zamboni robot project

15
Team Astana: Zamboni Robot Project i Team Astana: Zamboni Robot Project Andrew Burns, Victor Godoy-Cortés, Kelsey Newman, Cody Pederson, Westin Smith Professor Melde 30 October 2009 Engineering 102, University of Arizona

Upload: werdna-snrub

Post on 30-Mar-2016

222 views

Category:

Documents


0 download

DESCRIPTION

Design the lightest working NXT robot capable of picking up fifty randomly placed Lego pieces in under four minutes.

TRANSCRIPT

Page 1: Zamboni Robot Project

Team Astana: Zamboni Robot Project i

Team Astana: Zamboni Robot Project

Andrew Burns, Victor Godoy-Cortés,

Kelsey Newman, Cody Pederson, Westin Smith

Professor Melde

30 October 2009

Engineering 102, University of Arizona

Page 2: Zamboni Robot Project

Team Astana: Zamboni Robot Project ii

Executive Summary

The goal of this project was to design the lightest working NXT robot capable of picking

up fifty randomly placed Lego pieces in under four minutes. The design decided best fit to

achieve this goal was a remote controlled robot with a flexible scoop in the front to gather and

remove randomly scattered Lego pieces from a 36 inch by 36 inch walled enclosure. This design

was very effective. It picked up all of the Lego pieces in a time of 42 seconds while weighing a

total of 0.649 kilograms, giving an overall PI of 77.04.

Page 3: Zamboni Robot Project

Team Astana: Zamboni Robot Project iii

Table of Contents

Introduction ..................................................................................................................................... 1

Design Requirements ...................................................................................................................... 1

Design Description.......................................................................................................................... 2

Design Justification ......................................................................................................................... 4

System Model ................................................................................................................................. 6

Test Procedures ............................................................................................................................... 8

Test Results ................................................................................................................................... 10

Design Critique and Summary ...................................................................................................... 11

List of Figures/Tables

Figure 1 The rink with scattered pieces .......................................................................................... 1

Figure 2 Astana 100 from side ........................................................................................................ 3

Figure 3 Astana 100 from above ..................................................................................................... 4

Figure 4 Program flowchart ............................................................................................................ 7

Table 1 Breakdown of component weight. ..................................................................................... 3

Table 2 Light sensor false positive test ........................................................................................... 9

Table 3 Light sensor desk lamp test ................................................................................................ 8

Table 4 Final test run times ........................................................................................................... 10

References

The Zamboni Story. Retrieved October 29, 2009, from http://www.zamboni.com/story/story.html

Page 4: Zamboni Robot Project

Team Astana: Zamboni Robot Project iv

Rundell, R. W. (2003). High level of airborne ultrafine and fine particulate matter in indoor ice

arenas. Inhalation Toxicology, 15(3), 237-250. doi:10.1080/08958370390158256

Page 5: Zamboni Robot Project

Team Astana: Zamboni Robot Project 1

Figure 1 The rink with scattered pieces

Introduction

In 1949, Frank Zamboni patented a single operator ice resurfacing machine—the world’s

first self-propelled ice resurfacing machine (Zamboni). The process of cleaning the rink however,

involves in scraping a sharp, industrial sharp blade, across the ices surfaces while spraying a mist

of wash water (Zamboni). This complex operation is not easy, and a failure could be dangerous.

In recent years, many concerns have been raised about the air pollution Zambonis release as they

frequently resurface skating surfaces and its harmful effects on athletes (Rundell, 2003); and so

each new Zamboni model is subjected to countless tests to create a better, more efficient machine

that addresses these concerns (Zamboni). Similarly, the Robot Zamboni Project involves using a

Lego-brand robotic device, the NXT, to design, model and test a Zamboni-like robotic device to

solve the project’s problem in the most efficient way possible. This project is intended to give

students a feel for real world problem solving—through teamwork, problem solving, balancing

pros and cons, dealing with cost and technology constraints. Students should finish this problem

with a greater understanding of the rigors of the design process. Our team in particular, worked

on improving our time management and communication, two skills critical for employment in

engineering today.

Design Requirements

A rink, very similar to an ice rink, was

filled with 50 Lego pieces that must be “cleaned”

removed from the rink by the robotic device. The

success of the robot is measured by dividing the

number of pieces it successfully removed from

the rink in under four minutes by the weight of

Page 6: Zamboni Robot Project

Team Astana: Zamboni Robot Project 2

the robot in kilograms. This number was called the performance index, and the goal was to

achieve the maximum performance index possible. The project included a few explicit criteria

for/constraints on the design of the robot as well:

The robot must be able to drive itself through the “door” into the 36 inch x 36 inch

square rink.

