Radio/magnetic field: detect direction of antenna

[petterg]

Maybe there is a misunderstanding here of the orientation of coils.

I guess it is. When I say vertical coil I think of the coil produced when holding my finger vertical and spin a wire around it. The wire then will be somewhere near the horizontal plane, while the axis and magnetic field in the coil will be vertical. I guess you call this coil horizontal because it's wire is close to horizontal, while I say it's vertical because of the direction of it's axis.

Still, the method you suggest requires a vertical and a horizontal coil to change phase simultaneously, which they won't.

The phase of the signal is determined by the direction the flux passes through the coil, use your “Right Hand Rule”. The horizontal plane coil, flat with the floor, will always be above the wire so it will have a phase and magnitude determined by the left/right horizontal position error. The vertical plane coil will have a phase determined by the direction faced along the wire. That phase reverses if you turn the vehicle 180°, but so does the left/right phase picked up by the horizontal plane coil, so the phase detection still steers correctly along the line.

With that misunderstanding cleared out, let's check if there is another misunderstanding too. As far as I understand the second statement in bold from the quote above says the exact opposite of the first statement in bold. They both can't be correct. To my knowledge the first statement is correct, the second is not. If they both were correct the problem would be solved. (I hope you see a second misunderstanding here.)

 

[sophiecentaur]

Is it absolutely essential to use just one wire? If you can use some 2 dimensional array of wires, the problem becomes much more tractable because you can be dealing with time / phase rather than the amplitude of a signal. Hyperbolic navigation systems work so well that they were used very successfully until GPS came along. The implementation becomes almost trivial with the possibility of a bit of intelligence on board (Arduino board, for instance). You no longer need to be dealing with 'direction', which relies on complicated transmitting or receiving antennae and you just rely on 'timing', which can have a relative precision of less than 1ns (giving 1ft)

 

[petterg]

Radiobased hyperbolic navigation would surely be the best. But to my knowledge this would require free visual to transmiters That will not be the case here. It has to work even when obstacles move around. Making a grid if wires is not an option either.

I still haven't found any reason why the sawtooth-signal won't work. That is a good sign.

 

[sophiecentaur]

Why would it need "visual"? The presence of any conducting object would have more perturbing effect on the level of received signal than on the phase, so the time method still works out better. You would not need a 'grid' of wires - just three sources (minimum) or four (better) sources - short monopoles or vertical axis coils, placed one each end and one on either side.

 

[petterg]

You're right. Hyperbolic will be a lot better. If it works I can kick out all other attempts to keep track of position. They are all estimates with a huge error.

All I know about hyperbolic is: A antenna receives a wave from three (or more) sources and calculate the distance to each of them.
My biggest concern is that the timer in an ardurino won't have the resolution required to do the timing.

How does the receiver know which wave came from which sender? Do all senders send on different frequencies? Do the senders communicate in any way to synchronize when they send out the waves?
Is there a minimum/maximum distance sender-sender or sender-receiver?

 

[sophiecentaur]

The arduino needn't do the timing. The Arduino only needs to get the phase information from some other circuit (a mixed down version of the RF). In the Decca Navigator system (which used low frequencies appropriate for the large distances involved), four different carrier frequencies were used and the phase between them and four references was measured by having four receivers. In the Decca system, one transmitter was the master and the others synced with it. In this case, all four could originate from the same master oscillator and could have nice integer relationships. There's a lot of basic stuff about hyperbolic systems and the Decca Navigator on Wiki.

 

[petterg]

Now I've read quite a bit of several versions of hyperbolic navigation. It comes down to the fact that arduino runs at 16MHz, We must take into consideration that a signal may be received just after one clockcycle. Then it won't be processed before the next clock. If this should be based on timing then the time between two clocks alone makes an error of 18,73m. That's worse that what I already got without knowing which side of the wire the robot is on. Either I'm to dumb for understanding this, or timing based navigation cannot be used.

The other option is based on interfering waves. To get a reasonable resolution we're foreced to run on quite high frequencies. We hit frequencies that cannot be freely used. So lowest possible frequency that can be used is 2,4GHz band. With this high frequency we get pretty small grids of waves. Seems like some old wlan routers will be useful. But I can't figure out how to decide which point of intersecting waves the sensor is located in. It seems to me that when this kind of navigation was used, ppl knew approximately where they were located, so when they hit a point where waves were intersecting they know from their manual approximation which intersection that was.


