Introduction



Mechanical Design

The mechanical design of this robot has been a very extensive and time-consuming process. Because everything is designed to fit together seamlessly within a custom-made chassis, the design had to be iterated through many times in order to achieve an acceptable solution to any given design problem. Some of the final iterations are covered here, while a somewhat more exhaustive (though still not comprehensive) list is included in Appendix B.

We will begin with some of the manufacturing considerations, component selection, and notes on the general approach, and then cover individual portions of the mechanical design. Where possible, we have included diagrams, illustrations, photographs, sketches, and calculations, however, a tremendous amount of brainstorming, designing, and streamlining took place on dry-erase boards, chalkboards, or in software form which has since been overwritten in order to save disk space.

The documentation included in the body of this section and in the appendices should be sufficient to give a clear impression of the design process as well as the final design.

Manufacturing Considerations

The manufacturing considerations taken into account during this project were fairly simple: whatever we designed, we had to be able to make. Though there is money budgeted for tools, there is no space in the budget for machine shop time or professional tooling. Our design success metrics include “95% of manufacturing done in-house,” and we’re shooting for 100%. Furthermore, one of our risk mitigation actions is to design with manufacturing ability in mind. That is exactly what we intend to do.

Fortunately, this did not pose much of a constraint for the computer science portion of the project. All of the tools necessary for development are available open-source, and any software not available to use we should be able to develop ourselves.

The mechanical portion, however, is significantly limited by the available tools. Many small tools are readily available, including hand drills, screwdrivers, wrenches, and other hand tools. However, in order to produce a structure capable of supporting the loads under consideration here, we will need much more than that. An angle grinder, abrasive cut-off saw, and drill press can be found fairly cheaply, as can the disks, bits, and brushes necessary for their operation. There is, however, one very costly item that must be purchased if we are to successfully construct this robot.

We needed a welder. This was possibly the most uncertain part of the entire project. Everything else in this project dealt with something we had learned or school or built upon computer science or mechanical engineering studies. Welding, however, was the one field that neither of us had any experience or background in.

After extensive research and a lot of time discussing the topic with experienced welders, we have been able to learn a lot and have selected and purchased a MIG welder. The Hobart Handler 140 (pictured right) plugs into a standard 115V wall outlet and can weld material up to 0.25” thick. This should fill our need nicely.

This, however, led to one of our major design decisions. The chassis, axles, mounts, and structure will all be made of steel. This is because steel is a very easy material to weld, and with our limited experience, we’re trying to keep this portion of the project as simple as possible.

Other design considerations have driven our decisions to use as little plate as possible, as this cannot be easily or precisely cut with our tools, but rather, must be sheared at the time of purchase, which is very expensive. We have also chosen to use weld nuts (pictured left) in order to avoid the need for taped holes and reduce the number of broken taps.

Each time we began to design a new component or system, we simply kept in mind what tools we had available to make them with, and avoided from the start designing anything that we would not be able to build.

Component Selection

Because minimizing the size of this robot is so critical to its mobility, it has been meticulously designed and redesigned around the components in order to reduce its footprint. In order to do this, we first had to select our components and then develop a supporting structure that suspended the components in as efficient a way as possible.

Now, with the completion of all of our phase one six sigma work and input specification requirements documentation, we are ready to begin selecting the components that will power the robot. These components will define the behavior and capabilities of the robot and must be selected with special attention paid to the customer requirements. The translation of functional customer requirements to the mechanical specifications of the system is the first step in the mechanical design.

For example, we cannot sit down with our customers and ask them what voltage motors we should buy. Our customers do not care about things like this, and should not be required to figure it out for us. It is our job as engineers to take the requirements given to us in the form of deliverable features and, from those requirements, design a product capable of delivering them.

Motors

The easiest place to begin is with the motors, as this component will most visibly affect the performance of the robot. The combination of our customers’ requirements as well as our own limitations provides us with plenty of information to refine our search and make a selection.

There are four major options for motors of this scale. We can choose from electric, internal combustion, pneumatic, and hydraulic motors. Pneumatic and hydraulic options are much more complicated and take specialized hardware, tools, and knowledge. Due to this complexity, we have opted out of these two options. Internal combustion engines, while not quite as complicated, are more difficult to work with than electric motors. There are also issues with safe fuel storage, running in enclosed areas, and noise. For simplicity, we will be using electric motors. They are simple, environmentally friendly, quiet, and easy to keep powered.