The 50 Lego pieces must be randomly scattered throughout the ice rink.

The 50 Lego pieces vary in size and length.

No external Bluetooth enabled control allowed.

No destroying Lego pieces to create the device.

Additionally, other constraints were implicit in the project set-up. For example, as the robot was

controlled by the NXT brick, an item that weighed 0.106 kg, there was an upper limit on the

achievable performance index. Also, the environmental conditions such as lighting (the room

was not particularly bright) and sound (the room, especially during testing, was especially noisy)

limited the degree to which certain sensors could be used.

Design Description

The final design of our robot sought to minimize weight to the greatest possible extent.

Working off of the original thirty-minute robot provided at the start of the project, unnecessary

pieces were removed to reduce weight. Weight was a particularly important issue given the goal

of maximizing the performance index. Special intention was taken to the weight, and as shown in

Table 1, each individual piece of the robot was weighed to examine possible areas for weight

reduction.

Page 7: Zamboni Robot Project

Team Astana: Zamboni Robot Project 3

Figure 2 Astana 100 from side

part weight (lbs) weight (kg)

both motors 0.350 0.159

battery pack 0.234 0.106

brick (without battery) 0.328 0.149

both tires 0.0750 0.034

controller 0.0938 0.043

all cables 0.125 0.057

scooper arms 0.0563 0.026

back wheel and back wheel mounting 0.0375 0.017

remainder 0.125 0.057

single L-shaped piece 0.00625 0.003

TOTAL 1.425 0.649

Table 1 Breakdown of component weight.

As the goal purpose of the device was to remove Lego pieces from the floor of the rink, a

scooper, constructed of Lego pieces, was attached to the front of the robot to capture the Lego

pieces as it came into contact with them.

In our final design, the scooper arm consisted of

four pieces that formed an open cup shape, placing two of

the pieces at the bottom. The use of two pieces rather than

one allowed for minor flexing between the pieces, creating

a more adaptive scooper arm. The corners of the rink

allowed pieces to be trapped and difficult to extract, so

springs were added to the corners of the scooper arms to address this problem. When the arms

touched the wall, the cup shape pulled into a U-shaped and the corner pieces were flipped into

the scoop. The scooper arm was mounted so that it barely skimmed the ground. Special effort

was made to reduce the weight of the scoop to provide the maximum strength with the fewest

pieces. Each redesign of the scoop was tested for sturdiness by gently pushing the robot against a

wall until a piece dislodged, and using the and the effectiveness of the scoop was tested by

Page 8: Zamboni Robot Project

Team Astana: Zamboni Robot Project 4

running the robot through piles of Lego pieces and adjusting until it no longer allowed pieces to

slip behind the scoop.

The robot was propelled by two large frontal wheels attached to two NXT motors. A third

wheel, attached to a pivot, dragged behind the two front wheels to provide support for the back

end of the robot, and to improve the vehicle alignment and ensure it better moved in straight

lines. Unlike in the 30-minute robot, in our final design, the servo motors served as the

supporting structure. Many of the pieces from the NXT brick mounting and back wheel

mounting were removed, and virtually all pieces from the

underside of the robot were removed. The NXT brick was

placed with the bottom facing upwards at an angle. The

scooper arm was attached to an L-piece, which hooked onto

both the servo motors and the NXT-brick mounts, making it

rigid.

As the robot was remote controlled, the cables provided by the set proved too short to

comfortably control without tangles. Our solution involved using longer phone cables to the

NXT brick to its controller, and using an additional cable fabricated by connecting the cables for

the old Mindstorms motors. The shortest cables provided in the NXT set were used to connect

the motors to the NXT brick. The controller was built minimally, with the three sensors arranged

in a triangle.

Design Justification

The final design was selected upon recognizing the practical infeasibility of creating a

reliable autonomous robot within the limited time provided. Our original design, the product of a

lengthy brainstorming session, involved an autonomous robot controlled only by programming

Figure 3 Astana 100 from above

Page 9: Zamboni Robot Project

Team Astana: Zamboni Robot Project 5

and sensors. Our final product was remote controlled by touch and light tensors. In getting from

our original plan to our final design, our team divided responsibilities based on expertise and

resources. Cody, who not only had previous experience with the NXT system, but who also had

access to additional Lego pieces, was chosen to be both the team leader and lead designer, in

accordance with our team rules. Victor, who had access to the NXT software on his personal

computer, was placed in charge of programming the NXT brick, coordinating the design with

Cody. Kelsey and Westin were prototype testers and Andrew was charged with photographing

the prototypes and recording prototype performance to use in compiling this report and creating

the PowerPoint presentation. In general, the team roles were not entirely rigid, and most team

members gave input on robot and program design.

