Jordan John (ME), Shohom Bose-Bandyopadhyay (ME.

)
Julie Mittelstedt (ME), Steven Davidoff (ME)
Noah Sitar (ID), Kate Blake (ECE)

Advanced
Heads Up Display
& Control
For Delivery
Vehicle
Dr. Wayne Li, Liz Kelley
ME/ID Collaborative Design Studio

April 31, 2023

Executive Summary
The state of modern technology is leaning towards automation. However, the future of
automation development is trending towards semi-autonomous functions – that is, autonomous
devices that are still “piloted” or directed by active users. As a result, the development and testing of
user interfaces to interact with said devices is paramount, and this project begins to tackle what is
likely the most contemporary area of development within this field – the development of
autonomous delivery drones by the company Amazon.
This project aims to develop an effective U.I for Amazon Delivery Service Providers (DSP’s)
to pilot semi-autonomous drones from within their vehicles. Two Rotary Knobs with
magnetorheological fluid were provided to conduct this control scheme and strongly encouraged to
be used. Although the project's main focus was said Interface, this team also elected to develop an
ergonomic mechanism via which the knobs can be accessed, as well as a software algorithm that
redefines optimal delivery route planning under this context.
This project's significance comes from the trends underpinning its ideation and the results
garnered from its execution. As stated before, the development of autonomous processing software
and robotics is gaining momentum rapidly, and its implementation into practical market
circumstances is therefore crucial for launching the technology forwards. This project looks to make
that transition towards the practical application of autonomous technology.
The key stakeholders in the development of this project were the group members
themselves, the faculty sponsor, the company sponsor of Amazon, and the primary demographics,
which include both the DSP’s and the individuals receiving packages from Amazon. Although the
interests of group members and sponsors are important for directing the project’s development, the
needs of the primary demographics outline where and how solutions should be implemented.
The value proposition of this project is as follows: For a ground delivery service that wants to
use drones to deliver in a friendly and efficient way, (this team’s) product is a user interface that
provides delivery drivers with a user-friendly way to deliver more packages per stop using drones.
Unlike current delivery user interfaces, we use haptic knobs as controls and a unique algorithm for
point finding to provide information in a human-friendly way about multiple locations of drop-offs
and the best location for the driver to park.
Proof of evidence in this project falls into three major categories: evidence supporting the U.I,
the ergonomics of the implemented mechanism, and the software optimizing delivery routes.
Ergonometric evidence involves user testing and anthropometric data, U.I testing involves user
testing and time/efficiency measurements, and testing for the delivery algorithm involves
simulations and formulaic demonstration.
The final design solution is two adjustable armrests to be installed to the sides of an Amazon
Van’s driver seat, each of which houses one of the knobs. The mechanism must be comfortable to use
and easy to store, in addition to fulfilling practical manufacturing and assembly standards. The User
Interface is to be displayed on the van's windshield using OLED technology. The U.I must also be
comfortable to use, rely on predicate platform standards provided by Amazon, and be efficient in use.
The delivery optimization algorithm must also be demonstrably capable, effective, and integratable
into the overall concept.

Table of Contents
Executive Summary ................................................................................................................................................................ 1

1
Table of Contents ..................................................................................................................................................................... 1
Introduction ............................................................................................................................................................................... 3
Predicate Technology: Existing Products, Systems, and Patents ......................................................................... 3
Regulatory Considerations .................................................................................................................................................. 5
Value Proposition .................................................................................................................................................................... 6
Design Process Overview ..................................................................................................................................................... 6
Stakeholder Engagement ................................................................................................................................................. 6
Mounting Concept Generation and Selection .......................................................................................................... 7
UI Concept Generation and Selection ......................................................................................................................... 8
Positioning Algorithm ....................................................................................................................................................... 8
Specifications .......................................................................................................................................................................... 10
Prototype Evaluation ........................................................................................................................................................... 11
GPS Simulation Time Calculations ............................................................................................................................. 11
UI Testing ............................................................................................................................................................................. 11
Cost Analysis ....................................................................................................................................................................... 12
Ergonomic Testing............................................................................................................................................................ 12
Final Design Specifications and Details ........................................................................................................................ 13
Realization and Deployment ............................................................................................................................................. 13
Design For Manufacturing and Assembly ............................................................................................................... 13
Conclusion ................................................................................................................................................................................ 14
References ................................................................................................................................................................................ 15
Appendices ............................................................................................................................................................................. 17

