30 August 2000

Mowing Robot Project. (UDO)

After extending our lawn, the idea came to me to build an automatic lawn mower. In view of the surface area of 55 m², the existing Husqvarna machine was too big.

Objectives:

- Roughly dimensioned to fit an A4 sheet.
- Should also be capable of cutting the edges properly.
- Should operate fully automatically.
- Should mow approximately 4 hours per day.
- Light weight (energy saving, in future on solar energy)
- Sturdy design (no plaything)

Well, here we go.

In the first design I used two stepping motors taken from a 5.25 inch IBM disk drive, a Basic Stamp II computer, model-making wheels with an 8cm diameter made by Conrad and a lot of luck.

7 September 2000
After doing a lot of thinking, I come to the conclusion that in connection with mowing the edges the mower should be constructed as a front mower. This is made of an A4 epoxy chassis plate (printed circuit board) provided with two drive wheels and a trailing wheel at the rear. Because the wheels are too small to mount the motors directly onto the wheels, they are mounted on top of the chassis using a chain/gear drive on a shaft rotating in ball bearings underneath the chassis (reason: chains are rather insensitive to contamination). The stepping motor control system consists of an L297 plus an L298 for both motors. The first detection is done by way of microswitches.

14 September 2000
After a lot of building and a bit of software, the prototype was ready for its first performance in the living-room.
But alas, the motors can't even manage a mere 2% grade, so I can put the garden right out of my head. The motor power is too low.
Back to the drawing board. Because I also have two more powerful stepping motors of approx. 1.2 A, as against the 150 mA of the old motors, this should work. (And imagine me thinking that these would burn rubber like hell!)
I have rebuilt the frame and I have applied a different motor controller using the UMC-4508. The new motors are of the unipolar type.

One of the objectives, light weight, is already down the drain. The motors weigh 600 grams a piece, and the machine should be provided with a more powerful, therefore heavier, battery. (But then nature is cruel, and you just want to have a go at it.)
The trial with the new motors in the living-room is going well. The machine is driving nicely along the skirting-boards of the room.
And now to the garden for the first trial run without a mowing unit. I have put off the mowing issue to be dealt with later on.
Disappointment everywhere. The thing won't budge an inch. I had even pre-mown the grass with the hand mower and said to my UDO: "Look, that's the way to do it." (We have already started talking to it.)
And guess what: Nothing doing, nada, zippo. The cause is that the grass stalks are approx. 4 cm high.
This means that the machine, with its weight of 3 kg and wheels of 8 cm, has to overcome an upward gradient of 45 degrees and that each motor has to raise 15 kg.
But what's the way to go about it then? Bigger wheels? This means that the entire construction should be rebuilt. In short, I'm going about it in the wrong way. Back again to the drawing board.

21 September 2000
I come to the conclusion: do not build just like that and work by approximation, but calculate everything and start from scratch.
The wheels should be bigger, that's for sure. So, I've ordered new wheels having a 16 cm diameter.
But should it actually be a drive based on stepping motors? These motors have a poor efficiency and are heavy.
Moreover, what's the added value of a mower that can be controlled to an accuracy of 0.125 cm?
After developing a mathematical model in Excel and inputting the old constructions, the result is what I have experienced with the previous models.
The first one has a forward thrust of 100 grams and the second model has a thrust of 800 grams. That's simply impossible! I have consulted some technical literature from Conrad, and in it I found gear motors of 180 N/cm. They use up 1.5 W and weigh 200 grams. This should do the trick, I feel.

Objective (light weight) back on the agenda, and a forward thrust of 2.5 kg per wheel. I don't know how long these motors will last, but they are not expensive.
The new construction gives good results right from start at half the supply voltage (i.e. 6 V). Fine so far, the consumption of the drive is now 200 mA.

30 September 2000
Now I'm going to start with the mowing unit. I have ordered, again with Conrad, motors having a speed of 5,600 r.p.m. and 11,000 r.p.m. Using these motors, I'm going to experiment with a cutting blade and a saw plate.
I have made a feeble attempt to calculate this but I don't know how. So, according to the "we'll just see what happens" model.
I myself have the feeling that a saw plate is going to work, because it can be balanced without a problem. But a good friend says that a cutting blade is best, because this can be bent like a fan and it does not blow away the grass but sucks the grass into the blade. He may have something there. So, this is what I'm going to try first.