I think I'm falling back to the sawtooth again. Still nobody have said anything of why it shouldn't work.

 

[sophiecentaur]

You are confusing phase resolution with processor speed. Unless your robot is moving very fast (a large fraction of a wavelength each processor cycle) then the processor speed really doesn't matter. It just looks at the phase of a much lower frequency signal (beaten down version). Time difference, appears as phase of the IF frequency, which can be chosen to be anything convenient (it's a Delta t 'magnification' process' due to the frequency ratio). This is the same idea as when a sampling scope looks at extremely high frequencies and sub-samples at a manageable frequency for its circuitry. A time domain reflectometer can resolve distances of a fraction of a cm in this way, too. I think the processor speeds used to be just a few MHz in the TDRs I used in the 80s.
For the powers you are interested in (very short range operation), there is a pretty wide choice of operating frequency - the WAN frequencies would seem a good choice as the antenna would be pretty small.
BTW, you needed to know nothing extra about where you were for Decca to tell you - as long as you were on the appropriate map of the North Sea / English Channel etc. The channels of the transmitting stations would tell you that, in any case.
But Decca was real steam radio stuff, operating with carriers less than 1MHz and with ancient valve equipment, originally. It also had virtually no built in 'intelligence'.

Your suggested amplitude-based method would be full of sources of error, due to parasitic components around the set up and unknown directivity patterns. I don't know of any examples of a system for assessing position which relies on signal strength measurement. I must say, I haven't looked in detail at your idea because it doesn't, afaik, have any precedent. If you can quote one then you might be in with a chance, though.

 

[Baluncore]

(I hope you see a second misunderstanding here.)

Do you agree with the following four statements.

1. Consider a vehicle with a coil lying in the horizontal plane, having a vertical axis, near the signal wire. The heading of the vehicle is not important. The phase and amplitude of the induced voltage will depend only on the axis position relative to the transmitter wire. When the vehicle/coil is on one side of the wire the phase will be the opposite to when on the other.

2. Consider a coil mounted on the vertical plane of the vehicle with the coil axis left/right. The induced voltage will have a phase determined by the heading of the vehicle along the wire.

3. When following the wire the left/right steering information derived from the vertical axis coil will depend on direction of travel. If the left side magnetic field is up then the right will be down. Travelling in the opposite direction will view the right side as up and the left side as down. That reversal must be handled or the vehicle will not follow the line.

4. The product of the voltage induced in the two coils performs that correction automatically.

 

[Baluncore]

I don't know of any examples of a system for assessing position which relies on signal strength measurement.

The Lorenz beam landing system was used, (with cross beams), for bombing navigation over Britain in WW2. It compared the amplitude of two signals to define a straight line that was accurate to 300m from over 100km away.

I'm trying to design a circuit (for use in a robot) that can detect if a wire carrying an AC-current is to the left or right of the detecting circuit.

The OP here specified following a wire, not deriving a 2 dimensional navigation system to rival differential GPS.

Following a wire is trivial. It is done by telecommunications engineers who need to follow and mark the position of buried cables.

 

[petterg]

You are confusing phase resolution with processor speed. Unless your robot is moving very fast (a large fraction of a wavelength each processor cycle) then the processor speed really doesn't matter. It just looks at the phase of a much lower frequency signal (beaten down version). Time difference, appears as phase of the IF frequency, which can be chosen to be anything convenient (it's a Delta t 'magnification' process' due to the frequency ratio). This is the same idea as when a sampling scope looks at extremely high frequencies and sub-samples at a manageable frequency for its circuitry. A time domain reflectometer can resolve distances of a fraction of a cm in this way, too. I think the processor speeds used to be just a few MHz in the TDRs I used in the 80s.
For the powers you are interested in (very short range operation), there is a pretty wide choice of operating frequency - the WAN frequencies would seem a good choice as the antenna would be pretty small.
BTW, you needed to know nothing extra about where you were for Decca to tell you - as long as you were on the appropriate map of the North Sea / English Channel etc. The channels of the transmitting stations would tell you that, in any case.
But Decca was real steam radio stuff, operating with carriers less than 1MHz and with ancient valve equipment, originally. It also had virtually no built in 'intelligence'.

The robot moves about 30cm/s at top speed. (At least that's what I hope. The motors haven't arrived yet.) Speed shouldn't be an issue.

