Radio/magnetic field: detect direction of antenna

[petterg]

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.

Could you elaborate that, please?

 

[Baluncore]

Because the wire is near the surface it can be followed closely by the vehicle. The wire can be laid as a large irregular closed loop which defines the return current path in a useful way. There will be a very slight error in the sensed position of the wire due to the returning segment's far fields. That difference between the position of the physical wire and the virtual null will disappear in the calibration process.

 

[sophiecentaur]

Could you elaborate that, please?

This suggestion would involve feeding the rf along the two conductors (signal and earth), which acts as a transmission line laid out in a pattern just below the surface. Some of the signal (say 1%) gets to the far end and is absorbed in a load and there will be a loss of signal along the length of wire and the fields will be very local to the cable where the signal leaks out along the whole length of the cable. The best form of transmission line would depend upon the frequency you choose. You would need to experiment with a range of types - twisted telephone cable, mains cable, low quality TV feeder with a very open weave screen, etc.. Whatever you can get samples of. You only need a low level of signal in the vicinity of the wire. Some simple compass control could take care of the situation where the robot loses the cable and it would just track at right angles until it hits the cable and RF again. You could either get the robot to follow the cable , straddling the cable or use the cable as a 'fence' to limit where the robot goes. There are a number of algorithms to achieve this, using an array of, say four receivers under the mower to keep it tracking on the line. It's been done many times, optically, with 'line-following' robots on factory floors etc. and it would be straightforward to adapt for following a transmission line, leaking RF.

I must say, I don't have a clue why the robot needs GPS to tell it which garden it happens to be in - or indeed to tell it anything except, possibly what direction it's heading in.

The choice of line design is fairly critical. The loss per metre of the run needs to be low enough for the signal to survive to the end so the robot can still detect it. The receiver needs a dynamic range that's enough to cope with the high level signal at the start and the lowest level at the other end - but you could use a number of cable runs in parallel, perhaps the stripes could all be fed in parallel at one end in a 'star connection' (not a star shape - just a star topology)

 

[Baluncore]

Leaky feeder systems are no real problem these days, and the RF receiver chips with an RSSI output as used for FM reception are very common. The problem here is that they are optimised for extending communication systems in tunnels, not for navigation systems.

Leaky feeders are operated in air, not under a variable density layer of water. One problem is that the direction information sensed in a dry soil with a forward wave coupler, may become a backward wave coupler when the soil is wet. The dominance of Odd–Even coupling modes under different soil conditions would need to be modeled before burying the cable and waiting a couple of seasons for it to get lost. Replacing the feeder would temporarily fix the problem due to airing the soil. But then it would fail again.

As I understand it sophiecentaur is advocating the use of a leaky feeder because it has no return path while I am advocating a lower frequency loop with a usable return path.

I must say, I don't have a clue why the robot needs GPS to tell it which garden it happens to be in - or indeed to tell it anything except, possibly what direction it's heading in.

When the vehicle is used on several different lawns at a few different locations it would know which internal lawn map to use without needing to be reprogrammed when it was moved. It could be owned and operated by a contractor to service many customers. If it had some clever mobile phone ability it could let it's operator know it's immediate status and inform it's manufacturer where it was now. That anti-theft technique would be very useful.

 

[petterg]

When the vehicle is used on several different lawns at a few different locations it would know which internal lawn map to use without needing to be reprogrammed when it was moved. It could be owned and operated by a contractor to service many customers. If it had some clever mobile phone ability it could let it's operator know it's immediate status and inform it's manufacturer where it was now. That anti-theft technique would be very useful.

Considered that all the cheap robots are recommended to run 24 hours / day, only interrupted by charging batteries, to handle the given area I don't think moving it to other yards are very practical. Basically they mow so much that you can't use the garden for anything but watching the robot. (One was even recommended to run 27 hours/day.)
Bosch navigation keeps the mower within some smaller areas until (it believes) it's finished with the area. Then at least you can use parts of the yard safely for a couple of hours.

Using a robot in multiple gardens are unpractical for most robots, unless you have one dimensioned for a significantly larger yard than the sum of those you'll use it in. It's also impossible for Bosch, because it can only save one yard map.
So, that's surely not the reason of having a gps in it.

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

I'm starting to think of a cable loop around the area + a RF transmission line in the center. Does the transmission line need to be straight? (There's a house in the center.)

Sounds like some kind of automatic calibration will be required for RF. It shouldn't be too hard; follow the line end to end, count wheel rotations and record the strength.

 

[Baluncore]

The observation that the Bosch is not now capable of multiple maps, cannot preclude the possibility that the next version will have that capability. Indeed I would be very surprised if the original specification did not include a multiple maps option. That is why it was designed with an internal GPS.