Meanwhile, I've been doing endurance tests with the drive section in the hall of our house. These go smoothly and UDO manages to work its way out of every situation.

3 October 2000
Jan, a friend of mine, has access to a CNC milling machine at work and is making the frame from rigid foam PVC.
Meanwhile, the whole "infernal machine" has been input into AutoCAD, which makes life a lot easier. For, now I can put all components in their proper place before any building takes place. (Already a full box of them by now.) The frame is ready and I'm in the process of building up the machine.

The first test goes well. I've opted for 10 microswitches for the front sensor.
I still can't find another solution to make the front sensitive to touching an obstacle.

Tests performed by a Swede with an IR remote sensor came to nothing because the surface to be measured was very small.

To mount a servo on it, to look around like an eye, seems too much of a good thing to me. A rubber band combined with a press button switch is perhaps a possibility.
The microswitches are in place now and the system works well. The mowing motor is mounted near the drive shaft with a pulley drive to the shaft of the cutting blade. The reason for this is that it is not desirable to have too much mass at the front of the front mower. The mowing motor drive is made using a BUK-100. This is a power FET with a TTL input and is provided with all kinds of protective devices. Nice.

4 October 2000
At the moment I'm trying to come up with something involving an RF sensor for demarcating the garden using a buried wire. I'm going to have a try using an MK484 receiver and a transmission wire at 100 kHz. I'm also thinking of a green sensor at the front of UDO so that the mower will reverse when it is no longer green underneath the machine.
An RF wire is probably better. Such a wire can also be used in future to locate the battery charger.

UDO has been let loose in the garden fitted with a brass cutting blade. Disappointment all round, the blade immediately bends double upon touching grass. After 1 minute the blade looks like a whipped cream beater. Immediately after this, I have mounted two break-off blades of a hobby knife. Perfect, it's mowing.

10 October 2000
Lately I've measured the current through the mowing motor and it is still too high, approx. 1.5 A. I also intend to experiment with a lower supply voltage, but the grass will have to grow a bit more to be able to determine the result. The microswitches should actually be protected by a plastic flap, but various attempts ended in failure. The current plan is to make a circular front frame with the switches fitted in the centre. The advantage of this is that the switches are positioned lower, 7 cm instead of 10 cm, and the rubber front flap is easier to mount.

20 November 2000
During the past few days I've been working on the RF receiver for radio demarcation. The use of the MK484 gets stuck due to the fact that reception below 150 kHz is not possible with this chip. Now I have made a selective receiver operating at 70 kHz using an LM339 comparator. It's functioning well now on the workbench and I'm going to connect it to the processor. However, I do hope I won't be paralysing all DCF-77 time receivers in the neighbourhood. It won't come to this, I expect, because the receiver has a high sensitivity so that I won't need so much transmitting power. I have ordered another processor unit, the C-Control made by Conrad.

I get totally fed up with this clumsy instruction set provided with the Basic Stamp controller. Every software construction makes my skin crawl and causes cramp in my fingers. (That thing should be banned forever.) Tonight I'm going to find out whether the 70 kHz pulse to the output of the comparator will cause problems to the IO input of the Basic Stamp. Jan is busy building his own CNC milling machine using my old stepping motors. I hope it will be operative very soon so that it will become possible to make more complex mechanical components.

10 December 2000
Haven't had time for a short while to keep the diary, but I've made quite some progress.

Jan's CNC machine works incredibly well and the 70 kHz solution works very well and I have continued to go down this road. I even managed to convert the signal to an AD value. However, this design was dropped because of the fact that I misused a comparator as a signal converter, the result of which is that it is hard to reproduce. A really good design drops due to the complex electronics it requires. It has kept me busy for at least three weeks.

I have also experimented with a Hall sensor, but this doesn't work because I would have to send too much power through the transmission wire, probably transforming the garden into a microwave. I've also experimented with a Maxim current sensor, the MAX472, but this one, too, is too insensitive.

