Realtive Position Sensors

[JFChalmers]

I am trying to find a position sensor that can measure the relative position in respect to a control line (not point) in 2-D. I do not need to know the height of the object, just the XY position. The control sensor I assume would be 2 separate sensors creating a line in the +- y directions at the x origin. These sensors would be able to determine the x position of the third sensor placed on the object. Essentially I need the perpendicular distance of the third sensor to the control line. The y line created by the 2 control sensors (if that is the best way of doing so) could be anywhere from 1 to 150 feet from the third sensor on the object. Also, the third sensor on the object will be obstructed from view (does this eliminate IR?) by other objects such as people, clothing, weather etc. The sensor needs to be durable (I assume you can create a safe environment for any sensor).

I am up for using any kind of sensor as long as it is available to the public and can be purchased and setup at a reasonable price. I am highly experienced in matlab/simulink and would prefer using that in my system.

So far people are have suggested IR and CNC, are these viable options?

Thanks for your help and I hope there is a available solution.

 

[berkeman]

It's a bit difficult to visualize what you are asking. Is the Z dimension vertical? And you want to know the X,Y position of the sensor in a room full of people, like in a department store or something? You say the sensor may be obscured from view from the reference line, and may be from 1' to 150' away?

The "obscured from view" part will make this hard. RF is probably not an option due to multipath issues. Can you say more about what you are trying to do, in case that helps us brainstorm ideas for you?

 

[JFChalmers]

The z dimension is vertical. I do not need to know the z-value though in this particaulr application. In the setup there is a datum in which I want to measure the x-value of an object away from the datum to determine when the object crouses the datum. I assume to create a datum you would need two sensors defining the datum running in the along the y-axis. Essentially I want to measure the distance perpendicular to this y-datum-axis. The datum would be about 150 feet long. I do not need to know the y value of the object but I was using the value of 150 because of the assumed length of the datum line and where I figured you'd have to place the control-datum-sensors.

The object being measured from the datum place will often not in clear line of sight from the datum, again being obstructed by people, clothing, weather conditions, debris etc. The purpose of this system is to detect exactly when the object crosses the datum plane.

besides visual sensors, an idea I have is relative GPS. does this exists/is it able to be created from several components/is it accurate to or less then an inch? Another Idea I have was magnetic sensors...could the datum plane contain many control magnectic sensors and the object be rigged with their counter parts? I am open for any ideas and want to get the ball rolling ASAP. Thanks for all your help and I hope there is a available option.

 

[JFChalmers]

I found that GPS is only accurate to 1 to 10 meters. I have yet to find much information on magnetic sensors and will continue to do research.

Motion capture for sports video games is optical correct? hence the large white balls attached to the athlete. That was another thought if it is not optical due to the obstructions.

 

[berkeman]

If you can have the sensor go to a known starting point first, then you might be able to use accelerometers to sense changes in position and an RF link to have the sensor report those changes to your base station.

Alternately, if you could put marks on the floor, you could have the sensor look down, to see its absolute position (like reading barcodes on the floor?)...

 

[mgb_phys]

I found that GPS is only accurate to 1 to 10 meters.

RTK GPS measures the relative position between two antennae and is accurate to mm, but the antennea are large (10-12in dia) and a setup costs around $50k

I have yet to find much information on magnetic sensors and will continue to do research.

Generally magnetic systems are very limited in range.

Motion capture for sports video games is optical correct? hence the large white balls attached to the athlete. That was another thought if it is not optical due to the obstructions.

Optical would probably be the easiest, it sounds like you are measuring motion across a plane rather than toward/away from the camera? A video system is perfect for this.

If you need to measure position along a straight line then a laser rangerfinder is another good choice. Phase systems like a 'Leica Disto' will measure to <1mm but take a few seconds per measurement. Time of flight systems can measure >1000 times/second and are accurate to a few cm.

 

[NoTime]

I found that GPS is only accurate to 1 to 10 meters.