This application will require a high power electric motor. We expect the robot itself to weigh somewhere in the ballpark of three hundred and fifty pounds, and it should be able to carry an additional three hundred pounds of payload. Therefore, our motors should be able to perform the required tasks while driving a six hundred and fifty pound bulk.

The power requirements of this robot are fairly simple. It should have a top speed of approximately twenty-five miles per hour, and it should be able to climb stairs at a controlled speed. The stairs should be the most difficult task that these motors will be required to perform. If they are capable of climbing steep stairs at ten miles per hour, they should provide plenty of power to drive the robot.

If we assume that a steep flight of stairs climbs at an incline of forty-five degrees, we can easily calculate the amount of power necessary to drive up those stairs at a given velocity:

So, with roughly 12.3 horsepower, this robot will be able to carry a 300 pound individual up a flight of steep stairs at ten miles per hour. Just to make sure this will be sufficient for flat ground driving, the time to achieve top speed with constant acceleration can be easily calculated (We assume 5.657 average horsepower over the range of velocities).

So, starting from rest, given 12.3 maximum horsepower output, and carrying a 300 pound person, this robot should be able to accelerate to a top speed of 25 miles per hour in around 4.4 seconds.

This calculation of acceleration includes some sweeping assumptions. It assumes that the motors will be able to produce constant power over a range of speeds and provide a perfectly constant acceleration. Despite these assumptions, the results above do give us a reasonable approximation. Acceleration to top speed in under five seconds is, if anything, too fast. If these numbers are off, even by a factor of as much as five, it will be within acceptable levels. At 4.4 seconds, we will make use of the acceleration-limiting feature of our speed controller to increase this time for safety reasons.

So our motors will be required to produce roughly 12 horsepower. For simplicity, we plan on using slip steering to control the robot, so we will require two motors. This means that each of our motors should be able to provide a minimum of 6 hp.

Now that we have an idea of what will be required of the motors, we can look at some alternatives.

The PMG-132 (pictured right) is a permanent magnet DC motor that operates in the range of 12 to 72 volts. It weighs 24.8 pounds, provides 50.2 rpm per volt, 27 ozf-in per amp, and draws 110 amps continuous. At 48 volts, its peak efficiency is 86%, its peak power is 15.1 hp, and it stalls at 960A. The price for this motor ranges between $695 and $1200.

(, ).

The Black MAX (pictured left) is another permanent magnet DC motor. It operates between 24 and 36 volts, weighs 15.7 pounds, provides 140 rpm per volt, and can produce up to 3.8 hp at 24V. It draws 200 amps under normal use. This motor costs around $250 to $300. (, ).

The S28-400 Magmotor (pictured left) is a very powerful, compact motor. It weighs less than 7 pounds, and is only 3 inches in diameter. It is rated at 24 volts, but can be run higher with the proper precautions. It can produce a maximum of 4.5 hp, 3720 oz-in of torque, and 4900 rpm. Its max draw is 390 amps with a peak efficiency of 83%. This motor costs roughly $350. ().

The Etek electric motor (pictured right) is manufactured by Briggs and Stratton for use as drive motors in golf carts. It can operate between 24 and 48 volts, weighs 21 pounds, and is nearly eight inches in diameter. It produces 1.14 in-lbf per amp, 72 rpm per volt, and a peak efficiency of 89%. It draws a maximum of 330 amps for two minutes before burning out and produces up to 15 hp at 48 volts. The price of this motor ranges between $375 and $450.

(, ).

In addition to these four motors, we briefly considered countless models. These four represent the best-suited alternatives to this application. Our final choice was fairly clear. The Briggs and Stratton Etek motor fits the requirements quite nicely. The PGM 132 will also do the job easily, but it is far more costly and has the capability to run up to 72 volts, far more than we have any intention of using. The other two options fall short of the horsepower requirement. We considered using two Magmotors for each side briefly, but decided that there was insufficient benefit to warrant the added complexity.

From the documentation we have found for the Etek motor, we have opted to run them at 36 volts. This should provide 85% or better efficiency, roughly 2200 rpm, and 5 hp while drawing only 120 amps. Provided that our speed controller can handle 120 amps continuous, it should be able to handle significantly more for non-sustained periods of time. This should allow for the additional 1 hp per motor required for climbing a flight of stairs. Provided that the robot is not required to climb a continuous flight of stairs for more than two or three minutes at ten miles per hour, this will be sufficient.

