Locking onto the row. OK so we have now seeded our crop in rows that are exactly one foot apart and in the rows the plants will be spaced evenly with 2 inch spacing. As soon as the seeds germinate and have small green shoots breaking through the surface of the soil, we can use these plants to locate and position vehicles and implements for subsequent operations such as weeding.
Using the analogy of a garden, with plants seeded in straight rows; as the plants germinate and surface, the row can be recognized. Anything that is green, and not in the row is, by default, a weed; it can and should be removed. The location of the rows also provides some guidance as where we can walk – we don’t want to stomp all over the good plants. Likewise, with the weeder; it is best to keep the wheels between the rows and off the good plants by locking onto the relative position of the rows to the wheels.
How is this locking process accomplished? With cameras. The implement will have a camera pointed straight down for where it would expect the row of crop to be. On the implement would be a red pointer or red thread, that indicates the implements, row center. When the camera has the red pointer in view, it indicates where in the camera’s field of view, the row should be: and therefore, the system can do a software calibration or row alignment without having to fine tune the camera’s mounting.
The frame of reference is our implement, and a red thread could be placed on our machine row center, and a blue thread could be the horizontal axis of our machine coordinate. The camera pointed straight down would see a red line, (vertical or y axis) and a blue line, (horizontal or x axis). This x,y coordinate system would have green pixels, and each glob of green pixels could be given a x,y coordinate. These coordinates could in turn be used to determine the row of good plants, and the offset of the machine-row to the actual row, and to position the weeds precisely relative to the implement. This would be necessary to remove individual weeds.
To detect the row of good plants, a histogram of all the green pixels in that vertical line of the cameras frame of view, would indicate where the dominant growth of plants would be – presumably the row of good plants. The horizontal histogram would have peaks (row) and a noisy floor (weeds in the midrow). The ratio of peak to floor values of the histogram would be a reliability factor in determining the confidence we have in acquiring the row. If we have a low reliability-factor we can use the seed spacing to further distinguish the position of the row, and to distinguish and precisely locate the good plants relative to our frame of reference. We also have six other rows that can help row discrimination. Since the rows are precisely one foot apart, we could average the row location of the seven rows on the machine to achieve an even more precise lateral position of the implement relative to the actual rows.
We also want a precise coordinate of all the bad plants, that are not in the grid row of good plants This can be done by using the implement as a frame of reference, and with each row camera providing a list of green blobs, each with an x,y coordinate. This weed and row location is useful information and has many uses:
Used to steer and keep the wheels of the implement in between the rows.
Placing the implement precisely over the rows for accurate activities such as:
Weeding, water knife, or mechanical individual weed removal
Working the soil, like roto-tilling or summer fallow between rows
Chewing up trash between rows
Watering, only the good plants
Fertilizing, only where and when needed (shade of green of row)
Sucking-up insects, and vacuuming the good plants
Hilling the good plants
Washing the good plants with a spray of water
Applying fungicide accurately on only the good plants
These nursing and tending operations would start immediately after seeding. The weeding operation could be done with the seeding. When seeding, if there is a particular weedy patch; all the soil could be roto-tilled. How long would it take for an 8’ implement, traveling at 2 mph to work a quarter section? In one hour, it would cover a 32’ strip. Therefore, it would take 2640 / 32 = 82 hours or 3.43 days. Let’s say 4 or 5 days. Once finished it would turn around and start all over again, weeding and tending the crop. So maybe a dozen applications would be done for a single season in the first 60 days. The last 40 days, as the crop matures, may not need any applications.
The Sowing Machine is an implement that plants grain seeds in a row, evenly spaced. There are two seeders that will be discussed: the wheel sowing machine and the punch seeder. There are advantages and disadvantages of each. There will be a discussion at the end on the merits of each. Both, of these machines, have in common, the ability to suck and hold one seed and plant it at a precise spacing and an accurate depth.
The Sowing Wheel is shown in the figure below, moving to the left. A large spring harrow tine is used to clear the path of trash, straw and debris. See Figure 2.
Figure 2. The Sowing Wheel — the parts are numbered
The row needs to be cleared of trash and debris. A spring harrow tine scratches the ground and pushes aside the surface debris and straw. This is done so that the following colter will not hair-pin straw into the seed row.
2 The colter is a thin round disk that cuts the roots and straw down to the depth of the seed row. The outer edge would be sharp and serrated. The colter could be made to castor, to facilitate shallow turns; however, the overall design is for working in a straight line. The colter could have a collar that would roll on the surface for depth control.