For your basic standalone GPS that is fairly accurate.
Not quite sure how it is setup, but the GPS systems they use for geological ground motion and fault tracking claim at least centimeter resolution.

 

[berkeman]

For your basic standalone GPS that is fairly accurate.
Not quite sure how it is setup, but the GPS systems they use for geological ground motion and fault tracking claim at least centimeter resolution.

Maybe they just average bzillions of measurements to make it more accurate? The ground isn't moving very much, so maybe the GPS satellites broadcast offset data for non-military receivers, but the average per day is more accurate? I don't honestly know, but that would be a guess. I doubt the geological folks are able to license/use military-accuracy receivers -- there would be too big a black market for them!

 

[JFChalmers]

As for the accelerometer, would there be a way to constantly reset the starting point? That would work but the object would be continually moving, stopping, then moving. Every cycle when it is not in motion a sensor could take the starting position. Maybe a IR sensor at that point? The barcode idea intrigues me and I would like to hear more. Similar to how the microsoft surface works with IR and a patern on the ground being the barcode? How consistent and accurate would the "barcode" have to be? could you intigrate it into the ground somehow? preferably grass.

as for the 50K GPS that's just alittle less then a years pay so I think that is out of the questions. Also the object being measured is about the size of a bread basket.

As for what exactly is happening witht he object, it will be traveling in the xyz position but but I am only concern with the x value from the y datum and when it crosses it in the x direction.

You would have to show me one of these GPS's that are accurate to convince me but if there is one out there not for 50K, let me know.

Again, concerning the optical approaches which is the easiest, cheapest, and very accurate, there is still the issue of obstructions. There WILL be people, clothing, and weather conditions in the immediate surroundsings and is part of the objective, to find a way to deal with this.

A continued thanks for all the help.

 

[mgb_phys]

the GPS systems they use for geological ground motion and fault tracking claim at least centimeter resolution.

It's called real-time-kinematic (RTK) GPS. You have multiple receivers = a base station and any number of rovers. The system measures the phase of the clock signal at the base and the rover simultaneously, it uses a radio data link to compare the values.
The system uses the phase difference to measure the relative position of the receivers very accurately.
The absolute position accuracy is only as good as the GPS position of the base ( 1m with WAAS or correction) so you often put the base station on a known survey point.

 

[JFChalmers]

accurate up to 1 inch or less? That is exactly what I am looking for, just not that expensive or large. The control datum has no size constraints, but the test object again is small, the size of a shoebox, and the weight cannot be effected drastically. Can this be accomplished? It would be very helpful if I could find the orientation of the object also, which seems like the case with this GPS. Please continue to explain.

 

[NoTime]

Maybe they just average bzillions of measurements to make it more accurate? The ground isn't moving very much, so maybe the GPS satellites broadcast offset data for non-military receivers, but the average per day is more accurate? I don't honestly know, but that would be a guess. I doubt the geological folks are able to license/use military-accuracy receivers -- there would be too big a black market for them!

I don't really know how they do it, but it might be worth looking into.

A particular GPS receiver (one you can buy in the store) can often go back to the same position within a meter or so, even though its general accuracy to a map position might be 10 meters.
I suspect the difference is in the quality of the internal time base (a large expense item).
Within a short period of time a relative displacement of 150' may be fairly accurate, even for a handheld unit.

At one time the military degraded GPS accuracy for commercial users.
It's my understanding that they turned off the degradation feature some time ago, so I don't know that there is actually a difference between military and high end non-military receivers anymore.

 

[mgb_phys]

Can you prepare the surface the robot (shoebox) is moving over?
If you can draw a grid either opticaly or by buried wires then it's very easy for the box to sense it's position on the grid/wires,this avoids the obstruction and doesn't need any data link to the box - it's used for factory robots and automatic trucks/freight carts.

A camera pointing at the box can give an accurate position if the range of movement isn't too large. Assume the box can move say 10m and you can measure it's position to a fraction of a pixel you would be able to measure the position to a few mm with only a 640x480 type camera. You might be able to avoid the obstruction by having multiple cameras, having IR leds on the box and IR filters on the camera makes it easier to pick the box out of a complicated scene.