Batteries

We have examined a vast number of options as far as batteries go. The electrical system of this project draws on one of the most limited parts of our knowledge, and so it has been quite a learning experience. We know that the batteries must be capable of providing enough current to drive the motors. This would mean 120-240 amps under normal use and 330 amps for 2-minute surges. We also know that our customers require a 30-minute battery life between charges. We want to run the motors at 36 volts. Finally, there is a major safety requirement of the system. We cannot guarantee that this robot will never roll over or be called upon to operate in unusual orientations. This means that we must use dry cell batteries. This means that car batteries or other, non-sealed, electrolyte batteries are out of the question; these batteries are not safe if not operated right side up. We know that if we do use a lead acid or similar battery, it must be sealed.

What we do not know, however, is what type of batteries to use. After some brief research, we found four major options: sealed lead acid (SLA), lithium ion, nickel cadmium (NiCad), and nickel metal hydride (NiMH). Each have their own unique properties, pros, and cons associated with them, and were each suited to different applications. Lithium ion, NiCad, and NiMH batteries are the type of batteries typically seen in laptops, cell phones, remote control cars, PDAs, and other electronic devices. They are lightweight, charge quickly, and have a long life span. They are also expensive. There is a company that makes custom BattlePacks () for use in combat robotics. These are high performance packs of lithium ion, NiCad, and NiMH rechargeable cells designed to provide the high current draws associated with such a rigorous application. They cost five to ten times more than SLA batteries.

Because we are not required to fit into some restrictive weight classification like many combat robots, this added cost is unjustifiable. We will be able to accommodate the added weight associated with lead acid batteries. There are, however, several drawbacks. Even though a sealed lead acid battery is safe to have overturned, they do not run efficiently when upside down. Also, though they will not leak fluid, they do vent flammable gasses. This means that they must be properly ventilated and isolated.

While we were researching our options in SLA batteries as well as the design considerations that would be included with using such a battery, we discovered another type of battery that was not included in the initial search. This battery is an evolution of the SLA battery; the absorbed glass matrix (AGM) battery. An AGM battery takes all of the lead acid and electrolyte and absorbs it into a glass material, which immobilizes the electrolytic fluid. These batteries provide deeper cycles, higher amperage output, less internal resistance, no flammable gasses, and full efficiency when operated upside down. They are slightly heavier and more expensive but still much cheaper than anything except SLA.

At first, we selected an Optima Yellow Top deep cycle AGM battery (pictured left) for use in the robot. This is a 12-volt battery, so we would need to use three of them wired in series to get 36 volts. It uses a sulfuric acid H2SO4 electrolyte and is encased in a polypropylene case. It weighs 43.8 pounds, has 0.0028 ohms internal resistance when fully charged, and is rated at 55 amp hours. Unfortunately, the relationship between current and discharge time is not linear. The graph below is from the spec sheet on a SLA battery rated to 12 amp hours, but the shape of the curve clearly shows how the life of a battery drops off with increased current draw. From the documentation available on these batteries, we can expect to get roughly 80 amps of continuous current for 30 minutes before the voltage dips below acceptable limits.

Because 80 amps was the maximum continuous current draw of the motors, we deemed these Optima batteries acceptable and moved on in the design. It was not until later that we realized that, while each motor would only be drawing 80 amps of continuous current, there were two motors and two channels that would each be drawing power. This took our predicted battery life to fifteen minutes or less and sent us back to the drawing board.

Optima does not make batteries big enough for our application. We needed 100 amp hours, and Optimas cannot be found with more than 70. With further research we found that the spiral cell technology implemented by Optima provided higher instant-aneous current draw and better vibration resistance but at the cost of capacity. Other AGM battery manufacturers make flat-plate batteries with tremendous capacities and some 100 AH batteries are not that much bigger than Optima’s 55 amp hour variety.

After a great deal of time spent researching battery manufacturers, we found a company named Concorde. Concorde makes three lines of batteries that could fit our application: Lifeline, Chairman, and Sun Xtender. The Lifeline batteries are house batteries, designed for backup power systems and long term storage. Chairman batteries are designed to be used on electric wheelchairs. Sun Xtender batteries are designed for storage on solar power systems.

Further research yielded the exact differences between these three kinds of batteries: their terminal types. Except for the type of post used to connect to the battery, the internals of each of these kinds of batteries were exactly the same. We looked around, found that Sun Xtender (pictured right) provide the lowest profile terminals and can be found for around $200 each, delivered. They measure 12” x 6.8” x 9.3”, weigh 65 pounds, provide nearly 850 cold-cranking amps at room temperature, and 100 amp hours at a 20 hour rate.