3 The plow is used to widen the seed path and to hold the path open to protect the seed being carried on the seed wheel. The colter is thin, perhaps only an eighth of an inch, and the plow widens the path to three eights of an inch. Normally a curved shank or plow will catch and carry trash; but, this plow wraps around the colter so that the colter would clear the trash. The trailing edge of the plow could be used to inject a water/fertilizer mix into the soil. A pulsed piston pump could be used to accurately meter the water between where the seeds will be positioned to keep the fertilizer deep; but, avoiding contact with the seed.
4 The Runner / Shoe is welded to the side of the plow and acts as depth control and somewhat protects the integrity of the seed trench made by the plow. It would extend about 0.75” from the plow. The runner / shoe would act much like a pressure foot on a sewing machine to hold the cloth in place as the needle pierces the cloth. In this case the soil is being held firmly in place.
5 The Seeder Wheel is the main part of the sowing machine. The diagram shows 50 hollow spokes that go to the outer rim of the wheel for 2” spacing, giving a circumference of 100” and a diameter of 31.83”. The rim is only 5/16 “, to nicely fit in the 3/8 inch path that the plow made. The hole in the rim is the same diameter as the seed in question with a center pedestal that looks much like a valve stem on a car tire. The idea is to let enough air to suck up a seed and hold; but, not so big as to have the seed get stuck in the hole.
Picture the wheel rotating CCW. Starting at the 4:30 position, the rim of the wheel is entering the seed box. The seed-box is made with tight tolerance to fit tightly around the wheel, so that seed does not leak out. Vacuum is applied to the spoke, so a single seed is sucked onto the hole and is held firmly onto the edge of the wheel where it exits the seed box at 2:00 o’clock through an opening that is slightly bigger than a seed. The seed is held tightly to the edge of the wheel, all the way around to the bottom of the wheel and the vacuum changes to pressure and the seed is blown off the rim and pushed into the bottom the narrow trench made by the plow. The seed wheel could have a collar on it to ride on the surface of the soil for depth control.
A commutator in the hub of the wheel would be used to pressurize the tubes from 6:00 to 4:30 o’clock to blow off the seed at 6:00 and to clear any debris from the rim’s hole from 6:00 to 4:30 position. All spokes from 4:30, all the way around counter clockwise to 6:00 would have vacuum applied to hold the seed.
6 The seed box fits tight enough around the wheel so that seed grain will not leak out, yet not so tight as to bind. The very bottom of the seed-box has a scraper that keeps the wheel clean of mud and debris and may the occasional seed that is not ejected from the wheel. The seed box is kept about half full of gain – just enough so that there is enough seed for all holes on wheel can capture a seed, yet not so much that the seed is rubbed off. Vacuum is applied to the spokes from the 4:30 position, around the top, all the way to the 6:00 position in which it changes to pressure to force the seed off the rim, yet firmly planted in the seed trench.
7 The packer wheel is shown in Fig 1 packing to the bottom of the seed bed. It would practically be set somewhat higher than this; perhaps to just below the ground’s surface. The packer is also used to control the depth of seeding.
8 The fender wraps around the wheel, much like a fender on a bicycle, but; with a much tighter fit to protect the seed clinging to the rim as it makes its way all the way around.
This sowing machine with the wheel, has the advantage of being able to travel quite fast, but it has the disadvantage of preset seed distance, and much soil disturbance. The punch seeder, that will be discussed next, has very little soil disturbance, and better trash clearance, but; it would have a slower travel speed.
Punch Seeder: Both the punch seeder and the wheel seeder are predicated on the ability to suck up a single seed with a hollow tube and to hold it until it is pressed into the ground at a precise depth. The wheel seeder has many such suction tubes, whereas the punch seeder has only one. See Figure 3 below.
Figure 3. Punch SeederFigure 4. Path of Punch Tip and Seed
To create this J shape, a cam wheel is used to produce the vertical and horizontal displacement. Let’s start with the vertical displacement. Each cycle will be one revolution of the cam wheel which will be broken into 10° increments, 1 to 36 as shown as the horizontal axis in Figure 5. It is assumed that the punch must be lifted over the edge of the seed box of 1 inch. The ground level is referenced as zero of the vertical axis shown in inches. The seed depth is set to 2 inches. The cycle starts at 1 with the tip of the punch in the seed box, sucking up a seed. At 9 the tip of the punch is lifted over the seed box and hen proceeds down, to ground level at 14. At 18 the punch is at its lowest level, two inches into the ground. Vacuum in the tube is switched to pressure to force the seed off the punch. Air pressure continues at 22 as the tip surfaces and is lifted to its highest point at 27 to clear the seed wall. Air pressure is switched to vacuum as the tip of the punch goes into the seed box to suck up and load another seed at 36.