I did appreciate it in Maxim, though, that they sent free samples when I asked for them on the Internet. But be it as it may, we will proceed with the 70kHz comparator.

20 December 2000
Jan's milling machine is working beyond expectation and Jan has milled a front subframe with it, in which the front switches are now located in enclosed cells. The rubber flap is disappointing, it presses down the switches merely by its own weight. At a later stage, I'm going to build a latex bumper. This is available in liquid form and can be poured in a mould.

The mowing motor is now directly in front of the wheel shaft by means of a geared belt. The cutting blade has been replaced by a disk fitted with three surgical lancets fitted to the end. This solution was opted for because these knives are of equal length, and also light and rustproof.

2 January 2001
The base frame is meanwhile made of aluminium with PVC wedged in between. The original frame was made of rigid foam PVC, but this material tended to sag a lot.

6 January 2001
I just might be on to a little patent. I have been racking my brain lately about the Hall sensor because this is an elegant piece of semiconductor technique when compared to all this coil stuff. But it turns out quite difficult to measure a current through a conductor: the most sensitive sensor gets stuck at 5 A for a 10cm range. How do I solve this, without being stuck with a steaming garden?

20 January 2001
I got the idea to feed 30 A through a wire while applying very short pulses at a 1% duty cycle so that I will be able to measure a distance of 60 cm. I'm going to provide the receiver with a monoflop in order to obtain a constant output signal. The 30 A poses no problem with a large-capacity elco as a buffer. The theoretically consumed current will be 0.3 A now, and this is something we can work with.
Meanwhile, the PCBs for the 70 kHz receivers are being made in the milling machine.

21 March 2001
I have built up UDO using the new PCBs and it looks fine. However, I did have some problems with the receivers, these are too sensitive and the 77 kHz has also been a disappointing experience. In favourable reception conditions, UDO even received the atomic time signal in Germany. This was solved quickly by connecting a 10nF capacitor parallel to the receiving coil, it is now operating at 50 kHz.
Only reception remains too sensitive, and because the coils are at an angle of 45 degrees, and add to this the fact that they receive on both sides, UDO is all round sensitive to reception, which sometimes 'confuses' the software, resulting in UDO leaving the ring. The solution is simple, as good solutions always are: Aluminium foil around the rear half of the coils.

So far, UDO has been running for 5 consecutive hours without a hitch. The experiment with the Hall sensors comes to nothing, I can't get the sensitivity optimal. This is put on ice, but I will have another try later on.

I have designed an internet site for UDO and Jan's CNC milling machine. I've also applied for and obtained a domain, its name is http://www.mowbot.org.

I have also come up with a new idea to enable UDO to find the charging station. The idea underlying this is: Let UDO ride round the garden at right angles to the transmission wire and put the charging station somewhere on the wire. When the machine senses the wire, it will ride inwards. Two birds with one stone: The edges have been mown and the charging station has been located. But I'm still having difficulties with the software, i.e. to keep UDO at right angles to the wire while managing a sharp corner.

27 March 2001
The concept of tracking at right angles came to nothing. It was difficult to determine the position in the corners by using two sensors. It probably might have worked with a box brimming over with software. Now UDO follows the wire in the longitudinal direction and that goes perfectly well, plus the software is compact and simple.

In the meantime, I have designed the base station called "UDO's Home (Udopia)" in AutoCAD. It consists of a small shed in the shape of a garage fitted with a mechanical guidance system to guide it inwards (catch). At the end of this catch are two contacts to recharge the battery plus a switch so that UDO knows that it should stop. So, now we can put Jan's CNC machine to work again to realize the whole design work.

28 April 2001
The bumper with the switches is still a thorn in my flesh. Now I'm considering fitting a kind of hose along the front, with an air pressure sensor at the end. I have already ordered a pressure sensor. There are also some problems with starting the mowing motor, it causes such an induction pulse on the power supply that the CPU resets. In addition, I'm not making progress as to how I can regulate the speed from the software. The M-Basic unit has only two PWM outputs and these are now used for the drive motors. I have tried to let the BEEP output act as a PWM regulator, but that ended in failure. I'm going to the German M-Basic forum shortly, I'll ask round whether someone has successfully managed to overcome this problem.