Maybe the multiple map option has been disabled in the Bosch v1.0 release because it will be used to encourage users to buy v2.0

 

[the_emi_guy]

I may be wrong, but my impression is that these perimeter wire sensors do not, by themselves, make any distance measurements or detect which side of the wire we are on. They are very short range and simply tell the robot when we are on top of the wire. The robot interprets this as a hard boundary and turns around. The scheme assumes the mower is started inside the perimeter wire boundary.

In other words if you start the mower outside of your perimeter, won't the mower happily mow the entire neighborhood with the exception of your yard.

 

[sophiecentaur]

It really would help to know whether this thread is discussing how this system is being achieved by Bosch or how to make a better one. If the latter, then it is necessary to state exactly what you want it to do. A specification is needed if you want to avoid even more wasted effort here. Inanimate circuit elements need to be told exactly what to do before you will have a working machine and design can only result from a properly stated requirement. What are your actual requirements for this, petter? You keep shifting the goal posts and randomly picking out ideas and systems.
Are you prepared for a network of buried cables? Do you want just cables around the perimeter? Is this machine just for your garden or do you want to create a design that can be used commercially for many different gardens? Will all these gardens have their own cable systems? If not, then why are we discussing a system with buried cables? Without them, you will need a map of every garden to be treated.
Please give us something concrete to address. (Actually, some green concrete could solve the problem of grass cutting for ever!)

 

[petterg]

Maybe the multiple map option has been disabled in the Bosch v1.0 release because it will be used to encourage users to buy v2.0

I think they intended to use gps for navigation, but failed to make the error within limits. So they left it there for possible use in later firmware.
Firmware update can be performed by inserting a bootable SD-card with software image on. Just that they refuse to make the updates available. Noone who have had a buggy v1 will put money in a v2 unless they get the money back for the v1.

 

[petterg]

I may be wrong, but my impression is that these perimeter wire sensors do not, by themselves, make any distance measurements or detect which side of the wire we are on. They are very short range and simply tell the robot when we are on top of the wire. The robot interprets this as a hard boundary and turns around. The scheme assumes the mower is started inside the perimeter wire boundary.

In other words if you start the mower outside of your perimeter, won't the mower happily mow the entire neighborhood with the exception of your yard.

That totally depends on which model you're talking about. What you say is true for the cheap brands, but not for the expensive ones.

 

[sophiecentaur]

So do you want to make one? If so, what specification are you aiming at? Time to get down to brass tacks, I think. If your not careful it will consist of 'perhapsatronic' circuitry and shave the pile of your living room carpet.

 

[petterg]

It really would help to know whether this thread is discussing how this system is being achieved by Bosch or how to make a better one. If the latter, then it is necessary to state exactly what you want it to do. A specification is needed if you want to avoid even more wasted effort here. Inanimate circuit elements need to be told exactly what to do before you will have a working machine and design can only result from a properly stated requirement. What are your actual requirements for this, petter? You keep shifting the goal posts and randomly picking out ideas and systems.
Are you prepared for a network of buried cables? Do you want just cables around the perimeter? Is this machine just for your garden or do you want to create a design that can be used commercially for many different gardens? Will all these gardens have their own cable systems? If not, then why are we discussing a system with buried cables? Without them, you will need a map of every garden to be treated.
Please give us something concrete to address. (Actually, some green concrete could solve the problem of grass cutting for ever!)

(I got interrupted by work, so the reply had to wait)

The way I work with projects is to look at what have been done with similar applications before - what have worked, what have not. That gives a lot of ideas for solutions. Some of which may conflict. Currently there seems to be three possible directions to go:
1) The border-loop-wire
2) The RF transmission line
3) The radio

Some ppl in the thread talk about one or two of these. Others talk of only one. The thread may seem a bit chaotic as subject is changing between those technologies. Maybe the technologies should be combined in the final system, or maybe we can exclude some?

Spec (ideally):
- Be able to calculate position (within 30cm error) and direction heading (within 3 degrees) for any location of the robot within an area of 20m from the cable, or a square of size 30x30m

- calculation of one of the two (position or direction) should be within limits even when obstacles like a house, a car, a person, are located between signal sender and receiver

- robot needs to know how accurate calculations a any given moment are. That way it can go for calibration and/or put a higher trust other methods like counting wheel rotations.

- total length of wire to dig down should not be more that 150m

- if system requires radio senders placed around or within the area, cables to these are included in the 150m limit.

- ground is not flat. There has to be taken in account that ground plane varies with +/- 50cm. robot height is max 40cm. Sensors can be placed high or low on the robot.