When you're writing it sound like you have the perfect solution. But when I read it all seems far to complicated to implement. Where do I start? There seems to be a huge load of ways to do this. I can't even figure out which one to concentrate on. Not that I need to follow any standards.
Goal is to make the robot navigate in a area of 30x30m. An error < 1m would be very good.

Your suggested amplitude-based method would be full of sources of error

Then I rule out amplitude-based.

I must say, I haven't looked in detail at your idea because it doesn't, afaik, have any precedent. If you can quote one then you might be in with a chance, though.

What does that mean? I lack some language skills to understand the meaning. (And google translate didn't help this time.)

 

[petterg]

The OP here specified following a wire, not deriving a 2 dimensional navigation system to rival differential GPS.

A differential GPS would definately be a great solution for this problem. All though, I think it will be too complicated. Something that can make the robot know exactly where it is, is my dream.

Do you agree with the following four statements.

1. Consider a vehicle with a coil lying in the horizontal plane, having a vertical axis, near the signal wire. The heading of the vehicle is not important. The phase and amplitude of the induced voltage will depend only on the axis position relative to the transmitter wire. When the vehicle/coil is on one side of the wire the phase will be the opposite to when on the other.

2. Consider a coil mounted on the vertical plane of the vehicle with the coil axis left/right. The induced voltage will have a phase determined by the heading of the vehicle along the wire.

3. When following the wire the left/right steering information derived from the vertical axis coil will depend on direction of travel. If the left side magnetic field is up then the right will be down. Travelling in the opposite direction will view the right side as up and the left side as down. That reversal must be handled or the vehicle will not follow the line.

4. The product of the voltage induced in the two coils performs that correction automatically.

1: yes
2: yes
3: yes, for the special case when the sensor is rigth above the wire. "No" when it's a litle bit on the side.
4: yes, when #3. While #3 is not true, no.

In #1 you say " The phase and amplitude of the induced voltage will depend only on the axis position relative to the transmitter wire."
If #3 should be true for other locations than rigth above the wire, then that quoted part of #1 cannot be true. #1 is true, hence #3 can not be.

So I guess the missunderstanding here is that you think the robot will be on the wire, while my question is how to detect on which side of the wire the robot is.

Here is an example of location and the information the robot can detect with my current knowledge:
Say you have a wire going north-south, previously called the y-axis. The wire is 3mm thick. The robot is 600mm long, 400mm wide. The distancecalculation indicates the robot is between 6000mm and 8000mm from the cable. (This distance calculation is subject for disturbtion.)
The vertical axis coil indicates magnetic field going up.
The horizontal axis coil indicates a weak magnetic field towards the front of the robot.

What I'm not able to figure out: Is the robot on the east or west side of the wire?

Well, if the current in the wire is pointing to the north, the magnetic field will be up on the west side. If the current is heading south, the magnetic field will be up on the east side. So with the robots current knowledge there is two posible locations.
If the wire signal is of a kind so that the robot can identify its phase, then it would know it's location.

The best way so far, within the limits of what I think I'm able to implement seems to be the sawtooth-signal. Although I think Sophiecentaur might put me on another track.

 

[Baluncore]

So I guess the missunderstanding here is that you think the robot will be on the wire, while my question is how to detect on which side of the wire the robot is.

I thought it would be close to the wire and would be attempting to follow the wire. My suggested arrangement detects which side of the wire the sensor is on. If it is exactly on the wire the output will average zero and the question “which side” will be answered as 0 = neither.

I'm trying to design a circuit (for use in a robot) that can detect if a wire carrying an AC-current is to the left or right of the detecting circuit.

I was assuming that the vehicle would be moving parallel with or following above the wire. I saw no precision ranging requirement for the distance from the wire, certainly not of the order of 5m. A Lorenz type system could be integrated into the system later that would achieve that; http://en.wikipedia.org/wiki/Lorenz_beam it would be a different system but a similar concept.

Now let's add a minor complexity to the simple wire. When current is flowing one way a small, (10mW), VHF radio transmitter transmits a carrier. When the current flows the other way there is no carrier transmitted. This gives you an independent phase reference so a vertical axis coil can now tell which side of the line it is on.

 

[jim hardy]

Perhaps it's simpler.

How far away is the wire?
I assume it's a single conductor not a two conductor lampcord where the resultant magnetic field would be nearly zero...