1 May 2001
The pressure sensor at the front is operating perfectly well, except for one slight problem, namely that the hose that is being used is a bit too rigid. Therefore, I'm now busy looking for a thin-walled silicone hose. I feel this is going to work.

6 May 2001
UDO has meanwhile been mowing the lawn successfully for 3 hours and there the real shortcomings came to light.

1 )The mower is still a bit too small: it is still tipping to the sides too much during travelling across uneven surfaces.
2) The wheels should have a larger diameter and they can be considerably narrower: The mower is travelling too much on top of the grass and not on the ground underneath.
3) The motors should become slightly more powerful.

I intend to solve these points by scaling up the construction about one and a half times and to step up the power supply to 12 V. (It is currently 7.2 V.)

17 May 2001 UDO 4 IS BORN!!

Though it may still have some shortcomings, UDO 3 has meanwhile been mowing the garden to my satisfaction for 20 hours.
For a prototype, it is operating satisfactorily. And even more importantly, I have been able to leave the conventional hand mower in the shed.
I have started to work on UDO 4 and it has now been input in AutoCad in its entirety, and the wheels have already been milled out.
These are the new specifications:

1) Wheels: 250 mm in diameter
2) Cutting blade: 300 mm in diameter
3) Overall width: 300 mm
4) Overall length: 510 mm
5) Power supply: 12V 8Ah
6) Maximum cutting height: 65 mm

When reviewing UDO 3, it struck me that this version had too small a wheel base and too short a wheel base between the drive wheels and the trailing wheel. Especially when the trailing wheel was turning while reversing. The wheel would then tip forward/backward too far.
The motors were too weak when rotating round the shaft of UDO. As a result, the software had insufficient influence on the angle of rotation. This caused problems in the control software. The software thought that the wheel had made a 45 degree turn, but this angle would in reality be only 5 degrees.

The drive wheels were too small.
As a result, the motor shaft dragged through the grass.

The drive wheels were too broad.
As a result, they were travelling too much on top of the grass and not on the ground underneath.

The trailing wheel was far too small and too broad.
It dragged through the grass and sometimes, even worse, it rode across the grass.

The mowing disk was located underneath the blades and had too large a surface.
This caused too much friction on the mowing disk with already mown
grass.

Yes, now we had again enough to deal with. The first problems occurred with the wheels, because the ones which had to meet my requirements, were not available on the market. But here Jan came to my rescue again. He commented: "Then we'll just have to make them ourselves, won't we?" No sooner said than done. I have put them in AutoCAD according to the requirements. This resulted in drive wheels having a diameter of 250 mm and a breadth of 12 mm. A trailing wheel having a diameter of 125 mm and a breadth of 10 mm.
The wheels now consist of stacked plastic disks provided with notches at the outside in order to create treads. The disk for the cutting blade is now a kind of tripod with the blades fitted at the ends.
Jan is now in the process of milling the frame. This frame consists of two plates which enclose the drive motors like a kind of box. This was now possible because of the higher wheel shaft. Now the entire mowing drive assembly including the RF sensors can be placed in the box. The total underside is now flat. Besides, the collision sensor can now be mounted in a lower position. The fact that the underside is flat yields a lot of advantages because the underside of UDO 3 can hardly been cleaned after its mowing operations. (It looks like a beast.)

Now I'm experimenting with other drive motors which have a lower speed. Because of the increased wheel diameter, this is really necessary, because even UDO 3 was travelling a bit too fast.

8 June 2001
UDO 4 has been milled by now and it looks neat, the wheels and the trailing wheel are also ready. The trailing wheel runs entirely in ball bearings (neat solution). At this moment I'm assembling UDO and the box construction is perfect: All electronics are now inside, even the pulley rotates on the inside. The motors I need are out of stock temporarily. So, I'm going to solve this for the purpose of the tests by reducing the voltage to the motors to some extent using the PWM regulator. I hope that the torque will stay high enough and that UDO will not run too fast without load. Now that only the battery is on top of UDO, it can be constructed near-waterproof. In addition, I have made a device allowing for the software to be downloaded via a connector without opening the machine.