- there are 4 wlan senders around the area and 2 inside the area. At least 3 senders has coverage at any location. They can be used, but not modified.

- System should work independent on weather conditions and light conditions. Components that are not suitable for outdoors use need to work even when put in a box.

- System should not make sounds that can be heard by humans, or lights visible to humans.

- System should not disturb other electrical systems. Including an equal system installed next door.

If it works out in the end, I guess this may turn into some kind of DIY-kit-list that can be distributed.

(Green concrete solves the grass cutting, but not leaf blowing or snow blowing.)

 

[sophiecentaur]

Leaky feeder systems are no real problem these days, and the RF receiver chips with an RSSI output as used for FM reception are very common. The problem here is that they are optimised for extending communication systems in tunnels, not for navigation systems.

Leaky feeders are operated in air, not under a variable density layer of water. One problem is that the direction information sensed in a dry soil with a forward wave coupler, may become a backward wave coupler when the soil is wet. The dominance of Odd–Even coupling modes under different soil conditions would need to be modeled before burying the cable and waiting a couple of seasons for it to get lost. Replacing the feeder would temporarily fix the problem due to airing the soil. But then it would fail again.

As I understand it sophiecentaur is advocating the use of a leaky feeder because it has no return path while I am advocating a lower frequency loop with a usable return path.

I'm only proposing the leaky feeder to be operated in very near field - not necessarily for direction finding - so variation of field along the feeder would not matter. It would be a relatively short run and not necessarily involve any modulation (perhaps an audio tone, to make the detection more rugged). Any standing wave, due to mismatches / reflections, would not be relevant in this context. It would be a matter of locating a maximum as the robot crosses the line and then sticking to the line of maximum field strength, just above the line.
I would not imagine that the stuff that's used for communication in mines would be a reasonable price (high spec, for sure) and the requirement is for a usable field strength a a distance of several metres at least in mines; in this case, you'd be talking in terms of 0.5m maximum and a minimum carrier to noise ratio of no more than 20dB. Trivial requirement, I sure and there's bound to be a very cheap solution (sticking my neck out but with some certainty about it). I would even think that twisted bell wire would be a suitable candidate.
The only downside would be that the cable runs would need to be installed initially - but a simple mole plough works fine for drainage of lawns and the routing is fairly non-critical (just how perfect do your lawn stripes have to be?

 

[the_emi_guy]

That totally depends on which model you're talking about. What you say is true for the cheap brands, but not for the expensive ones.

Just curious how you know this.

 

[petterg]

I'm only proposing the leaky feeder to be operated in very near field - not necessarily for direction finding - so variation of field along the feeder would not matter. It would be a relatively short run and not necessarily involve any modulation (perhaps an audio tone, to make the detection more rugged). Any standing wave, due to mismatches / reflections, would not be relevant in this context. It would be a matter of locating a maximum as the robot crosses the line and then sticking to the line of maximum field strength, just above the line.
I would not imagine that the stuff that's used for communication in mines would be a reasonable price (high spec, for sure) and the requirement is for a usable field strength a a distance of several metres at least in mines; in this case, you'd be talking in terms of 0.5m maximum and a minimum carrier to noise ratio of no more than 20dB. Trivial requirement, I sure and there's bound to be a very cheap solution (sticking my neck out but with some certainty about it). I would even think that twisted bell wire would be a suitable candidate.
The only downside would be that the cable runs would need to be installed initially - but a simple mole plough works fine for drainage of lawns and the routing is fairly non-critical (just how perfect do your lawn stripes have to be?

How would I start creating a test rig for this?
Is the sender just a regular sin-wave generator?
Transmission is done be grounding one pole, and connect the other to a wire (experiment with several types of wire). For testing I suppose it's ok to not dig the wire down. Should the far end of wire have some kind of terminator? Should it be grounded in the far end?

Receiver can use any kind of coil for sensing the signal?
How do I select the best coil for the purpose? When designing a filter at the receiver, do I consider the pickup coil as a part of the filter or as a signalsource?

 

[petterg]

Just curious how you know this.

By communicating with lots of robot owners. "Bwt, what happens with your robot if the power goes out while it's mowing? Does it keep going across the wire for greener grass other places in the neighborhood?"
Those who tested to pull the power while the robot was mowing could report that:
Bosch, Husquarna, Gardena and Robomow stops within a couple of seconds.
Worx and "the china robots" (those who sell under various names and colors depending on which store you get them from, but are essentially just 3 different models) keeps going random directions until battery runs out.
All though, the best story so far came from a owner of a bosch; He had his mower in the garage over night. When he got out next morning there was an other bosch docked in his charger. I don't think that was a power out though. Probably just another example of when the software tell the robot to turn the wrong way at the cable.