Figure 5. Vertical Displacement of Punch and linear Profile of Vertical Cam
The vertical cam on a 10”cam wheel is shown in Figure 6 with the cog being at the top, being pressed downward onto the cam with a spring. The cam wheel would rotate CCW.
Figure 6. Vertical Cam with follower Cog at the top. Rotates CCW.
The horizontal displacement cam will move the punch tip forward into the seed box, + 1”, and then back to the wall of the seed box which is referenced as 0, at 9. The punch enters the soil at 13 and the horizontal speed of the tip is set to match the ground speed of the seeder. Let’s assume that the seed spacing is 2 inches; therefore, we need to complete a cycle for every 2” of ground covered, — 2” per cycle. If the punch is in the ground for 90° (13 to 22) of the cycle, that would equate to a distance traveled of 0.5 “. At 13, the punch is 1” behind the seed box wall. Over the next 90°, the punch should be pulled back a half an inch to 1.5” behind the seed box wall in a linear manner to match the ground speed. The horizontal displacement cam would look like Figure 7.
Figure 7. Horizontal Displacement
From 22 to 36 the punch is pushed forward, back into the seed box for another load.
On a 10” wheel the horizontal cam would look like Figure 8.
Figure 8. Horizontal Cam with Vertical to Horizontal Conversion Lever not shownFigure 9. Path of Punch Tip
The x or horizontal axis is at ground level, and the y or vertical axis is the trailing edge of the seed box. The tip starts to the left (-1,0) inside the seed box and it is shown much to deep at ground level; it should be much higher. It then travels to the right, over the edge of the seed box (1,0) until it is behind the seed box over the ground. (1, 0.5). It enters the ground (1,0) and then the tip matches the ground speed as it is inserted 2” and extracted (1.7, 0.25). It is then pulled back quickly to the edge of the seed bin and back into the seed bin to load up another seed.
Another cam, on the same cam wheel is used to switch the vacuum to pressure in the tube. Pressure, 15 to 26; otherwise vacuum as shown in Figure 10 and 11.
Figure 10 Air Pressure / Vacuum CamFigure 11. Air Pressure / Vacuum Cam
. A hollow tube slide valve would apply pressure or vacuum to the punch. Vacuum to suck and hold the seed on the tip of the punch, and pressure to release and plant the seed. OR the cam could drive a piston in a cylinder to supply the vacuum, (when pulling back) or pressure (when driven forward) as controlled by the cam. The tube travels in a half circular arc; into the seed-box with suction applied to the hollow to suck a seed onto its tip. The seed and tip then travel in a circular arc, out of the seed box and onto the seed path and placed 2” below the surface, as air pressure is applied to the hollow tube to gently release the seed.
Figure 12. Punch Seeder Tip shown inverted with Seed Stuck on Top rather than Bottom
Whether a seed wheel or a punch; the tip of the tube where the seed is held must be designed carefully to allow as much surface area as possible for the seed to stick, yet, not so big as to get stuck in the hole.
The tip of the punch seeder is shown in Figure 12 with the seed stuck at the top of the tip. The barrel of the punch is hollow to have a vacuum applied to suck up a seed, or to have air pressure applied to remove the seed. The diameter of the barrel is close to the diameter of the seed. Oblong seeds would have the barrel the same size as the longest dimension of the seed. The pedestal in the middle is to support the seed, as it is being pressed into the ground so as to not get pushed into the barrel. The pedestal has a hole in the center and is hollow to allow for suction, and pressure to be applied in the center.
The punch seeder would have a ski, or a runner, or a shoe; much like on a sewing machine to partially clear the seed row of debris, but also to hold the soil firm for the punch and to control the seed depth.
How fast could we seed? Let’s assume the seed separation is 2 inches. I would think the limiting function would be the loading and securing the seed on the suction tube and then doing the semi-circular path to get the seed placed. Consider a valve in a car motor that goes up and down much like our seed punch. At 2400 RPM would be 40 times a second; there is more distance to cover and we need to allow for the seed to be secured and some experimentation would be required, but let’s say 10 cycles per second would be obtainable. Another example would be a sewing machine, certainly it can do 10 stitches in a second. Ten seconds at 2” would result in 20 inches per sec. or 1200 inches per minute, or 100 ft per min. or 6000 ft per hour. – Just over one mph. This might seem slow, but then again if this is a machine that goes 24/7 it could get the seeding done in a timely manner. The sowing wheel could potentially go much faster, but with more soil disturbance.
