Position System like GPS but localized

[JonB]

The Global Positioning Satellite system is a good model but I'm thinking smaller.

Each satellite transmits a very accurate time code and the receivers compare the time signatures to determine location. I am amazed but understand the concept.

But, I have a more localized application and it is inside a shielded building that makes GPS worthless. Imagine a circular area about 7 meters in diameter, made of steel and hemispherical. It has about 70 veritcal piping penetrations located symetrically around the surface. Because of hazardous conditions, a remote control camera is sent in to do visual inspection for leakage. A joystick control moves the robot around this maze, trailing an umbilical of video cables for its front and rear cameras. Each penetration must be looked at from four sides, so 280 video/picture shots must be taken to complete the inspection.

Here's the kicker - the pentrations are not labled at all, so the only way to know which one you are looking at is by "relative" calculation. Several people are involved tracking and counting pipes to verify that the correct penetration was viewed.

What would you put on that robot to be able to correctly and accurately position it? It is a 3-D environment because of the domed surface. I'm thinking about how easy it is for the Nintendo Wii to track its controller, but this is a harsh environment. It isn't wet but all the metal pipes will block normal IR sensors.

Thanks. JonB

 

[Born2bwire]

I know a way of doing this but I can't talk about it. Aren't I helpful ?


I can say this, you're going to have a hell of a time if this little robot guy is wandering around in steel pipes (don't know if the pipes are steel or if the automaton only travels within them) because you won't be able to receive an electromagnetic signal through the conductive steel. What you could do though, is use each pipe as a waveguide. Have a simple antenna at the mouth of each pipe that broadcasts at a frequency that is supported by the pipe and encode a simple number or other piece of information so that you can tell which pipe you are in. As for your position within the pipe, I would suggest using accelerometers and tachometers on the robot's wheels to measure distances traveled.

 

[JonB]

No, imagine this more like crawling across somebody's head where the pipes are their hairs sticking up. I don't care what is in the pipes, just what the base of them look like. I'd like to have an accurate remote indication of where the remote camera is (which hair). Knowing what direction it is pointing would be gravy.

 

[Born2bwire]

Originally posted by: JonB
No, imagine this more like crawling across somebody's head where the pipes are their hairs sticking up. I don't care what is in the pipes, just what the base of them look like. I'd like to have an accurate remote indication of where the remote camera is (which hair). Knowing what direction it is pointing would be gravy.

I think you could still have antennas in the pipes which the robot could detect the individual signals. So basically this is just a bigass metal sphere with pipes that feed off the surface, and you need to crawl around in there and monitor the areas where the pipes meet the surface for defects? Can you enter the inside of the sphere yourself?

 

[Nathelion]

Is this a nuclear reactor or something?

Why don't you just mount an inertial compass on the probe thingy, and have it keep track of how far you've gone and project that on a map of the area?

 

[TuxDave]

I've seen a Panera Bread wired up like that using some sonar based locationing so that the waiters can find where you sit faster. They stuck it in the coaster that you take that has a number when you order.

 

[PowerEngineer]

My first thought is to label the penetrations

Not sure I have a good picture of what this looks like. If the vertical penetrations don't reach all the way to the building's ceiling, then maybe a vertically mounted laser shining up to the ceiling could pinpoint its position.

Actually, I like the wave guide idea best; very clever.

 

[firewolfsm]

Any decent program can figure out where the robot has traveled, in distance and direction, just travel where you haven't until you cover everything. It's basically like leaving a trail of cookies, but avoiding them.

Or you could label everything you've taken a picture of, that is, each side of each pipe. This would add some extra cost to it though.

 

[CycloWizard]

This is eerily similar to the Roomba thread. Most of the techniques proposed there for determining position could be applied here. Your feedback is simply done through the trailing cables in this case.

 

[dkozloski]

Can you string a grid of wires across the floor, "ping" them like a graphics tablet, and have an antenna on the robot. You can get accuracy in the thousandths of an inch.

 

[Born2bwire]

Originally posted by: dkozloski
Can you string a grid of wires across the floor, "ping" them like a graphics tablet, and have an antenna on the robot. You can get accuracy in the thousandths of an inch.

I think the first problem is that the robot seems to be inside a metal sphere. Any electromagnetic radiation or acoustics are going to be bouncing around off the sphere, with some energy coupling and propagating down the pipes (depending upon wavelength). The multiple reflections will probably render anything that is sourced inside the sphere unusable. So having little beacons for a GPS like mapping would not work. However, if we did have a grid of wires, we could have a brush on the underside of the robot, like an inverted bumper car. This brush could complete a circuit from the robot to the grid and we could monitor voltages on the grids to see what two intersecting wires are being excited.