The final design was deemed far superior, both in its simplicity and efficiency, to various

other autonomous robot designs examined early in the design process. Autonomous robots were

too easily misdirected by slight variations in initial condition, and their programs required

complex iterative programming that required hours of careful adjustments to compensate for

various environmental variability, i.e. floor imperfections, unexpected frictional forces,

obstructive Lego pieces, etc. In examining the feasibility of autonomous programming, the robot

was programmed to turn 360 degrees one direction, move forward, and then move 360 degrees

the other direction. This experiment revealed the difficulties with working with autonomous

robotics. The most significant of these problems arose from the imprecision of the servos.

Initially, it was believed that even if the robot did not turn the same number degrees as the

program indicated, it would consistently turn the same number of degrees each turn and, after

each test, would end facing the same direction as it began; however, the robot never faced the

same direction at the end of this test no matter the number of degrees indicated by the program.

Page 10: Zamboni Robot Project

Team Astana: Zamboni Robot Project 6

The two servos provided with the NXT were both imprecise and inconsistent, and these two

qualities revealed the failures of the autonomous model. The user-controlled model, however,

offered greater flexibility and suffered far less from the variability in servo performance. The

final program provided for a complete range of motion: forward, backward, and laterally. Human

control added human problem solving skills otherwise unavailable to autonomous robots.

It should be noted that, though a few of the modifications used in the final robot appear to

contradict the project goal, they were deemed necessary for overall design performance. One

such modification is the way the robot's center of gravity is higher in the final design than in the

original. This higher center of gravity posed a risk of increased instability and roll-over; however,

the team decided that, because the robot did not run on 100 percent power, roll-over was unlikely.

To make sure, roll-over tests were performed whereby the robot was ran into the wall of the rink.

At no time did the robot appear unstable or at risk of toppling. The design had a higher center of

gravity due to the removal of unnecessary Lego pieces, and this design was deemed the best way

to reduce the weight of the robot as much as possible. It might seem contradictory then that we

decided to add additional weight with longer cords; however, after doing a cost-benefit analysis,

the team decided it was better to collect all the pieces quickly with slightly more weight than to

have the robot become tangled and fail the task.

System Model

In choosing a human-controlled, remote control system model, the most important aspect

of the system design was the program. The available sensors for the NXT system each carried

challenges as a control device. Several were considered and discarded. The light sensor was

originally considered as a means of guiding the robot. The light sensor would be affixed to the

front and a light shone from the direction of travel. This proved problematic because other

Page 11: Zamboni Robot Project

Team Astana: Zamboni Robot Project 7

sources of light interfered and the robot kept running even when no light was shone on it.

Eventually, the team settled on a remote control model using touch and light sensor. The touch

sensors would direct the robot to move left, right or straight, and the light sensor would instruct

the robot to move into reverse. Figure 4 shows a flow chart of the system operation. By reducing

the programming down to a set of iterative controls, the complexity was reduced and identifying

any system problems was greatly simplified. Careful adjustments were made to the program to

ensure that the robot responded at a reasonable speed. Early on, the robot responded too quickly

and the pieces were scattered across the rink, but by incrementally adjusting the speed,

eventually, a suitable speed was located and the robot performed quite adequately.

Figure 4 Program flowchart

In general, the programming of the robot worked on a trial and error basis. No one in our

group was greatly familiar with the intricacies of the NXT software, so this project was a

learning experience for everyone. Also, as mentioned earlier, the unreliability of the servos made

making adjustments touch-and-go—no one was ever quite sure what the eventual outcome of an

adjustment would be just by looking at the program. The program began simply, controlling only

Page 12: Zamboni Robot Project

Team Astana: Zamboni Robot Project 8

forward and reverse movements, and gradually, more and more intricacies were added into the

program by Victor, with feedback from Cody, so that all of the controls operated correctly and

the robot was as agile as possible. Through trial and error, it was determined that the scoop was

simply not large enough to capture all the pieces in one go. This meant that the robot needed a

reverse. The team decided against the bump solution to reverse because many members

expressed concern that, in the heat of competition, discriminating between a bump and push

would be overly cumbersome. The light sensor that we had once pooh-poohed as not useful was

made viable again as a mechanism to go in reverse. With this new reverse function, we were able

to avoid using complex algorithms or directions based on piece density to collect all of the pieces,

relying instead on a general counterclockwise move through the rink.