The punch seeder has less soil disturbance and better trash clearance, but it is slower.
The punch seeder has a more complicated motor driving the cam wheel to match the ground speed, whereas the sowing wheel, automatically rolls to the ground speed. The punch seeder has more flexibility in seed spacing, and is easier to change seed size.
Both seeders have the hollow tube going into a seed box to suck and hold a single seed; and hold it until it can be placed into the ground.
Arguments for 1 foot row spacing: I would like to make the case for one foot rows and then for equal spacing of the seeds in the row. There have been many experiments to determine just how the row spacing affects the yield for wheat, barley, canola and other grains. It was found that row spacing can be increased to 14, 15 inches before the yield is adversely affected. There are many advantages to keep the row spacing as big as possible:
More room to put the wheels of the implement without trampling the crop. With one foot spacing we could have an 8-inch tire, with 2 inch side clearance.
Better ability to cope with trash with the seeder; the wider the row the easier the trash slips through.
Less soil disturbance with seeding
Easier to lock onto rows with a need for less accuracy required for GPS +/- 6” After making a turn in the headland, the tractor needs to hit the next pass before it can visually lock onto to the row of good plants. The wider the row, the easier it is to hit the right spot after the turn.
Better chance for weed removal, with less chance for removing the good plants
Fewer openers are needed for the seeder. This is important as the openers become more sophisticated and expensive.
So we want the row spacing to be as large as possible, but less than 15 inches, so I will suggest that a one-foot spacing is a good compromise. It happens to be the same spacing as current air seeders.
Arguments for consistent spacing within row. OK so we have settled on one foot spacing between rows; but what about spacing the seeds evenly in the row itself. Current seeders just let the seeds fall randomly down the chute. Let’s say we have set the seeding rate to have spacing of 2 inches. There will be seeds in the row that are touching each other all the way to over 4 inches; with an overall average of 2 inches. Why is this so bad? Because, to have the highest yielding crop, one must give each seed access to the maximum resources of water, nutrients and sunlight. We don’t want the plants competing with each other for these resources. This competition is mitigated by having as much distance as possible from its neighbor – that is with equal spacing. So, why aren’t seeders made such that there will be equal spacing? Because such seeders are more complicated. It is not that easy to plant the seeds with precise spacing. Is it worth it to have such a complicated seeder? I think so, for the following reasons:
The crop should produce a higher yield if the resources are shared evenly.
We would need less seed to produce the same density of crop. I would guess that maybe 10% of the seed could be saved.
Anything that grows in the 2 inch spacing, not in the grid, would be considered a weed, and if we got really good at weed removal, we could even remove weeds that are in the row, between the good plants.
When the plants are used for positioning, the 2” grid would provide a tighter more reliable lock.
Inputs such as fertilizer and water could be applied in the row with precision as to not affect the good plants.
To plant seeds evenly in a row, a special seeder is needed, as will be explained in the next section – the sowing machine.
Arguments for Rows: Back in the good old days, when I was a kid, there were two ways to seed: either one used a press drill that seeded in rows with 7 inch spacing, or one used a discer; it spewed the seeds out randomly without any apparent rows. What’s the best? Well, you would want to give each individual plant the best chance at going after the resources of water and nutrients, so it should be as far away from its neighbors as possible. It should be evenly spaced. So one would think the discer would be best. But there is more to it than that. One also wants the seeds to be properly positioned at the appropriate depth and also be packed in nicely; that’s where the press drill is superior. Also being able to deal with trash is important for continuous cropping, and here the discer wins. However, with the advent of the air seeder, with one foot row spacing and more ranks, it would be able to control the depth, do the packing and still be able to deal with considerable trash.
Gardens are planted in rows, why? Why not just scatter the seeds about randomly, and this would give each plant a better chance at access to the resources of water, nutrients and sunlight. Well, there are many good reasons to have rows in a garden:
A gardener needs some place to walk that would not result in trampling the good plants.
A row of good plants is much easier to distinguish. If plants were just placed randomly, it would be an onerous task to identify and discriminate between good plants and weeds. If there is something green, between the rows, it is by default a weed because it is in the wrong place.
When seeded in a row, it is easy to pack the seeds in tightly in the row. And I would like to make the case for one foot rows and then for equal spacing of the seeds in the row.