15 June 2001
The drive motors are not performing satisfactorily. As could be expected, the torque is too low, therefore it will go like a rocket when riding without load and it will stop when the going gets tough.
The motors I'm looking for should make about 8 revolutions per minute, but the motors I could find that meet these specifications were too expensive for hobby purposes. Now I have also found motors running 17 revolutions per minute which are affordable and I have ordered these.
UDO 4 has meanwhile been subjected to some trial runs in the garden. The location of the RF sensors caused problems and there was quite some interference on the RF receivers, originating from the PWM regulator. These are the same problems I experienced when developing UDO 3. But I was able to solve them all, so that doesn't worry me much.

30th June 2001
The 17rpm motors have arrived and UDO 4 is back on its wheels, complete and ready for a new trial run. The motor failures on the receivers posed a serious problem because the CPU print was situated too close to the receiver. To overcome this, I have placed the receivers to the front and the CPU print to the back so that the feed wires to the drive motors have a minimum length. The receivers are now causing more problems than they did before because they were taken from UDO 3 and have become very dirty due to the grass flying about. Before long, I will have to build new ones because the receiving sensitivity of both is going to differ considerably. For the time being, I will solve this by screening the most sensitive one to some extent with aluminium foil. Now that UDO 4 has been fully sealed, I'm not going to let this happen again with reference to the new receivers.
In addition, I placed the reception aerials a bit further back just in front of the wheels. I have done so because UDO 3 while following the transmission wire rounded the corner a little too early when approaching a square outside bend. But this solution came to nothing because it was difficult for the software to determine where the UDO's front or rear side was. The aerials have therefore been mounted back again at the front at an angle of 45 degrees. The mowing disk, too, keeps causing problems. I have to put too much energy in it to obtain a satisfactory mowing result (one motor even burnt out).
Now I am devising something with a disk, underneath which the cutting blades are situated approximately 5 cm lower by using distance bushes. However, I only hope that the disk will not get out of balance too much. The 17rpm motors are performing very well. Should UDO 4 become a success, I'm going to order the 8rpm motors. The mower is still driving a trifle too fast. But now UDO 4 is mowing satisfactorily with UDO 3's smaller mowing disk mounted, and the bumper with integrated air pressure sensor has worked out really well. Next, we are going to conduct a 20-hour endurance test and then we are going to work on the charging station plus some software.

20th July 2001
UDO has stopped after having mowed for 15 hours. One of the driving motors has broken down. On closer examination, it turned out that the reduction gears had worn out completely. It also came to light at this stage that the supplier had exaggerated the specifications of these motors by a factor 10. Currently, I have not been able to find any affordable motors yet which meet the requirements. But there's Jan again, who remarks: "So far, we have made almost everything ourselves. Why don't we build our own reducing gear unit?" Well, in AutoCad I have been experimenting with gear wheels but that became too complex. However, a toothed-belt drive seems to be the best choice because it relieves the motors from the jolts which occur when stopping and when changing direction of rotation. One thing and another has resulted in a compromise, namely a gear motor with a belt transmission. We are going to mill it out on Jan's machine.

15th August 2001
A real success this drive unit is. According to calculations, it is capable of carrying 80 kg and the unit loads the gear motor only with half the capacity. In addition, it weighs less than 200 grams without motor. UDO re-assembled, and there we go again for a new run. The cutting blade still constitutes a problem. It requires too much energy to let the 280mm long blade rotate. This turns out to be caused by the fact that the freshly mown grass is thrown against the blunt sides of the blade in the centre of the disk. It also turns out that the lighter the mowing disk the better the results. At first I thought that if the disk has a high mass, it would keep turning when the going gets tough for a moment. Nothing turned out to be further from the truth, for the blade keeps cutting through the grass continuously. The grass that has been mown is constantly being replaced due to the driving action. Therefore, the mower has to cut continuously through a cake of grass. Now I am busy trying out various mowing disks.

20th September 2001
UDO is still performing properly. At present, the mowing disk consists of a 250mm ABS disk which is 4mm thick and has three Stanley knives fitted at an angle of 5 degrees. This disk has been milled off around the blades to a thickness of 0.5 mm to obtain a weight of 100 grams. The underside of this disk is completely smooth so that it meets with little resistance from the mown grass. The mowing motor is still a little too weak. At present, it is approx. 5 watts. I am going to order a 19-watt motor and see whether this is enough. At the same time I am examining the possibility to mount a stationary disk underneath the mowing disk. In so doing, the mowing disk will be subjected to even less resistance from the mown grass.

