for a brute force solution -- usually because it's exactly
what the customer wants. After all, engineering is solving problems
for people. If a factory owner's main goal is having you make
something he can fix himself, or that he can explain to any maintenance
person he hires, that's a good solution for him (assuming the brute
force solution will work). This isn't one of those times.

Good luck
Chris

 

[Bob Masta]

Hi all,
i am trying to experiment on building a smart navigational platform
for finding positions of multiple robots instead of using the
conventional camera - image processing.
( use - similar to Robo cup -
http://www.cs.cmu.edu/~jbruce/robocup.html)

So far i have only come been able to think of using infra-red sensors
mounted on the front and back of my robot
(http://www.segyung.com/shop/proc/vi...l_part_no=MARO-10 BODY&vendor=D099&maker=M272)

On scale the robot measures 12 cm by 12cm. So for experiment purposes
i will place a 3x3 grid (each grid box = 12cm x 12cm) so the most basic
design i have thought of is an infra red transmitters placed at the
centre of each grid, trasmitting their unique grid location.

The infra-red recievers on the robot will recieve these signals(grid
location) and transmit.

There are many loop holes in this design.
Any suggestions of other possible methods or even maybe other sensors
that can be implemented would be much appreciated.

If you have a fairly well-defined area to cover, you could
have an IR laser that scans the entire area. Modulate
the beam according to scan position, so that when any
robot detects the beam hitting it, it can demodulate and
read the coordinates directly.

Another approach might involve two different-wavelength
scanning lasers, running orthogonally. Each robot would note
the difference between arrival times to figure out where it
was.

There are probably lots more approaches like this.
Old radio-navigation systems like aircraft "omni"
used rotating beacons where the phase was
modulated according to angle. Upon receiving 2
beacons, you knew the angles of your craft relative to
those beacons and you could triangulate your
position. To do this with IR you'd need some way
to distinguish different beacons. But if you have
a master control over the whole area, you could
alternate bursts between two (or more) beacons
so they didn't interfere, and have each beacon send
some ID preamble so you know which is which.

You might even want to do this with RF instead
of IR, just like the old systems. Since all synchrony
issues are under your control in a localized area,
there may be some shortcuts compared to the
old methods.

Just some ideas to get you started. Best regards,





Bob Masta
dqatechATdaqartaDOTcom

D A Q A R T A
Data AcQuisition And Real-Time Analysis
www.daqarta.com
Home of DaqGen, the FREEWARE signal generator

 

[Rich Grise]

Thanks yet again Chris, u have given me more information to consider.


I am basically trying to go away from the on-field camera's - it will
still be a centralised system, will implement a nordic chip. At the
moment I am only trying to achieve positioning of the robots, will give
thought to control of movement of robots at a later stage.

Look into LORAN (LOng RAnge Navigation) - it somehow uses ranging from two
(or maybe more) different transmitters to triangulate itself - here's an
idea - two quadrature antennas, tuned to two separate transmitter
frequencies, with the two transmitters at known coordinates. The
quadrature signals from the two antenna sets should give you a direction
to each of the transmitters; do a little trig, and there you are!

Good Luck!
Rich

 

[James Thompson]

Hi all,
i am trying to experiment on building a smart navigational platform
for finding positions of multiple robots instead of using the
conventional camera - image processing.
( use - similar to Robo cup -
http://www.cs.cmu.edu/~jbruce/robocup.html)

So far i have only come been able to think of using infra-red sensors
mounted on the front and back of my robot
(http://www.segyung.com/shop/proc/vi...l_part_no=MARO-10 BODY&vendor=D099&maker=M272)

On scale the robot measures 12 cm by 12cm. So for experiment purposes
i will place a 3x3 grid (each grid box = 12cm x 12cm) so the most basic
design i have thought of is an infra red transmitters placed at the
centre of each grid, trasmitting their unique grid location.

The infra-red recievers on the robot will recieve these signals(grid
location) and transmit.

There are many loop holes in this design.
Any suggestions of other possible methods or even maybe other sensors
that can be implemented would be much appreciated.

Setup coils in each grid etc. have each robot have a transmitting coil with
its own distinct freq. as each robot get closer to a grid coil, its signal
is picked up there and you will know where each robot is withing the grid.
If your grid is 6 ft x 8 ft, then you have a coil centered on every sq ft
you have 48 coils to identify where the bots are. You could even use signal
strength recieved from multiple coils to plot then when between coils.
You could wire the coils up as a matrix, giving you 6 x 8 wire outputs where
2 wires would have the strongest signal. With each bot having its own freq,
then you have instant positional data for each bot.. Give some thought..
Jim