OK so let’s assume that we have all the technology to have autonomous tractors; and we do. The operator would be just sitting on the tractor with nothing to do but go along for the ride; but, not really necessary. So, let’s say we remove the operator; what are the ramifications. Obviously, if we don’t need the operator, we don’t need all the cab and all the operator amenities; but that is nothing compared to the realization that the only rational reason for having large, and I mean really large equipment is because of the operator. To best make use of the operator’s time, one must have large equipment. However, if an operator is no longer required, then all rational logic leads to smaller equipment. Some say that larger equipment is more efficient; but this is only in terms of the operator’s time. If there is no operator, this is a moot argument. I would like to demonstrate that smaller equipment is more efficient, more reliable, less costly and more convenient than larger equipment.
Let’s be specific; let’s compare a 60 foot and an 8 foot piece of equipment working a quarter section (160 acres). How much overlap will occur? Let’s assume that the lateral overlap has been eliminated with auto-steering, and that ideal turns can be made in the headland. Comparing the area covered in the headlands, it is found that this covers a half a circle for every one width pass, or a full circle for every double pass.
A quarter section is a half mile by a half mile (5280/2 = 2640 ft). A 60 foot applicator needs 2640/60 = 44 single passes, or 22 double passes. Each double pass requires a full circle area of radius 60’ = p 602 = 11,310 ft2 whereas the 8’ implement requires 2640/16 = 165 double turns x p 82 = 201 ft2. In finishing the field, the last pass would be, on the average, half an implement width, resulting in 30 x 2640 = 79,200 ft2 for 60’ and 4 x 2640 = 10,560 ft2. So a 60 ft. double covers an area of 11,310 + 79,200 = 90,510 ft2. Whereas a 8 ft implement would double cover 210 + 10560 = 10,770 ft2.
A quarter section is 6,969,600 ft2. 60’ has 90,150/6,969,600 = 1.3% double area coverage, whereas the 8’ is only 0.15% double coverage.
Another way to look at this is one would expect that a 60’ implement, going the same speed as the 8’ implement would take 7.5 times the amount of time to do a quarter section. (90,510 + 6,969,600) / 60 =117,668 time units. Whereas the 8’ would require (10,770 + 6,969,600)/ 8 = 872,546 or 7.41 not 7.5.
Taking the operator off the machine means we don’t have to cater to the operator; we can work 24/7. Assuming that an operator needs to eat, sleep, and take occasional rest breaks, let’s say we have 12 hours out of the day that he is actually driving. So, to do the equivalent of a 60’ machine only half of the 7.41 8 ft machines or 3.7.
When we have an operator driven machine, we must provide some over capacity to ensure the job gets done in case the machine or the operator break down. However, there is reliability in numbers. If the 60’ machine breaks down for a week, you are doing nothing for a week. However if an 8 ft machine breaks down for a week, and you have 4 of them, you still have 3 good ones working, or 75%. I would be so bold as to say that three 8’ machines working 24/7 would do as much work as the one 60’.
Smaller is cheaper. Let’s say that the 60’ cost $120K and that half the cost of a tractor goes into keeping the operator happy. So an operator-less 60’, without cab etc. would cost $60. Let’s also assume that the cost is proportional to the number of feet that the implement is. Therefore an 8’ would cost $8K and three 8’ would cost $24K. So, going from driver to driverless can save an enormous amount, $120K to $24K.
Operating costs could also go down. We don’t have to haul around the extra weight of the cab and we don’t need to provide 15 HP for air conditioning. If tendering can be done more frequently, smaller amounts of fuel, seed, and other inputs need to be dragged around the field. Reducing the speed of the machine, can greatly reduce the amount of energy required as well as the weight of the engine. Just roughly, I would think that fuel costs could easily be decreased by 10% or more.
The bottom line to all this, is that having a driver, as opposed to having an autonomous system is that: having a driver is a very expensive proposition.
An 8’ machine is also easier to work on, is less onerous in terms of safety, and is easier to transport when one consider the legal 8’ width for any road, highway or bridge. And with less weight comes less soil compaction. Many, many advantages to go small.
Less compaction, easier to store, easier to fix, we are down to less than 1% of overlap in headland therefore we can accommodate smaller fields with sloughs, trees, shelter belts and other obstacles with little consequence. Going smaller is environmentally friendly.
No operator: So, we are now at the point where we can drive our tractor to the starting point in the field, the designated starting point of the predetermined course and then let the tractor automatically follow this path as accurately as it can; all the while controlling the applicator and the tractor until it gets to the end point of the course. The operator on the tractor is literally going for a ride. Why are they there at all? Some would argue that an operator is needed to monitor the process and be able to intervene just in case ‘something goes wrong’. I harken back to the days when my Dad put me on the tractor and told me that if something goes wrong, just stop the tractor and walk home. The same thing could be said here, if the computerized control system detects something wrong, it could simply stop the tractor and message (text) the farmer of what fault it has detected. The farmer could then go out to the tractor and investigate and remedy the situation.