Test Procedures

In order to confirm that the light sensor would reliably tell if a finger was on it, multiple

tests were run. First, to determine when the light sensor would return a false positive, the robot

was placed on the floor and set to give light readings. This test was also repeated once with a low

Trials

Distance From Floor

(cm)

1 2 3 Average Intensity

1/√ℓ Equation-based

prediction

5 51 54 52 52.333

0.138 52.109

10 51 54 54 53.000

0.137 52.948

15 51 56 54 53.667

0.137 53.807

20 52 56 54 54.000

0.136 54.687

25 55 57 56 56.000

0.134 55.590

30 56 57 57 56.667

0.133 56.514

-800 THIS IS A DEMONSTRATIVE ESTIMATE 10.019

Table 2 Light sensor false positive test

Page 13: Zamboni Robot Project

Team Astana: Zamboni Robot Project 9

battery. In a second test, the light sensor was placed near a desk lamp and covered with a finger.

Readings from three trials for both tests appear in Table 2 and Table 3, respectively. We found

the data could be reasonably modeled by a linear regression of distance versus

, where l

represents the relative light intensity. The values for these models also appear in both tables.

Distance from Light

(cm) 1 2 3

Average Intensity

1/√ℓ Equation-

based prediction

Low

Battery

5 76 83 78 79.000 0.113 77.610 79

10 70 74 71 71.667 0.118 69.029 71

15 62 66 64 64.000 0.125 61.797 62

20 55 54 53 54.000 0.136 55.644 57

25 47 50 47 48.000 0.144 50.366 50

30 44 46 43 44.333 0.150 45.806 48

35 41 43 37 40.333 0.157 41.837 47

40 39 42 34 38.333 0.162 38.363 42

45 39 41 33 37.667 0.163 35.305 40

155 THIS IS A DEMONSTRATIVE ESTIMATE 9.829 Table 3 Light sensor desk lamp test

During the build phase, minor testing of how the scooper would react when striking a

wall was performed, both with the scooper attached and not attached to the robot. Although these

tests were brief, they were vital for observing how the elasticity of the scooper behaved and at

what angles the robot could strike a wall and recover.

Some testing was done with autonomous robots. This included a crude program aimed to

accomplish the design objectives, as well as a simple turning calibration program. The turning

calibration program would turn the robot (the amount was estimated to be one full turn), then

drive forward and turn the same amount in the opposite direction.

Before any testing on the actual robot, the program was briefly run in a free environment

to check that everything worked properly. This involved turning the program on and trying the

controls to see how the robot would respond. For performing full performance tests, the final

testing details were mirrored closely. The robot would start outside the rink, drive itself in, and

Page 14: Zamboni Robot Project

Team Astana: Zamboni Robot Project 10

take as many pieces out as it could within four minutes. For these tests, the designated robot

controller, Cody, was used. These tests were repeated multiple times.

Test Results

From our testing of the light sensor, it was determined that it would work accurately

when further than 1.5 meters from a desk lamp and within about 10 meters of an overhead

light. This represents a good range of working conditions, and during real-world tests the light

sensor worked in all but several dark areas.

Preliminary tests on the complete robot served primarily to fine-tune the

programming. From these tests, it was determined that the reverse speed, turning speed, and

light sensitivity should be fine-tuned. These tests also established the need for and dependence

on the length of the wires.

The final test runs went very well, and were very indicative of the final run. Table 3

reveals the tests returned times consistently under two minutes, and revealed no significant

problems in the design. Indeed, the final run went without problems and took only 42 seconds to

complete. The performance index for ever final trial run and for the final run was found to be

77.04.

Test Number

Time (s)

1 61.78

2 78.16

3 103.46

4 55.15

5 74.07

6 59.34

7 63.36

8 41.5

9 54.02

10 66.12 Table 4 Final test run times

Page 15: Zamboni Robot Project

Team Astana: Zamboni Robot Project 11

Design Critique and Summary

Retrospectively, we wish we would have shaved more weight off of our robot. While we

had considered using only one motor, we never figured out how it could be done. Seeing as one

team did it, we now wish we had been more innovative in our design. Apart from that, we cannot

think of anything else we would have done differently design-wise. The Astana 100 performed

spectacularly, snatching up all of the pieces in less than one minute. As for team dynamics, the

team wishes we had procrastinated less. Rushing to complete presentations and projects brings a

lot of stress. Otherwise, the team worked well together and everyone contributed equally.

Perhaps if we adhered to our team guide more completely, especially the sections on setting

meeting agendas beforehand and keeping detailed records of each meeting, the process would

have been less hectic. Indeed, it is our goal to more fully embrace our guide next project.