Their increased size would require a huge chassis modification, but that was far better than a mere 15-minute battery life. We are very happy with the results of our research, and are confident that three, 100 amp hour Concorde Sun Xtender batteries will provide sufficient power for this robot for 30 minutes in all but the most demanding driving conditions. If operated with low gear reduction at top speed up steep grades continuously, the battery life may dip below 30 minutes of useful life. Under these most extreme of conditions, this is acceptable.

Power Transmission

Now that we have selected our components, we are ready to move on and begin designing the various components and systems of this robot. The easiest place to begin this process is with the drive train. The drive train is the system that will take the power generated in the motors and deliver it to the ground.

Our drive train has several requirements placed on it. First of all, each side of the system must be capable of transmitting, at a minimum, six horsepower. Ideally, it should be able to handle much, much more to allow for shock loading and surges in the electrical system. Also, our customers have requested three speeds of operation: 25 mph, 10 mph, and 5 mph. For this requirement, hardware and software are logically equivalent. It would be possible to program the computer system to limit the speed of the robot without making any physical gear reduction, but that would require running the motor at a wide range of speeds, which would adversely affect efficiency.

In order to avoid this, we have opted to have one physical gear reduction, and one software gear reduction. The software reduction is a consideration for the computer science portion of this project and is discussed in that section.

As far as the mechanical design, we will require a drive train capable of driving at a maximum speed of either 25 mph or 10 mph, depending on which gearing is used. This means that the gear ratios necessary to achieve these speeds must be available and the user must be able to switch between the two in some way.

As published by the manufacturer, Briggs and Stratton Etek DC electric motors will deliver 72 revolutions per minute per volt. The anti-ratchet snowmobile drive sprockets we intend to use for the driving wheels are 7.5 inches in diameter, and snowmobile tread is roughly one quarter of an inch thick, giving us a 4 inch radius from drive axle to the outside surface of the tread. This gives us:

We will achieve the reduction from 10 mph to 5 mph through software means, so these two gear reduction options should be sufficient.

We spent some time considering a two-stage reduction, but, in the end, there was no room in the design for the additional hardware. Furthermore, we could not determine a simple mechanism for changing gears or maintaining the necessary tension with this setup. Our final design includes a one-stage reduction. The gear change can be achieved by removing the pinned chain link, removing drive sprocket from the motor, and replacing the driving sprocket with a different sized one.

We have a few options for power transmission at our disposal. We can use chain and sprocket drive, belt and pulley, or gears. We discounted gears early. We have no way to machine mountings with the necessary precision to mount our own, and a quick search of the available pre-made gearboxes revealed prohibitive prices. After examining the available chain and pulley systems available, we have selected ANSI standard #40 chain and sprockets for the drive train.

We will mount a 60 tooth, 10-inch diameter sprocket on the driving axles and either an 11 tooth or a 30 tooth driving sprocket on the motor, depending on the desired gear ratio. Number 40 ANSI standard steel roller chain has a maximum working load of 810 pounds, which should easily support the expected loads. Given a torque constant of 1.14 in-lb/amp, 180-amp draw, and a 1-inch radius small sprocket, we have:

1.14 in-lb/amp * 180 amps / 1 in = 205.2 lb

So the chain should easily be able to handle the loads created by the motors.

Suspension Elements

The suspension system for the robot is one aspect of the design that required many revisions before we arrived at a solution with which we were satisfied. Each solution was either very costly, introduced a great deal of complexity to the design, or simply did not fit inside the structure of the robot that we had envisioned.

The first and most straightforward option we came across was go-kart shock absorbers (pictured right). These provide two inches of compression under 450 pounds of load, which is appropriate for our application. The problem with these, however, is their size. The eye to eye length on the smallest shock absorbers we could find was 12 inches. The entire height of the chassis of the robot is 10 inches. These simply will not fit in the structure of the robot as we have designed it. ().

Another option that has some rather attractive implications was the use of air springs. These springs allow the user to adjust their stiffness by adding or removing air, which would allow us to fine-tune the suspension of the robot. They are also quite compact in design and are very robust. They are, however, quite expensive. The smallest and cheapest (pictured right) run approximately $30 and only compress one inch. An air spring with two inches compression costs more than double that figure. Despite the high cost, however, it would have been worth it for such an elegant solution, but their compression versus resisting force curves (below) show that they would be quite poorly suited to our application. As the springs are compressed (from right to left on the horizontal axis), the load increases through the first half of the desired range in an acceptable manner. However, from 2.5 inches to 2 inches, the load supported tables off drastically. This eliminates half the already limited useful range, and is unacceptable for our design. ().