Another argument for having an operator is that someone is needed to look out for intruders in the field such that the tractor does not drive over them. I see three levels of security here. First there could be a motion detector or infrared detector that could detect intruders. Secondly, a mechanical bumper could be placed in front of the tractor and applicator such that if it encountered or touched anything it would immediately stop the machine. And thirdly, an independent program could be checking to see that the tractor is always within the bounds of the field. If the tractor is ever caught outside the perimeter of the field, or within the boundary of an obstacle, it would immediately shut down.
There is also the thinking that driving a tractor is much too complicated for a machine to anticipate all that could go wrong. A little story is in order: Imagine years ago, maybe in your grand parents era, that a travelling salesman comes into the farm yard selling automatic washing machines. He explains to your grandmother, that all you have to do is put your clothes in, set the thing going, and you can walk away and do something else. A half hour later you can come return and your clothes are done – clean, and ready to hang on the line. This is incredible, and your grandmother says this is hogwash, impossible and she chases the salesman out of the yard. Later she goes in for dinner and tells everyone the incredible story of the automatic washing machine. Crazy, just how would this machine be able to pick up the clothes and put them through the wringer? How would this machine be able to monitor when things go wrong – like when the clothes wrapped around the wringer. No way the machine would never know how to reverse the wringer. The point here is that machines can be designed to be automated, and they can mitigate the faults. Manually driven machines, often fail when the operator pushes that machine beyond its design limits. If you get too close to that slough, you will get stuck.
To remove the operator from the tractor and have it completely autonomous; more than a course to follow is needed. The control system must have complete reliable control of the tractor, the steering, the powertrain and the brakes. Also control of the applicator is needed – turning it on and off at precisely the appropriate time. Monitoring the tractor and the applicator for overheated bearings, oil pressure, water temperature, and air pressure in the tires. Instrumenting and monitoring the tractor and applicator are all within the realm of today’s technology and in fact most of this monitoring is already being done today. The difference is that the results are now being sent to the operator, whereas with an autonomous system the monitored measurements would be relayed to a remote location – perhaps to the farmer’s cell phone while sitting in his half ton. Security to make sure that the tractor does not drive over someone or something. This could be achieved with layers of security. As earlier mentioned, the first level would be a motion detector and/or an infrared scan. The second tier of protection could be a mechanical bumper, and a visual video analysis of what is in front. To ensure that the tractor is always in the field, an independent software program could always be checking. A good reliable communication link would be needed from the tractor to the farmer, who may be miles away. If a fault or something unusual was detected, or perhaps the tractor just needs servicing, fuel or more inputs such as seed or fertilizer – it would need to be able to inform the farmer of its status. In most cases it would quietly do its job, and only demand attention when extenuating circumstances arose.
The advantages of removing the operator from the tractor are many. The most obvious is that you do not have to utilize the time and abilities of the operator on such a mindless task. The tractor could be much cheaper without the cab, air conditioner, and all the other comforts and amenities an operator requires. So, the tractor itself would cost less. Yes, you would need the computer and electronics to do all the steering and automatic stuff, but it pales in comparison to the cost of the cab and operator. I have heard that half the cost of a tractor goes into the comforts and conveniences of the operator. And since we don’t need the air conditioner and will have less weight without the cab, the fuel use could be reduced. An air conditioner can use 15 HP. Another advantage in having the process automated is that since we are not concerned about the operators time, the machine can operate 24/7, it needs no rest, it does not take holidays and it can drive in the dark. The speed of the tractor can be lowered and kept within the design limits of the machine, and at speed, optimum for that operation. This will save fuel and minimize the amount of break downs since it is not being pushed to the limit. But perhaps the greatest advantage in going automatic without an operator is that one can now have smaller tractors and smaller implements which it turns out are more efficient, as will be demonstrated in the next section.
OK so let’s say we can track a course-line and do a U turn in the headland. What’s next? Could we maybe define a track for doing the entire field, headland, straight passes, turns; the whole field, everything? I think we could. A field course could be computer generated, but the computer would need to know a few things: the perimeter of the field, of which it is understood that the perimeter can not be breached. Also, the perimeter of any obstacles such as sloughs, wet spots, bushes, utility poles, rock piles buildings and what have you. The tractor or implement is not allowed inside this obstacle perimeter. The computer must also have knowledge of the width of the implement, its drag distance and the minimum turning radius as well as the maximum change or the upper limit of the turning rate. (this will define the transition distance in a turn). The start position, or field entry, and end of the track position (field exit) would be needed. The desired pattern, round and round, or parallel passes would be needed. If parallel lines are selected, then the direction of the parallel lines would be input, unless you let the computer find the optimum direction for best efficiency.