4th October 2001
So far the mower has been performing trouble-free for some 20 hours. We are well under way to 60 hours. It is planned to construct UDO 4 again in its entirety during the coming winter, the final version for next summer. Meanwhile, however, I keep experimenting with various mowing disks and mowing motors. Just now I realized that only twice have I had to mow the grass by myself and that UDO has done the rest. So, it is really going to work. Now I have also the opportunity to improve the software.
 
20th January 2002
Have concerned myself far too little with UDO. This is caused by the fact that I have got a new job which demands all my attention. Jan is pushing me on by saying that the grass will be growing again in March.
The new design has to undergo two modifications. First, the wire receivers, these are difficult to reproduce without sufficient knowledge of electronics and that is not intended, because everyone should be able to build the mower by him/herself. I have got the idea to use a DCF-77 receiver chip as a receiver. These chips are crystal-controlled and do not have any adjustment points. I have been able to find a chip which fits our purpose nicely, namely the U4224B. This is a DCF receiver chip for 40, 60 and 77.5 kHz. On the outside of the chip are the AGC (Automatic Gain Control) components located. The idea is to use this AGC as a signal intensity output. The second point to be addressed is the energy flowing to the mowing motor. This will be effected by a homebrew PWM controller with two TTL chips and a Power FET.

3rd February 2002
Jan has milled the PCB and the first tests have been conducted. This is going to work, we feel, the recipient is now crystal-controlled at 60 kHz to avoid possible interference caused by the atomic time signal and at the same time I can use the AGC signal to transfer the acknowledgement signal on to the CPU. No doubt, I will have to incorporate an OPAMP because this signal may not be loaded.

23rd February 2002
It didn't work out with the DCF recipient, after all, the load of the AGC controller had to be so low that as much as a greasy finger on the print already caused problems. Therefore, this was useless with a view to people constructing it themselves.
I found two people on the Internet who had built an RF recipient for their robot. I have sent them an E-mail asking them how they had dealt with it. One person wrote back that he had converted a dog deterring device by connecting a neon light to the deterring device which activated a photocell. The other person had developed something that was quite similar to my receiver.
This answer set me thinking. My receiver wasn't so bad after all, but it needed further development.
A bout of hard thinking resulted in the following idea. I'm going to build a PLL receiver by using the NE567. This way, I'll kill two birds with one stone. Firstly, it is not wide-banded anymore so that I will hopefully get less interference from the PWM controllers and secondly, it can be adjusted very well now.
First, I'm going to try to connect the receiving coil directly to the chip and if it should turn out that it is too insensitive, I'm going to build a small amplifier for it. To be continued.
 
17th March 2002
The receiver with the NE567 is working well. Initially, I built it without using a preamplifier, but this was too insensitive. After that, I placed an amplifier with an LM393 downstream of it. Bingo!
It is working very well now, with only a small problem to be solved, namely that when the aerial coil is too close to the receiver print, it oscillates. This should be easy to solve, I think, by shielding the print by means of some tin plate.
The CPU print has been drawn again and will be tested shortly. The modifications to the print are:
Thicker print conductors for the power supply and the motor drive.
These were too thin on the old printed circuit board.
Now it has been constructed in such a manner that it can be easily mounted in UDO.
The leads can now be connected by means of plugs.
Some glue electronics have been slightly adjusted; as a result, it is easier to control the LEDs.

And I have developed my first wire transmitter so that the function generator is no longer necessary.
This is a crystal-controlled 62.5kHz transmitter with a CD4060. The first tests are promising. The only problem I'm faced with is how to control the power. Currently, this is done by a potentiometer in the power supply, this is not exactly a textbook solution but, what the heck, it works.

I have not been sitting still, as you may have noticed, but neither has the grass. Speedy action is required.