Were the robot pretty well non-magnetic , two identical magnetic sensors on opposite sides of the robot would receive signals differing in strength by square of their respective distances from the wire, which would differ by width of robot.
A simple amplitude comparison might do the trick if the distances are not great compared to size of robot..
Sum the two signals and square up the result with a Schmitt trigger, use that for CLOCK input to a D flip-flop
Difference the two signals and square up the result with a Schmitt trigger, use that for DATA input to same flip-flop.

OUTPUT of flip-flop is direction L or R.

I think that's sort of like the multiplication suggested by Baluncore

but a TV show on hammerhead sharks made me think of it.

Might need to delay the summed signal by a fraction of a line cycle so that comparison is made well after sinewave's zero crossing.

A job for LM324's and 555's . (I know , I'm showing my age ).

Is this practical or am I having a midsummer night's dream ? old jim

 

[petterg]

Here's a document that gives a good overview of various ways of using wifi for navigation
http://www.snet.tu-berlin.de/filead...12/snet-project/wifi-positioning_henniges.pdf

 

[petterg]

Thank you, Jim

I think your onto something.
I need to look into the circuit you describe and see if I understand it.

It is a single conductor wire. Well it can be anything. My intetion was to use a single conductor. If it turns out that an other wire makes stuff easier I'll change it.

While being a single conductor it also has to be a loop. Hence a coil. Being a coil, the magnetic field can be made stronger when using a multi-conductor and connect the conductors in series. So maybe a single conductor is not prefered.

The robot will be at most 20m from the wire. That is 30 times the size of the robot. I guess there needs to be at lest 3 sensors in order to make the theory work while the robot is able to turn.

 

[sophiecentaur]

The robot moves about 30cm/s at top speed. (At least that's what I hope. The motors haven't arrived yet.) Speed shouldn't be an issue.

When you're writing it sound like you have the perfect solution. But when I read it all seems far to complicated to implement. Where do I start? There seems to be a huge load of ways to do this. I can't even figure out which one to concentrate on. Not that I need to follow any standards.
Goal is to make the robot navigate in a area of 30x30m. An error < 1m would be very good.

Then I rule out amplitude-based.

What does that mean? I lack some language skills to understand the meaning. (And google translate didn't help this time.)

You are right; my proposal is certainly complicated and it would do more than just to tell you which side of wire your robot is - which may not be needed.
If all you really want to know is that then the solution may be simpler - because you do not need to measure distances accurately.
The problem still remains that the robot needs to know something of its orientation. The idea of an asymmetrical waveform for the current on the wire is a good one because it gives information about the sign of the field vector. A single coil will not tell you which way to go to get to the wire (it will just tell you which side you are on) and your robot would need to 'hunt' in order to find the slope of the amplitude / distance relationship. Three coils in quadrature could tell you the direction of the magnetic field and give you a rough idea of its strength - hence the distance from the wire plus which side you are on. However, if your robot is in or near the same horizontal plane as the wire then the field will appear vertically up or vertically down, which gives no information about the direction to go to reach the wire. If the robot carries the coils high enough to give an 'angle' then - less of a problem.
Imo, there is no substitute for some form of phase /timing measurement, if you want to know your distance from the wire. A single measurement of distance from the wire, over a limited range, could be a compromise.
Take two oscillators, using any high frequency you choose, with their frequencies differing by exactly 100MHz (say) and feed them to the wire. The distance of the robot from the wire can be found by the phase difference, which would repeat every 3m. Again.there would be a non-trivial bit of circuit design (quite do-able, though).

The problem would become much more tractable is you wanted to keep the robot between a pair of parallel wires, using a phase difference approach.

The final paragraph in my other post was just saying that, if you can't find evidence of anyone else using just amplitude, then you are unlikely to make it work. I am ignoring the systems used for automatic aircraft landing because they involve a complex pair of transmitting antennae to form the right beam shape.

 

[jim hardy]

yes, the current must get back from whence it came even if by multiple paths...
Maybe you can get creative with the solid geometry of the room.
It's time for some experiments, I think.

Baluncore and Sophie are better versed in math and field theory than I, they probably are fluent in "Poynting vectors" and "intensity" .

On a low level practical note - What are your constraints?

Were the 'wire' to be a light rope made of LED's with something different about the ones pointing left and the ones pointing right, it'd become a visual sensing task .

Issue to solve: The current in the wire has to be AC.