How do we establish the perimeter of the field and the obstacles? Some have suggested that this could be done by remote sensing or from Google Earth. I would argue that that is not accurate enough. One must be on the ground making continuously scrutinizing as to where the perimeter is precisely. What’s a wet spot and what’s not? To make that call, one must be right there. I would suggest that a small quad ATV could be used equipped with a GPS receiver; the coordinates would be collected to define the perimeter as the farmer drives the quad precisely where the perimeter is to be established. Most perimeters are stable and the remain the same year after year. However, there are some perimeters or parts thereof that would require updating such as wet spots. In the end, the perimeters of the field and the obstacles would be a set of coordinates that could easily be input to a computer. With all the other fore mentioned parameters; we would be ready for the next step – to develop a series of coordinates, from the start to the finish that would define the track that we want the tractor to drive. This would include the straight lines, the turns, the headlands – everything. Defined by a long series of coordinates. The objective of the computer program that makes the course would be to have the entire field covered, without any missing, with the shortest distance for the greatest efficiency. By allowing some flexibility in some of the inputs, it might be possible to work the field with an even shorter distance. For example, if one adjusted the angle of the parallel lines. The best direction for parallel passes, is to be parallel with the longest part of the field which provides the longest passes, with the fewest turns. The nice thing about predetermined computer-generated courses is that they can be done repeatedly for the best possible course. With manual driving, we find that the last pass always involves considerable overlap. With the computer-generated courses the last pass can be adjusted to minimize the amount of overlap.
How much more efficient would a computerized course be? In one study, driving a quarter-section (160 acres) with a 60’ sprayer: driven manually was 42 km but a computerized predetermined course was 34 km. This was 23% more. As a rule of thumb, manual driving can be improved by 10% by eliminating lateral overlap, and another 10% by making optimum turns and smart decision in adjusting the patterns position and in negotiating turns.
A hybrid computer/manual means of developing a predetermined course by having the operator of the computer manually design the course by hand. This might be a way to get computer aided courses before developing a sophisticated computer program.
If the tractor is now able to follow a predetermined course from start to finish without human intervention, one may ask: Why do we need an operator sitting on the tractor?
U turn: So far we have described how to do parallel passes that are straight and with negligible overlap; but, when we get to the headlands at the edge of the field we need to turn around and catch the next pass. We need to do a U turn. An optimum U turn would be a turn with the shortest distance in the least time. Manual U turns are anything but optimum. The operator goes deep into the headland in order to be given time to merge onto the desired course line. This is exacerbated by the distance the applicator drags behind the tractor. Also, if the passes are not precisely perpendicular to the headland, extra distance is required to make up for the wedge.
Executing a U turn in principle is the same as staying on course for a straight pass. The difference is that a U turn requires an arced or curved path, whereas the parallel pass is straight. The turning rate for a straight path is zero, whereas the turning rate for a U turn can be calculated, knowing the width of the implement and the speed of the tractor. Assuming that we want the applicator turned off during the turn, it is also necessary to know where the headland starts and the drag-distance of the applicator. Let’s list all the parameters that are required to define the desired course of a perfect U turn: width of the implement or applicator, the drag distance of the applicator, edge of the headland, angle of headland to parallel pass. And to execute the steering to the desired course of the headland we need to know the position of the tractor, the speed of the tractor, and the turning rate. By using all these parameters, it would be possible to automatically do a U turn, by steering the tractor automatically onto the desired predetermined U turn course. Yes, it is more complicated, with more input information required, but it is doable.
A U turn course is much more than a half a circle because some time is required for the tractor going straight, to actually turning to a rate that defines a half circle. We could call this the transition time to get into the turn. The same transition time is required to come out of the turn. The transition time is set by the time it takes to change the turning rate and it could be significant for a larger massive tractor. The transition, makes the calculation of a U turn much more arduous.
The application control is a part of the U turn. We want the applicator turned off in the headland; yet, we don’t want to miss anything. By knowing the delay of the applicator control, it would be possible to precisely turn the applicator on and off at the precise time.
There would be an advantage in doing the headlands at the very end. With manual driving, the headland marks where we are to turn. With auto-steering, there is no need to have a visual mark, as the system knows the location of the headland. In doing the headlands last, it would be possible to cover all the turn marks and ruts made with the parallel pass turns.