7th April 2002
The receiver, transmitter and the CPU have been put on the print and in a diagram.
The prints are being milled at Jan's. The tests with the receiver were satisfactory to good.
First, I tried to mount the aerial coil onto the receiver print, but that didn't work at all. The oscillator of the NE567 was picked up by the aerial coil so that it always detected something.
Now the receiver print has been 'tinned' and the aerial coils have been placed outside the receivers using a short coax cable.

15th April 2002
The CPU print is working perfectly, there is not even a single fault to be found on the print. Even the new features such as the programming adapter and LED control were operating properly, the first time. The advantage of the integrated programming adapter is that UDO now possesses a real RS232 port. I have used two pins of the 9-pin RS232 port to set the M-Unit to programming and run mode by enabling the M-Unit jumper through a connection in the RS232 plug.
Everything has been mounted in the old frame now to subject the new electronic components to a trial run. This went completely wrong. When near the wire, UDO turned into any random direction every time. It took me a day to find out the cause of this phenomenon. But having whiled away some time under the shower, I found the solution (You see, I tend to come up with the best ideas when I'm having a shower. Apart from that, I'm quite all right.)
This problem was caused by the fact that the receiving coils were used while being too far from their original 77.5kHz resonance frequency. I used them at 62.5kHz. As a result, the connection coax was just as good an aerial as the coil itself. By connecting an extra capacitor across the coil, I managed to tune it exactly to 62.5 kHz. I did this tuning by connecting the receiving coil to a scope. The aerial transmission wire has been made by connecting a measuring flex to a function generator. By turning the function generator up and down around the 62.5 kHz, the scope shows you exactly where the reception of the aerial is maximal.
This was the solution, because UDO is now running even better than last year and it has already logged 8 hours behind the wheels, without as much as a single error.
The receivers have now become so sensitive that I will have to modify the mini transmitter a little so that it will send out a weaker signal.
The CPU print has now also been provided with a frequency measuring input. I'm going to use this to measure the mowing motor speed by means of an optocoupler on the shaft of the motor. I will use this measurement to let UDO travel more slowly when the speed drops too far. The first tests on the working bench went smoothly.
The electronics are now working fine with the advantage compared to last year that the receivers can now be adjusted smoothly according to a fixed adjustment procedure. In addition, they are now insensitive to interference from external sources (CPU and PWM controllers).
Now the time has come to design the mechanics. This will not differ very much from the previous version. It is really necessary to have a new frame because the current one has become a holey cheese as a result of all modifications.

5th May 2002
UDO's frame has been milled again by Jan and we have fitted a heavier mowing motor. We also managed to measure the rotational speed of the cutting blade. This has not been done by using an optocoupler but by using a Hall sensor plus a  small magnet on the shaft. ("HI-TECH", nothing is foreign to UDO). Should the speed drop below a set value now, UDO will retreat and then go onto the attack again. The new mowing motor is also a success and fits nicely in the frame without any protruding parts.
Jan has milled new wheels as well, first toothed ones but that was too rocky. Now the wheels are entirely round, having staggered teeth.
The mini transmitter is working perfectly now and because of the fact that it is crystal-controlled, I do not need to adjust anything to it anymore.
UDO has logged another 20 hours of mowing, with one problem still to overcome, namely that the pressure sensor is set slightly too insensitive when bumping sideways against the tree in our garden. I'm not particularly worried about it because this can be adjusted when it is being serviced for the first time.
Time to start making new plans. And there are plenty, the most important being that UDO should be able to find its battery charger automatically. At present, I am examining the possibilty to charge UDO using an inductive coupling instead of electric contacts.
The advantage of this method is that UDO can remain waterproof and that no poor contact can develop due to oxidation. And a few things just for fun. UDO will be provided with a plexiglass bottom with built-in LEDs. At present, the front has already been fitted with LEDs which represent the sensors. A colleague of mine had a good idea. Dick suggested that UDO should have a state of mind. I'm going to try and achieve this by applying a stress factor within the software which will result - on the basis of the rotational speed of the cutting blade and the number of rotary movements within a set time - in a stress value. So, when the speed is low and UDO has to do a lot of turning within a short time, UDO will change colour from green to red. Don't you worry, we'll bring UDO to life, as long as it doesn't take to drink, because we're only too happy to down a few ourselves.