Is that a constraint? Fluxgate or Hall magnetometer can sense a DC field, three of them can resolve its direction.

Good luck and have fun !


old jim

 

[sophiecentaur]

yes, the current must get back from whence it came even if by multiple paths...
Maybe you can get creative with the solid geometry of the room.
It's time for some experiments, I think.

Baluncore and Sophie are better versed in math and field theory than I, they probably are fluent in "Poynting vectors" and "intensity" .

On a low level practical note - What are your constraints?

Were the 'wire' to be a light rope made of LED's with something different about the ones pointing left and the ones pointing right, it'd become a visual sensing task .


Is that a constraint? Fluxgate or Hall magnetometer can sense a DC field, three of them can resolve its direction.

Good luck and have fun !


old jim

I think the message is that 'on the whole, we'd rather not start from here'. I reckon that what's called for is a modification of the initial requirements. There are so many more ways of approaching this than insisting on using a single wire. As Jim says - you also need a return current path, which is highly relevant in a practical situation. You can still have something that looks like a wire but why can't it be more complicated? Two wires and a third wire between them to act as a reflector could be arranged to fire two beams in the left and right directions but this, of course, would involve some RF and the spacing of the wires and reflector would need to be a quarter wavelength - but why not? 3cm microwave systems are available as School demo kits (Gunn diode) and would be suitable for a few experiments. Then there's the optical (infra red) beams solution, which would actually be a piece of cake.
Basically anything but the single wire.

 

[petterg]

...

The reason I started out this thread with the purpose just looking for which side of the wire the robot is on was that I already have (in theory) methodes to estimate the location. Those methodes are:
+ Estimate distance from wire based on received signal amplitude, probably 40% error.

+ On occations when none of the engines are running, get a compass reading. A compass in front and back can hopefully averange out distrbance from souroundings.

+ Count wheel rotations. That will give robots rotation and traveled distance in each direction.

+ GPS readings - look at difference with previuos reading. GPS reading can be 100m off. But it doesn't really matter if it's 100m off as long as it's still 100m off afer moving 1m. Sometimes the GPS position can jump 40m in a second. Such events has to be filtered out. I considered multiple GPS'es, but after walking around with three GPS's I realized that they tend to get the same error. However, two GPSes 60cm apart may have the effect that only one of them are located at a spot disturbed by an obsticals.

+ Every time the wire is crossed, calculate each of the aboves errors. Count for thees error when moving on.

= I'm hoping the sum of all of this will return a position that is accurate down to 50cm. It's acceptable that location is off by a few meters as a startup. This should be adjusted for when the wire is crossed. Then the presission will drift until the next time the wire is crossed.

IF I manage to see how to implement radio navigation with better pressision I can throw away all the above. Hence the logic might be simplified by doing something complex. Alternatives to location implementations are very welcome!

When I started the thread I was convinced that radio navigation would be far to complicated to implement (When I slightly understand the basic theori, I'm still FAR away from being able to make an implementation.) Also I though it would be less accurate than the sum of the above methods.

The idea of using dual frequency is also a good one. (I understand the basic theory without looking it up.)
The field will be close to vertical most of the time. Sensor at most 0.3m above horizontal wire plane, and up to 20m away. But who says there can't be several coils at different angels on the robot?

I've ordered quite a bit of components today so I can start playing in a few weeks. (Ships from Hong Kong)

 

[petterg]

..

The wire will be digged down a few cm in the ground. It won't be visible.

The constraints is that this needs to not be distrubet by the wires to the street lights. The equipment should be as discret as posible (wires across the area will block for people moving around). Digging down a wire or two is ok. Digging down a grid of wires is not ok.
And there's quite a bit of wlans in the area that could be used for positioning. And finally, it shouldn't send out signals that's disturbing the suroundings.
Other than that limitations can be adjusted.

Hall sensors (or anything that uses DC) will be distrubed by random magnetic fields.

 

[jim hardy]

I had no idea what you were up to - maybe robot wars or something.

I was even wondering if we could detect the Earth's magnetic field in the vicinity, and modulate it slowly with our loop. That'd give us robot heading information twice per cycle at zero crossings of our modulation... just daydreaming...

Sounds almost like you're taking the wireless dog fence to next stage... I wonder what's to be gleaned from taking one of those apart... it might beat reinventing the wheel.

 

[sophiecentaur]

You now say that you are considering using other positional information. That will give you a much better chance. Dead reckoning (wheel rotations plus direction) would give you some good results,
This should be a really fun project, I think.