We also considered using leaf springs for a long time. We found very small trailer leaf springs that measured 20.5 inches from eye to eye, and were rated to only 750 pounds. We went through several very extensive design iterations using these but finally were forced to abandon even these as the complications they introduced became too numerous. They took up a lot of space, required very unique geometry to allow for their travel, introduced angular displacement in two dimensions, and were difficult to mount.

There were many other possibilities considered during this process. We considered several different linkage solutions, but each was abandoned as the costs of bearings, bushings, and supporting members added up. We also considered several compression spring combinations but could never constrain the system without the use of a linkage, and the risk of buckling under extreme conditions was unacceptable. We considered torsion springs of several different configurations, but each of those solutions introduced angular displacement in planes that would have destabilized the tank tread of the robot. It seemed that we were very quickly running out of viable options.

Then we considered what seemed an impossibly simple solution. We will suspend the robot with rubber bands. At first, we envisioned molded rubber tie down straps used widely in trucking and marine applications. These could be cut into strips and strapped over the axle to provide the necessary downward force. We were concerned with the lack of documentation available on these straps, but at $2 each, we could easily afford to buy some and test them to find out for ourselves. The results of these tests are discussed later.

With a little research, we found molded polyurethane straps that are superior in every way to the rubber ones, and, after testing them, we are quite happy with their performance. They will provide 40 pounds of force when elongated to double their original length, and they even have acceptable levels of creep when loaded in this manner. With five straps per side, we will need a total of 30 straps to provide the necessary force, but even at that many, they are far cheaper than any other solution considered.

Tread and Wheels

One of the initial design decisions of this project has been to make this a treaded, tank style robot. This decision has been questioned frequently during the design process. It has complicated the design significantly without adding any significant advantage over a wheeled design. We have, however, decided to stick with it. This design decision was made simply for the “cool” factor.

We spent a lot of time at the beginning of this project with Kano Analysis and going to Gemba for a reason. As engineers, we understand not only what this product will be required to do but also the nature of its intended use. This is not going to be a device to transport overweight people up stairs, even though that is one of its capabilities. The purpose of this robot is to look cool. It will be used as an advertising icon, and one of its first requirements is to look distinct. We feel that the treads are essential to the aesthetic approach to this project and have gone to great lengths to keep them in the design.

After examining what is available, we will either have to construct our own tracks from scratch or modify snowmobile tracks to do the job. In an effort to design with respect to the available tools, we have opted to use snowmobile tracks (pictured right), because they are very sturdy, reinforced, and we can buy drive sprockets designed to interface with them at the high speeds that we will be using.

This places a few constraints on the design. At first, we attempted to use a tread length used commonly by snowmobiles, but that would make the robot over five feet in length, which would limit its mobility to an unacceptable level. Even with cutting the tread, however, we must use an integer multiple of 2.52 inches of length, because the pitch of the sprockets and track is 2.52 inches. We will also need to design the wheels to interface properly with the inside of the tread.

The wheels themselves proved to be an unexpected cost. We expected that the cost of a few hard plastic wheels would be negligible, but we ended up having more requirements for them then we initially expected. They need to be able to support a lot of weight and be wide enough to hold the tread. Also, in order to machine out a groove to run the tread in, they can not be pneumatic or non-homogeneous. We found a lot of caster and idler wheels but most that were the size we wanted were around $20 each, and, for some designs under consideration, we need over twenty of them.

The tread design went through many iterations. The wheel diameters required ranged from four inches to fourteen inches, and the number of wheels varied widely as well. We have finally settled upon a design that requires six eight-inch wheels (right) two four-inch wheels, and four two-inch wheels.

Tank Tread Design

In designing the structure of the robot in its entirety, we iterated through many, many different possible solutions. An early consideration was the tread design. Through tinkering with the layout of the wheels and the general profile shape of the robot, we were able to get a good idea of the entire structure. Once we achieved a tread design we liked, we then considered where the through axles needed to exist in space and what constraints that would put on the internal structure of the chassis. With that taken into account, we evaluated the design, considered the limitations, and then went from there.

At the beginning of the first conceptualization of the tank tread system, we had a few initial constraints. First of all, the robot had to be able to mount a flight of stairs. I spent a few weeks with a tape measure in my pocket and measured every stair I came across. An average tall stair is eight inches tall, while the tallest stair I found anywhere was on a fire escape in an alley that measured ten inches tall. With this in mind, we have designed a front axle that is eight inches off the ground and at the very leading edge of the robot. This will allow the robot to easily mount any normal height stair, and, if it encounters a particularly high obstruction, the wheel will encounter the vertical surface and the nose of the robot will climb a short vertical distance in order to mount slightly taller objects.