 

[mgb_phys]

so I don't know that there is actually a difference between military and high end non-military receivers anymore.

The only difference is that military sets use both GPS frequencies which allow you to cancel out some of the atmospheric effects. But you can also use a second broadcast correction signal (USA=WAAS EU=EGNOS) to give accuracy of a couple of m.

Simply averaging the signal doesn't necessarily give improved accuracy, it removes random noise in the receiver but not larger scale atmospheric effects.

 

[JFChalmers]

I can prepare the surface, but the object is moving eradictly and quickly. And again I am only concern about it crossing the datum line. So the only time and position I am measuring is exactly during, and after it crosses the datum (+- 0 inches). The goal is to set an alarm off when it cross the datum and while it is in the area beyond the datum. So If I were to prepare the surface it would just be directly before and completely past the datum?

I am still interested in the RTK GPS, is this a realistic approach?

If the grid were buried wires, what sensors would be used there?

Can you prepare the surface the robot (shoebox) is moving over?
If you can draw a grid either opticaly or by buried wires then it's very easy for the box to sense it's position on the grid/wires,this avoids the obstruction and doesn't need any data link to the box - it's used for factory robots and automatic trucks/freight carts.

 

[NoTime]

If all you are interested in is the crossing, then a single high contrast line and optical sensor would probably work. Perhaps white chalk on grass.

 

[JFChalmers]

the problem is still the accuracy of the crossing. The optical sensor would see the datum as the object approaches it, not just as it crosses it. The object will be moving eradictly. Is there a solution for this?

for the RTK GPS, can they be small enough to fit inside a shoebox size object?

Explain more about the grid also, a grid of wires? what would the sensor be?

 

[NoTime]

the problem is still the accuracy of the crossing. The optical sensor would see the datum as the object approaches it, not just as it crosses it.

I'm not sure why that would be true.
I would think that a vertically aligned sensor with a constrained field of view would only see the line as it passed underneath the sensor.
Multiple brightness sensors or a video camera as a sensor with some software should be able to determine if the robot crossed perpendicular to the line or at an angle and measure the width of the line in the process.

 

[JFChalmers]

The sensor will not be vertically aligned unfortunately. It is not a controlled environment in this case. The object will be moving and articulating erratically in all directions and again will be obstructed often. I will look into it but believe that it won't be completely accurate because of this. A pattern on the object that the datum recognized is still my most viable solution, but still does not resolve the obstruction dilemma. A series of cameras could be place all along and around the datum in the yz plane looking inwards as to catch a glimpse of the object being obstructed possibly. Can you narrow the optical sensors vision to a narrow line like so? Also, what optical sensor would be best for detecting a pattern on the object? IR? Thanks.

I'm not sure why that would be true.
I would think that a vertically aligned sensor with a constrained field of view would only see the line as it passed underneath the sensor.
Multiple brightness sensors or a video camera as a sensor with some software should be able to determine if the robot crossed perpendicular to the line or at an angle and measure the width of the line in the process.

 

[NoTime]

You can't have on sensors on board the object?
That would seem to eliminate the buried wire idea and GPS as well.

No transmitters either (optical or RF)?

Sounds like you can paint it, but that doesn't seem to make a lot of sense if there is a constraint on modifying the object.

 

[JFChalmers]

There can be a sensor on bored, yes, but it can't change the weight or CG significantly (i.e. about 1 oz.). What is the buried wire idea, I am still confused about that. Is it optic sensors in the ball looking at wires or some other sensor. there can be a transmitter, on the object or the ground.

If I could find some kind of paint or pattern that sensed by the sensor, what sensor would I use?

 

[NoTime]

Ball?
What sport are we talking about here?

Having the ball fly thru the air, rather than being in contact with the ground, would probably eliminate optical and wire proximity sensors.

 

[JFChalmers]

The object will be moving eradictly in xyz space but I am only concerned with the 1-D x value of the object from the datum. Neither will work if it is not in contact with the ground?