 

[Baluncore]

There are too many possible solutions for this undefined problem. Random guesses at moving goal posts should be quite successful at preventing completion of the project.

Without a specification of the performance requirements and the operational environment it will not be possible to advise a sensible solution.

Are we permitted to know what the vehicle is for?
Is it armed and dangerous?
Is it reasonable for us to assume it is a robot vacuum cleaner or lawn mower?
What are the maximum dimensions of the field?
Is knowing position sufficient or is vehicle orientation also needed?
How accurately must position and orientation be known?
How fast does the vehicle travel?
Is it inside a building or outside?
Must it work in the dark and in bright sunlight, night, day or both?
What obstructions to visibility are there in the field?
Can there be several points in the area, three of which are always visible from the vehicle?
Will the vehicle be alone or with others, friends or enemies?

 

[sophiecentaur]

@petterg
If you want this thing to operate 20m away from your wire you will need hundreds of amps of current in order to exceed the Earth's magnetic field. (use this link to calculate it) You are forced to use RF techniques at this distance.
I can echo Baluncore's frustration at the lack of specific details you are 'leaking out', when the solution to sort of project relies totally on the values of the variables involved. Just buying a load of components is not the way forward; planning is necessary.

 

[petterg]

I'll try to answer all questions in one post.

I tried to narrow the subject for this thread down to the one particular issue. That turned out to be a mistake. Sorry about that.

The background for this project is that I bought a robot lawn mower from Bosch. It has an advantage over other similar robots in the fact that it's mapping the working area. It learns where obstacles are located, it knows where it has been working, and works systematically in the areas where it hasn't been for a long time.

That is in theory.

What Bosch has managed to do is the navigation. Some think it's using gps, but it's not. (It does have a built in gps chip, but it's not in use by the software for other than setting time/date.) It's navigating by following a wire, counting wheel rotations, and when it has moved past a couple of wire bends and counted the wheel rotations between the bends it's got a reference point. From this point it's navigating to where ever it needs to work. The navigation logic works well. Direction is also assisted from a compass - if you try to lift up one of the wheels so that it changes direction, it's instantly compensating for that by stopping the wheel still on the ground, and wait for the lifted wheel to recover grip so the course is correct. Error seems to be less than 1 degree.

Anything but the navigation logic is a total failure on this product. It's software is so full of bugs that it couldn't even be called a beta release. The mechanic is extremely poor. Even the razorblade-kind of knifes that is fitted to a spinning disc - the some samples of the product is shipped without tightening the screews holding the disc. (And such issues aren't related to a faulty batch. It occurs on a large number of samples sold i several countries, marked with different versions of the product.)
I can't mention all the faults here, I could write for days. (Really. I've probably spent more than 100 hours writing bug reports to bosch.)

I got my robot replaced by bosch service (after it had spent more time in the back of my car to/from service than it had spent on my lawn). The replacement with the newest updated firmware, fixed some bugs - and introduced just as many new ones.

Bosch seems to refuse that there are issues with their robot. Bosch is so hard to communicate with that loads of Homedepot-kind of stores stop selling them. The customers return the faulty products to the stores, but the stores only get to return a few of them to Bosch. One customer got his robot replaced 5 times in 8 weeks.

Then, 6 days after I got the replacement robot, it was stolen. (They even took the base station, the digged down wire and the neighbors garden furniture - while we were home!)

This left me with 4 choices: Not have a robot again, get a new Bosch, get a new robot with random mowing pattern, or make something better.

You might have guessed my choice is to make something better.

While commercial robots leave 10-30cm uncut grass around obstacles and edges, this new robot will have a edge-cutter function. That's why this robot will have 7 motors spinning, creating the "random" magnetic fields that makes navigation harder. It's quite important that the navigation can say that when a know obstacle is hit, it should know if this is an obstacle marked in the map, or an unknown obstacle. This will make the difference if the robot is cutting grass around a tree or the fur of a curious dog.

Most of the well known robot mowers recognizes which side of the wire they're on - and that's their only navigation. Also, if the robot is 20m from the wire and you cut the wire, it stops. As for Bosch there is a square wave of about 150mA in the wire.


I'd like the wire not to be an absolute boundary wire, as most others. I'd like to mark in the garden map that in some areas it's allowed to go outside the wire until it hits obstacles (or wire signal get below some threshold), while in other areas it's not allowed to cross the wire at all.