Also, we want all of the ground wheels to be idlers. This serves several purposes. First of all, we will not have to worry about suspending the drive wheel. We also won’t have to worry about additional loads on the critical drive axle due to terrain and extreme driving conditions. Furthermore, this will facilitate the transport of the robot unpowered. In the event of a serious electrical failure where the robot must be moved by a limited number of people, its bulk could make carrying it very difficult or even dangerous. Instead, with all ground wheels as idlers, it will be possible for a single person to remove the track and simply roll the robot to its destination.

As a starting point, we have used a Model S90-A Japanese Ground Self-Defense Force Battle Tank (pictured above). We are not striving for historical accuracy, nor do we expect to maintain any level of correlation to this, we simply chose it as a good spring point for the design. From there, we have modified it to meet the requirements already outlined.

After tinkering with the idea for a while, we determined that the idler wheel that maintains the tension in a tank design (the fore or aft wheel) would not work for our layout. The front wheel must be fixed in order to provide the climbing effect discussed earlier, and the rear wheel is the drive wheel, so it cannot be free-floating. This means that we must add an additional wheel to maintain the tension on the track as the suspension compresses.

One of our first designs (pictured below) has several major flaws. First of all, the travel of the tension wheel is unacceptable. It adds far too much height to the chassis of the robot. Also, the travel of the suspension is too great. The travel must be less than the radius of the wheels, otherwise the chassis will impact the ground before the full range of the suspension is used.

Also at this point, we began to realize just how much wheels were probably going to cost. If we wanted to keep the three inches of suspension and not bottom out the frame, we would need at least eight-inch wheels, which were quite expensive. Then we considered making use of the junkyard for this particular component. We could find small fourteen-inch car rims with bearings for incredibly cheep in a junkyard; they would be easily strong enough to support the design loads, and we could fill the rim with liquid plastic to create the machineable surface necessary for running the treads. It would also provide a radius easily large enough for plenty of suspension and room for a more efficient tensioning wheel in the rear of the track. A preliminary sketch is included above.

This solution provided problems of its own, however. First, the steel car rims would add a significant amount of weight to the design. While this was not unacceptable, it was negative. The unacceptable part of this design was the size. With the completion of this design we did a quick reality check on the dimensions. A robot built with these specifications would be roughly the size of a couch. This is far too large for the mobility requirements of our customers.

At this point we realized that the only way to meet the requirements was to cut the track. Up until this point we had used a tread length of 121 inches, a common length of track for a small snowmobile. This simply made the robot too large.

Once we removed the track length constraint, we were able to reduce the length and overall size of the robot a great deal. We moved to three ground wheels, eight inches in diameter each. We spoke with our customers about the optimum size, and we ended up agreeing on some rough figures. We’re aiming to design a robot that is less than 30 inches wide, and approximately 42 inches long.

An early sketch of the new design is included below. It is quite rough and includes some scratch work used to determine the idler travel of the new layout, but it shows our first concepts of this smaller design. As you can see from the positions of the compressed and extended ground wheels, at this point we were still planning on using leaf springs. There is a horizontal displacement associated with compression that caused a lot of problems with this design.

As we transitioned to our next major design (pictured below), there were two significant changes. First of all, we moved the idler wheel to the front. This means we had to lower the front wheel to its final eight-inch high position, but due to the climbing effect discussed earlier, this should not be a problem. This re-positioning also allowed us to make the top surface level. The other major change is in suspension. Here we have moved to a completely vertical suspension, eliminating the drawbacks of leaf springs.

We made a few minor adjustments to the previous design and modeled the result in SolidWorks (pictured below). The line at the top of the drawing, dimensioned at 95.76 inches, is driven by an equation that sums the length of each segment of the track. This is exactly 38 pitches long. This shows that when the suspension is fully extended (there is no load on the wheels), the idler arm will be at an angle of 43.5 degrees from the horizontal. When the suspension is fully compressed, the idler’s arm will move to an angle of 77.6 degrees from horizontal.

These results are all very acceptable and were deemed worthy of “final design” status. Unfortunately, this was all changed when we discovered that the Optima 55 amp hour batteries previous destined to exist between the axles were found to be lacking. With the selection of the larger Concorde Sun Xtender, we were forced to do yet another redesign.

The redesign eliminated the back bottom wheel, and moved the drive wheels forward, changing the general shape of the track greatly. The new tread length calculation image is shown below. This new design has an idler arm travel from 54.25 degrees to 90 degrees, and the overall tread length has changed to 98.28 inches, or 39 pitches. This design (below) still provides the eight inch front wheel height as well as the same footprint shape developed in the previous iteration but facilitates batteries of nearly double the capacity, while adding only one inch to the robot’s height and a half an inch to its length. Furthermore, due to some streamlining done during the redesign process, the entire robot is three quarters of an inch narrower.