Knowing when to turn in a V trap is another interesting dilemma for auto-steering. The system must always be cognizant of having enough room to turn before being trapped; even though it might result in some misses.
Headland Alert: OK so we can do parallel lines; and this would go a long way in reducing the overlap; but what is next? What about the turns, or at least a warning of the turns? We learned early on, when we introduced auto-steering, that the driver’s attention waned, with little to do as the tractor drove itself across the field. One was now free to make a call on a cell phone, or have lunch, or maybe even have a nap. We heard of all kinds of horror stories of farmer’s falling asleep and the tractor faithfully following its line right into the neighbors field, or onto a highway, or into a slough. Consequently, we introduced a headland alert, which was nothing more than a beeper that went off just before the headland. However, it was still incumbent upon the operator to recognize the warning and to make the turn in the headland. The next evolutionary improvement would be to make the turn in the headland. A perfect U turn.
Auto Steer: There is more to driving a nice straight line down the field than one first thinks. Yes, one needs to judge the distance to the last pass, but just as important is the angle of approach and the rate of convergence/divergence to the last pass. When we drive manually, we combine all these visual cues and make a quick calculation as to how much we should move the steering wheel. We do it without even thinking – it becomes so automatic. It’s not perfect, but one can get pretty good at judging the distance and the angle to the last pass. Can we mechanize the control of driving a farm tractor? The distance to the last pass can be done with a positioning system like GPS, but how do we measure the angle? And we don’t have all day to do this, it must be done quickly, it must be done now – without delay. A delay in a control system becomes the juggernaut Eventually, we used GPS for the position, even though it’s accuracy and delay were questionable. Later we will see how the plant row can be used to improve the accuracy. An electronic gyro would be used to acquire the angle, but it was not without challenge. Electronic gyros put out a voltage that is proportional to the turning rate. One must integrate this to get an angle. And there is drift in the output with age and temperature. Corrections and compensations must be done in real time. Also, the vibration of the tractor, if close to the sample rate of the gyro, could result in egregious angle errors.
In Engineering we are taught that a control system with the least error in tracking is one with critical damping. So, if we are driving and we are off the desired line, we want to get back to where we belong as quickly as possible and this will result in some over-shoot. There will be a little loop in the corrective action taken. When doing the auto-steering trials; I soon learned that farmers wanted to forego this optimum response. They did not want to see any little loops or squiggles – they wanted a perfectly straight line no matter how long it took or the resulting accumulated overlap error. They wanted a controlled correction that was over-damped.
There were three things used in determining the steering correction: the offset, the angle and the tractor’s rate of turn. Let’s look at each of these more carefully.
The Offset: This is the perpendicular distance of the center of the tractor, to the desired track line. It is shown in Fig 1 as the offset. The desired track line is set by the driver doing a strike line across the field. From this strike line a set of parallel lines is calculated one implement width apart. The desired track line is the parallel line closest to the tractor’s position. When the positioning systems coordinate (GPS) is received on the tractor a calculation is made using a point to line formula, yielding the offset. GPS is accurate to ten feet, but it can be improved by using differential GPS, DGPS to get to within about a foot. This can be made even better with Real Time Correction, RTK GPS, to get to inches, especially if we are using relative measurements within minutes of each other. However, all positioning systems do not report the current position of the tractor, they report where the tractor was a moment ago, a fraction of a second ago. This just adds to the difficulty of steering.
The Angle: The tractor is shown on its course line to the * on the desired track line as shown in Fig 1. The angle j is the angle between the actual course line and the desired course line. Some call it the angle of attack. When we have a large offset, we need a large angle of attach, but as we get closer to the desired line, we want to make the angle smaller and smaller so that we don’t overshoot the line. One could think that driving is actually just aiming a few hundred yards down onto the desired line.
Rate of Turn: The tractor could be turning, so the turning will determine what the angle will be in the future. So it is another factor in determining the steering correction. An electronic gyro is used to determine the rate of turn, such as degrees per second. The gyro we used put out a voltage from zero to five volts, with the mid voltage of about 2.5 volts being no turn. A voltage higher than this was a left turn, and one lower than this was a right turn. We called this no turn voltage zero_w. It was not exactly 2.5 volts. It was initially calculated by averaging 32 readings when the tractor was standing still. When the tractor was moving, a reading from the gyro minus zero_w gave the turn-rate. Digital integration of the turn-rate gave the angle j. Angle j is acquired and adjusted by taking the inverse tangent of the difference of past offsets over the distance traveled. The correction to the steering = offset – angle – turn-rate.. This correction was a pulse applied to a hydraulic valve that put a squirt of oil into the steering ram.