And, if it works, why not fit it with a leaf blower that also can be used on snow.

-----------------------

Are we permitted to know what the vehicle is for?
Yes

Is it armed and dangerous?
Yes. Hopefully within limits

Is it reasonable for us to assume it is a robot vacuum cleaner or lawn mower?
yes

What are the maximum dimensions of the field?
A square around the field will have sides of 28m. Although if the robot gets more than 20m from the wire the navigation has failed.

Is knowing position sufficient or is vehicle orientation also needed?
Both is needed. It will have reference points where it can go to calibrate it's sensors.

How accurately must position and orientation be known?
To make the wheel rotation counter work as part of the positioning the direction is required to be quite accurate. If position can be accurate with other kinds of sensors the direction doesn't have to be all that accurate.

How fast does the vehicle travel?
30cm/s is a goal.

Is it inside a building or outside?
Outside, or under the deck, or under trees.

Must it work in the dark and in bright sunlight, night, day or both?
Bright sun might introduce a risk for overheating. Other than that it should work under all kinds of light conditions.

What obstructions to visibility are there in the field?
A house, garage, trees, rocks, cars, poles, bushes, walking people. In worst case a dog or a thief. (Aim for the last one!)

Can there be several points in the area, three of which are always visible from the vehicle?
No. But there can be 8 points where at least 3 of them will be visible at all locations. Most locations will have visual to 4 of them.
There's already at least 4 wlans available at any location.

Will the vehicle be alone or with others, friends or enemies?
If it turns out well, I guess the neighbors would like to build them self a copy. If those turn out friendly or not depends on if the signals from ones area will disturb navigation for the neighboring areas.

 

[jim hardy]

original question was

Now the problem: How can we know if Xr=x1 or Xr=-x1 ?

and I'm guilty of getting distracted by the idea of a cool robot. Sorry about that.

Seems to me next step is to lay out some wire, take some measurements with coils near your wire and see what the signal looks like.

 

[sophiecentaur]

@petterg
I am still not clear whether you want a cure for the faulty device you already have, to reverse engineer it so you can find out how it works or to build another one with an entirely different system. Either way, you have managed to waste a lot of your time and PF's. The general rule on PF is that you attempt to write a clear question about a particular problem and you will invariably get some sense out of us. As it is, you have managed to get a load of idle ramblings, so far.

 

[petterg]

Yep. Components for generating signal (bunch of opamps, various conductors and resistors), an oscilloscope, and some arduino pro micros are on the way.

How do I find the ideal coil for picking up the wire signal? A large coil will give stronger signal, but might not cope with the frequency?
And for filtering the signal when the coil is the source of electric signal in the circuit; Should I consider the coil as part of the RLC-filter, or just as a source for a RC-filter?
(I can't figure how to simulate a coil as the source of signal.)

 

[sophiecentaur]

It would be normal to design a system before ordering the components. How is it going to operate? What principle and what values of current, frequency etc. etc. are you planning to use? Random assemblies of components tend to result in nothing happening - except, sometimes, burnt components.

 

[petterg]

The above post was for Jim

I am still not clear whether you want a cure for the faulty device you already have.

The two I had is gone. First one returned. Second one stolen. This will be a build from zero.

you have managed to waste a lot of your time and PF's. The general rule on PF is that you attempt to write a clear question about a particular problem and you will invariably get some sense out of us. As it is, you have managed to get a load of idle ramblings, so far.

I thought the problem was clear. Cause it was the remainding one problem I hadn't figured out to make all of the mentioned position calculations within the needs.
Allong the way someone came up with complex ideas that can potentially replace all the ideas I had for navigation in the first place.

I'm still at the side-of-wire-detector issue because the alternatives seems to complicated to me. (I understand the basic idea, but not well enough to implement it.)

 

[petterg]

It would be normal to design a system before ordering the components. How is it going to operate? What principle and what values of current, frequency etc. etc. are you planning to use? Random assemblies of components tend to result in nothing happening - except, sometimes, burnt components.

I have the sender planned. Opamps/conductors sending sawtooth at 4880Hz.
What I don't have is a plan for the receiver. It seems like I need to measure a coil around the field while the sender is sending and make the design based on those measurements.

 

[Baluncore]

As for Bosch there is a square wave of about 150mA in the wire.

What was the frequency of that square wave?