 

[dkozloski]

Originally posted by: Born2bwire

Originally posted by: dkozloski
Can you string a grid of wires across the floor, "ping" them like a graphics tablet, and have an antenna on the robot. You can get accuracy in the thousandths of an inch.

I think the first problem is that the robot seems to be inside a metal sphere. Any electromagnetic radiation or acoustics are going to be bouncing around off the sphere, with some energy coupling and propagating down the pipes (depending upon wavelength). The multiple reflections will probably render anything that is sourced inside the sphere unusable. So having little beacons for a GPS like mapping would not work. However, if we did have a grid of wires, we could have a brush on the underside of the robot, like an inverted bumper car. This brush could complete a circuit from the robot to the grid and we could monitor voltages on the grids to see what two intersecting wires are being excited.

A graphics tablet uses a receiving coil in close proximity to, but not touching the grid. The grid is pulsed and the signals received at the coil are compared for timing to locate the coil. I believe there are also systems where the coil is pulsed and he signals at the grid processed for position.

http://en.wikipedia.org/wiki/Graphics_tablet

 

[JonB]

Nathelion: Is this a nuclear reactor or something?

Actually, it is the hemispherical pressure head of a commercial pressurized water nuclear reactor. Good guess.

You can't put anything IN these vertical tubes; they are the control rod tubes and they have something in them already.

Something optical may be the only usable method. I don't know why they (the "metallurgy guys") won't label anything; possibly a chemical contamination concern for ink or glue, and there's nothing on which to hang a metal label.

I am thinking about the wire grid idea, though. Perhaps if the first thing the robot did was drag a "leaky antenna" wire across the head in two axes. The strength and phase of the received signals might give a good position.

 

[AeroEngy]

Originally posted by: Nathelion
Is this a nuclear reactor or something?

Why don't you just mount an inertial compass on the probe thingy, and have it keep track of how far you've gone and project that on a map of the area?

Mentioned already but I didn't see any response, why not inertial navigation.

Here is a link to a sample unit

That one is pretty small only 9 in^3 and under 0.5 lbs.

The only real drawback is that they tend to drift over time and you have to tell the unit where it is when you turn it on. But when used in short durations they can be pretty accurate.

 

[Throckmorton]

Isn't RFID used for this? I guess you'd need at least 4 base stations.

 

[AeroEngy]

Originally posted by: Throckmorton
Isn't RFID used for this? I guess you'd need at least 4 base stations.

I don't know of any inertial navigations systems that use RFID.... that doesn't mean that one doesn't exist that uses something like that to initialize its position.

For anyone that doesn't know how a INS works here is a very basic explanation.
The unit typically has 3 linear accelerometers to measure the accelerations along X,Y and Z. It typically has 3 angular accelerometers to measure rotations about X,Y, and Z.

The unit determines it position from the initial state (which must be input in some fashion) and integrating the measured accelerations to determine velocity (angular and linear). Then integrating again to determine position and orientation. This is continuously done so the INS always knows where it is ... with some error.

The drawback is that there is error in the integration from noisy sensors etc. this error is compounded with every computational cycle. So the longer the unit is on without given an positional update from an outside source the worse the error gets. But for relatively short durations or if you use it with a GPS unit and a Kalman filter it is pretty accurate.

 

[kylef]

Back in 2001 in graduate school for a sensor networks class, a few of us worked on a hybrid radio/ultrasonic position localization system. We only got it to work in 1 dimension before we ran out of time, but it was pretty cool.

The basic idea is that, for small distances, the radio chirps arrive at your location "instantly", and the ultrasonic chirps arrive at the speed of sound. You can calculate the differential and obtain the distance to the transmitter, down to a granularity of about 1 centimeter. Obviously this can be extended to 3 dimensions without too much more effort.

We bought PIC microcontroller boards that had onboard FM tranceivers, and some ultrasonic transmitter and receiver/amplifier boards. We hooked the ultrasonic receivers up to the ADCs on the PIC to digitize the signal.

Getting the sampling right and tuning it took quite some time, but we eventually got the thing to work right. For a while we were fighting ultrasonic reflections from nearby surfaces, but eventually we worked out a pretty reliable algorithm for rejecting reflections.

Obviously nothing like this is available "off the shelf" at Fry's, but you could build it with enough time and money.