This process of iterating through tread designs has helped us greatly in designing other systems. The room available in the structure for components like motors, batteries, and drive train was greatly impacted by the positions of the axles. The axles needed to have a range of movement to accommodate the suspension, and that range constrains what volumes may be used to contain structure and components.

Suspension Design

As mentioned earlier, in order to select the tension elements of our suspension, we had to conduct some tests. Virtually every form of traditional suspension had been considered and deemed unacceptable for one reason or another, and so we were in a position to design and build a custom suspension from scratch. This is no easy task.

There were several requirements on the suspension. First of all, it had to take up as little space as possible. In order to keep the robot’s center of gravity as low as possible, the batteries are to be situated very low within the chassis; the same space the suspension elements will occupy. Also, they need to provide a rather specific tensioning force. The structure has been carefully designed to place the center two wheels one half inch lower than the other four. This is to reduce the friction in the “scrub zones” of turning, resulting in lower impact on the environment and better control.

However, in order for this to work, the suspension must be tuned. The elements should compress one half of an inch under roughly 90% of the robot’s unloaded weight. This way, the robot will sit flat and level on the ground (not teeter across the middle axle), but still be able turn about a very stable axis and not produce significant frictional forces on the environment.

These straps have the possibility to very effectively fill all of these requirements. They can be located at different points along the axle, reducing their impact on the surrounding component's placements, and they can be layered to adjust the downward force provided. However, because this is not the intended use of tie-down straps, there is no published data on the critical factors of the available materials.

In order to determine this data, we have conducted some simple experiments on various types of molded rubber straps to determine their capabilities. We tested three straps, pictured right. The top strap (blue) is polyurethane and is rated to stretch to twice its original length. The middle strap is molded rubber and is rated to stretch to 2.5 times its original length. The bottom strap is also molded rubber but is only rated to stretch to 1.5 times its original length.

Each strap was cut into short strips, and holes were drilled into their ends. These holes were stress concentrations and were, not surprisingly, the primary point of failure. This is exactly how we intend to mount them in the robot, however, so the results are perfectly valid. The drilled strips were each loaded into a tensile tester and stretched (pictured left) until they either broke or the maximum lower limit of the tensile tester was reached. We took readings periodically in order to determine the properties of each strap. The results are tabulated on the following page.

In addition to the spring rate and ultimate strength of these materials, some other properties are important to our application and must be determined. Possibly the most important of these is their tendency to creep. We performed two very simple experiments to get a feel for this. First, we took a four-inch strip of each material, suspended a forty-pound weight from each, and left them for six days. For the other experiment, we took another sample of each material, stretched them to a fixed length, and left them for the same six days. After the time was up, we returned and measured the unloaded length of each sample. The results of these experiments are also presented on the following page.

Of additional interest were the abrasion properties of each sample. This was not measured in any clear way, but at the conclusion of each experiment, the specimen’s surface where it was bolted was examined for abrasion damage. This also played a part in our final decision.

Based on the results of these experiments, we have selected the polyurethane straps (rated to 2 times their original length). They did not break in any test administered and had the minimum creep. They also provide good resistance to deformation, as well as some positive traits not tabulated here. They have superior resistance to abrasion, are easier to work with, and have better compression around mountings.

Tensioner

The tensioner assembly is the final component that works with the tread and the suspension in order to make the entire system work. As the suspension compresses or extends, the shape of the track changes to accommodate the uneven terrain, and something must take up the slack, or else the tread could fall off of the wheels. This is accomplished through a tensioning arm with 2 inch idler wheels on the end of it. This arm, as seen in the tread section, travels between 54.25 degrees and 90 degrees in order to maintain a constant tread length of 98.28 inches.

However, in order to do this, we need some input force to drive the tensioner. We briefly considered many of the spring types addressed in the suspension element selection section, but many of them required linkage solutions. The desired solution directs the input force against the axle of the tensioning idler wheels. This makes the tensioner arm a two-force member, with no required moment-supporting capability.

In order to achieve this, we have selected compressed nitrogen extension springs (pictured right), which are often found on car trunks and some doors. These come in a variety of extension forces and stroke lengths and can easily be mounted using their threaded ends. Now we simply must determine the desired extension force.