It may well be an RF system that satisfies the navigation requirements, but that is not certain. In principle, any AC modulation can operate with immunity from the Earth's magnetic field. I have built systems that use a buried wire that operate at 1kHz with a range of over 100m.

It is easy to fixate on the Hall Effect sensor as the solution to magnetic field detection problems. Unfortunately, Hall Effect sensors are not optimised for AC field detection as they are biased by the background field. A high gain Hall Effect sensor will saturate and so has a limited sensitivity when AC coupled. A pickup coil can be better optimised for small signal detection.

Most systems that measure phase have a very high conversion gain. Interference from power lines and electric motors is often overstated. It is not that much of a problem when making phase measurements because phase is not instant, it is temporal, it has the time to get it right. I suspect that a shaped pattern loop of wire is a local reference used to correct for slip / drift of the wheel navigation.

The GPS is almost certainly there to identify which lawn it is on. Since simple GPS is only accurate to about 2m it could only be used to detect extreme boundary excursion.

The wire loop does not have to be “ranged”, it is only required so the vehicle can recalibrate it's position by following the wire line around one bend.

 

[sophiecentaur]

It is easy to fixate on the Hall Effect sensor as the solution to magnetic field detection problems. Unfortunately, Hall Effect sensors are not optimised for AC field detection as they are biased by the background field. A high gain Hall Effect sensor will saturate and so has a limited sensitivity when AC coupled. A pickup coil can be better optimised for small signal detection.

I agree. afaics, there is absolutely no need for a Hall sensor. Simple induction (to a tuned coil) would work just as well. I can see a problem with the (audio) field around a wire on a random route varying considerably as the return path is pretty well undefined. Hence my suggestion for using RF and a feeder, with its own, integral return conductor. I reckon that a piece of mains twin cable would probably serve as a good feeder with lots of leakage. Matching would hardly be a problem, either.

 

[petterg]

What was the frequency of that square wave?

A pattern that looks like -_-_--__-_-_--__
Where each - or _ lasting for about 100ns. (It would look better in text if - and _ had the same length.)

It may well be an RF system that satisfies the navigation requirements, but that is not certain. In principle, any AC modulation can operate with immunity from the Earth's magnetic field. I have built systems that use a buried wire that operate at 1kHz with a range of over 100m.

Sounds promising.
What kind of receiver did you use? Sender was a sin-wave?

It is easy to fixate on the Hall Effect sensor as the solution to magnetic field detection problems. Unfortunately, Hall Effect sensors are not optimised for AC field detection as they are biased by the background field. A high gain Hall Effect sensor will saturate and so has a limited sensitivity when AC coupled. A pickup coil can be better optimised for small signal detection.

Considered the motors noise I think a coil will perform better than a Hall sensor.

Most systems that measure phase have a very high conversion gain. Interference from power lines and electric motors is often overstated. It is not that much of a problem when making phase measurements because phase is not instant, it is temporal, it has the time to get it right.

It's a matter of filtering out the signals to compare, I guess.

I suspect that a shaped pattern loop of wire is a local reference used to correct for slip / drift of the wheel navigation.

The GPS is almost certainly there to identify which lawn it is on. Since simple GPS is only accurate to about 2m it could only be used to detect extreme boundary excursion.

When the bosch stops on the lawn and wait for a human to press a button (which it does hourly on a good day, every few minutes on a bad day) there is a choice for "continue mowing" or "go home". When I took it to service 105km away, and powered it up I got the same choice. I just had to try "go home" to see what it would do. It was thinking for minutes, then concluded no signal from cable.

When I got the replacement it was tested at the service location before I brought it home. A home it started out quite well before it figured that the cable boundaries did not fit with the stored map.

A guy reported that he'd tried the bosch in a concrete parking garage/basement. He concluded that it worked better there than in the yard. Probably because the wheels had better grip. Clock remained on 00:00 during his test, so no GPS coverage.

For those experienced I conclude that GPS is only used to setting the clock.

The wire loop does not have to be “ranged”, it is only required so the vehicle can recalibrate it's position by following the wire line around one bend.

It will need two bends and compass reading to have know which straight it just passed.
Bosch is following the line so much that all grass is gone and it's tracks turn into slippery mud. Hence the wheels lose traction and the wheel count along the wire fail.

Using phase timing, maybe it's possible for the robot to just stop over the wire and figure how far it is from the sender?