 

[Modelworks]

Hall effect sensors.
Instead of the sensors being on the robot, place them on each column.
If thats not possible, make a grid that contains enough sensors to cover the area and lowered just far enough to be activated by the robot .

The robot then carries rare earth magnet(s).
As the robot nears the sensors they will show the position.
And hall effect sensors can have very high resolution if needed.
They also can be packaged so that they can be placed in even the worst of environments.

 

[degibson]

There are accurate through-the-air sonar sensors available... I recall using them in college physics. Stick four of these things on your robot, couple it with some kind of directional indicator (in a big steel dome, in a PWR, magnetic compass might not work so well... suggestions?), and you have positioning via sonar rebounds. Its not clear to me if the rods are pulled when this procedure is going on-- obviously if the rods are at robot-level the sonar will have some problems.

I'd advise against the use of the accelerometers, due to error accumulation, but tachometers will be helpful, provided you have highly accurate control of your wheels / tracks / whatever.

 

[Throckmorton]

Originally posted by: AeroEngy

Originally posted by: Throckmorton
Isn't RFID used for this? I guess you'd need at least 4 base stations.

I don't know of any inertial navigations systems that use RFID.... that doesn't mean that one doesn't exist that uses something like that to initialize its position.

For anyone that doesn't know how a INS works here is a very basic explanation.
The unit typically has 3 linear accelerometers to measure the accelerations along X,Y and Z. It typically has 3 angular accelerometers to measure rotations about X,Y, and Z.

The unit determines it position from the initial state (which must be input in some fashion) and integrating the measured accelerations to determine velocity (angular and linear). Then integrating again to determine position and orientation. This is continuously done so the INS always knows where it is ... with some error.

The drawback is that there is error in the integration from noisy sensors etc. this error is compounded with every computational cycle. So the longer the unit is on without given an positional update from an outside source the worse the error gets. But for relatively short durations or if you use it with a GPS unit and a Kalman filter it is pretty accurate.

I was talking about triangulating the position of an RFID tag

 

[dkozloski]

How about something like the now defunct MouseSystems mouse arrangement that was used on Sun Microsystems computers that optically tracked a grid printed on the metallic mouse pad? The repeatability was very impressive.

 

[gsellis]

Originally posted by: Throckmorton
Isn't RFID used for this? I guess you'd need at least 4 base stations.

That is what I would think to do so that you could use off the shelf parts. Very localized, and would require premapping to place and encode the tags.

 

[JonB]

Originally posted by: degibson... (in a big steel dome, in a PWR, magnetic compass might not work so well... suggestions?), and you have positioning via sonar rebounds. Its not clear to me if the rods are pulled when this procedure is going on-- obviously if the rods are at robot-level the sonar will have some problems.

The control rods or their motion wouldn't be a problem. The vessel head is detached and moved to a work stand at this point during the refueling. While the vessel head is electrically dead, it is still a strong radiation field and that's why we use a robot.

 

[degibson]

Originally posted by: JonB

The control rods or their motion wouldn't be a problem. The vessel head is detached and moved to a work stand at this point during the refueling. While the vessel head is electrically dead, it is still a strong radiation field and that's why we use a robot.

Ok, so I'm a complete radiation n00b, but does the presence of all this radiation disqualify RF-band solutions? Moreover, are there enough high-energy particles flying around to affect the robot's processor significantly? That would change the playing field significantly.

If the answers above are yes and no, repsectively, I'd try the sonar approach. If a poor rural high school could afford them in 1997, I'm sure an organization with a PWR could get a dozen of them for well within their budget.

A five-second google search yields: http://www.acroname.com/robotics/parts/R93-SRF04.html

The kicker again is the radiation -- I don't know how the sensor will be affected.

 

[johnpombrio]

Think simple here.
Remember, people are terrific at remembering patterns. Just figure out a way to create a visual pattern and people would be the best navigators.
Take it you cannot go in. If you could, just paint the damn things a different color.
If the robot can just go in, have it carry a spray can and a solinoid to trigger the spray can.
Stick a magic marker or a sponge full of paint on the end of a stick and run it into a pipe to create a dot pattern.
If the pipes are hot, stick a marble with a high temp slow set epoxy onto the end of a stick. Stick it to the pipe. Use different colored marbles or different number of marbles.
Stick a kids sticker of a different flower shape on each of the pipes or on the sphere near the pipe.
Create vertical stripes on the ground with paint in various patterns.