The first step in this calculation was to determine the tension in the tank tread as a function of the input force. This is done at the fully compressed and fully extended positions in static equilibrium. Simple free body diagrams and equations of static equilibrium give us the functionality we want, and we find that the minimum tension in the tread occurs when the track is fully compressed, giving us a tension of only 1.23 times the input tensioning force. These simple calculations are given on the next page.

Now that we know the track tension as a function of input force, we can determine what input force is required to hold the treads. We pick a minimum acceptable track tension of 50 pounds of force and then use the tension differential created by dynamic power transmission through the tread to back-solve for input force.

Each motor is capable of producing 1.14 inch pounds of torque per amp drawn, and the speed controller will allow up to 120 amp bursts of current. This, times the greatest gear reduction, divided by the moment arm created by the drive wheels gives us the maximum tension differential in the tread:

This means that the tension in the track approaching the drive wheels will be 213.2 pounds force greater than the tension in the track leaving the drive wheels. So, if the tension leaving is fifty pounds, the force on the high-tension side will be 263.2 pounds force, and the average tension (which should be the static tension), is equal to 156.6 pounds of force. This gives us the following:

T = 1.23 F = 156.6 lbf ( F = 127.3 lbf

The nitrogen springs are available in 25-pound increments, meaning that an input force of 150 pounds will ensure that the tension in the treads, even during bursts of high current, will never drop below 50 pounds.

Chassis Design

The chassis design followed logically from various points in the tread design. One of the earliest fully developed concepts is shown below. This design was based on several constraints. First of all, there are two mounting points for leaf springs. These are the most solid points of the structure, and the entire chassis is designed to rest on these. Also, we wanted to keep the design reasonably balanced, so we located two batteries ahead of the middle axle, with the motors and remaining battery aft. This design also includes a two-stage gear reduction. This configuration is the only way we could reasonably expect to fit in two stages of gearing, but we failed to take into account the space taken up by the rear leaf springs.

In order to make the springs fit, the motors had to be moved up, and the second stage of reduction no longer fit. It was at this point that we abandoned a two-stage reduction. We ran a few more iterations with a single stage reduction before throwing out the idea of using leaf springs entirely. It was becoming increasingly obvious just how sloppy they were making the design, and their elimination simplified the chassis structure a great deal.

Pictured below is a chassis that accommodates a single stage gear reduction, idler sprocket for tension, and vertical suspension. This is very close to the design that we ended up using in the first three dimensional models.

At this point, it was simply getting too difficult to draft our ideas in two dimensions. We drew several of our iterations in layered views, spanning multiple sheets of paper and carefully correlated and indexed to show the internal structure of the chassis. With a binder thick with notes and drawings, we began drafting in SolidWorks.

Shown here are some of the first three-dimensional drawings done of the robot. Shown first is the chassis structure with the motors, batteries, and wheels attached but with no top covering. This design has the motors in the middle and the rounded cylinder shapes of the Optima batteries. The same design can be seen below, from a different angle, top covering included. In these two drawings you can see some what was a reasonably good design. Unfortunately, all of this had to be thrown out when we realized that our battery capacity was extremely lacking.

It was at this point that our last major structure redesign began. The smaller Optima batteries were far to limited to provide the desired 30 minute battery life, so a new chassis had to be designed to support the larger Sun Xtender batteries.

With the previous design, the drive train, gearing, axles, and wheels all fit in as pictured below.

There was not much in the way of wasted space in this design, but we had to optimize where we could in order to fit in the larger components. In order to do this, we moved the motors to the back end of the robot. As can be seen in the previous design, there are several inches available between the motors and the batteries. We were unable to close this space because that would have put the second and third pairs of wheels too close together. With the motors in the back compartment, however, we were able to make use of this wasted space.

With that major revision in place, it was a simple matter to iterate through three or four designs before coming to one that we were happy with. The result (pictured below) has enough room to contain batteries of nearly twice the capacity and is only half an inch longer and one inch taller. Also, due to some other optimizations we performed while iterating through these designs, it is three quarters of an inch narrower.

For more on the iterative process, see Appendix B.

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

[pic]

[pic]

[pic]

[pic]

[pic]

[pic]

[pic]

[pic]

[pic]

[pic]

[pic]

[pic]

[pic]

[pic]

[pic]

[pic]

[pic]

[pic]

[pic]

[pic]

[pic]

[pic]

[pic]

[pic]

[pic]

[pic]

[pic]

[pic]

[pic]

[pic]

[pic]

[pic]

[pic]

[pic]

................
................

In order to avoid copyright disputes, this page is only a partial summary.

Google Online Preview   Download