2
Introduction
Autonomous delivery services will disrupt the transportation industry as we know it today,
and the role of the delivery driver should change accordingly. This shift presents a design opportunity
to reimagine the driver's role in the context of the evolution of autonomous delivery vehicles.
Whereas previously, drone delivery carried packages from the warehouse to the destination which
uses the conventional transportation method of a delivery van until the last leg of delivery, where the
driver's role shifts to piloting a fleet of semi-autonomous drones. This approach increases efficiency
and enables utilization of emerging technologies in a humanistic way, intending to avoid displacing
many current delivery drivers, and incorporates new technologies while maintaining the driver’s
integral role in the process.

This project aims to design, demonstrate, and test a UI (user interface) for semi-autonomous
drone delivery with a fleet of up to five drones using two programmable haptic knobs and a
transparent OLED HD Display. To best utilize the group's skills, there was an additional extension of
the prompt to mount the knobs in a way that was comfortable and easily accessible in the context of
the current delivery Vehicle. After researching demographic needs, UI principles, and prototyping
actualized UI solutions, this project created a product that reduced the time needed for delivery.

Considering the market growth and the unfortunate work conditions drivers face, the need
for this project became clear. Today the implementation of drones in delivery lacks efficiency, with
one drone delivering at any given time. This project will increase efficiency by allowing one person
to control more drones to deliver more packages. The success of this project will address the problem
of poor work-life balance faced by delivery drivers everywhere.

Predicate Technology: Existing Products, Systems, and Patents

There are currently multiple examples of heads-up displays and user interfaces for drone
control. Heads-up displays have existed as physical screens inside vehicles and projections onto a
surface such as a windshield and date back to the 1940s in WWII fighter jets. Moving forward, General
Motors began plans to implement a HUD in their automobiles. They actualized it in their Oldsmobile
Cutlass in 1988 by using reflective optics to show the driver basic information, such as speed. BMW
followed suit in 2003, and the market has continued to grow. In 2019 the market was valued at $1.68
billion, with expected growth of 10.96% to $13.69 billion in 2027. The De Havilland DH.82B Queen
Bee Aircraft was the earliest credited modern drone, a British radio-controlled aircraft flown in 1935.
Over time, drone usage evolved from military to personal and industrial use. They were first
acknowledged as a viable tool for delivery services in 2013, including Amazon, UPS, FedEx, Google,
and Uber. The drone market and the user interfaces necessary to control them were valued at 20.8
billion in 2021, with expected growth of 57.5% to $501.4 billion in 2028.

While assessing needs relevant to the delivery market, communication was attempted with
people who currently work for Amazon, namely the drivers. This was done by exchanging contact
with drivers as they delivered, as well as visiting an existing Amazon hub location and asking to

3
exchange contact with someone there. After contacting these stakeholders, the overall conclusion was
that drivers are stretched thin and that the job as it exists today causes a toll on mental and physical
health. The stress tends to cause drivers to put the job before themselves to the point where according
to drivers, they will use a bottle in the vehicle before considering stopping to use a bathroom. This
conclusion is supported by existing literature. An article by Deanna Cuadra in 2022 details the stress
placed on drivers and mentions that delivery drivers were ranked 8 in the most dangerous US jobs,
with 28 fatal injuries per 100 workers.

As it stands, mobile delivery services utilize workers to deliver items on foot to each location.
The current technology is minimal, requiring drivers to exit the van at each stop and traverse across
streets, walkways, and stairs. This practice is time-consuming and cumbersome while it lasts the
extent of a workday. Amazon has tested and is always improving Artificial intelligence and machine
learning algorithms to aid in the delivery process. This includes weather, traffic pattern, and truck
capacity monitoring to optimize efficiency (Gebel). Amazon Flex drivers currently use a barcode
reader to scan items. Since 2021, Amazon has been installing AI camera monitoring drivers for safety
and behavior. With this, the system is linked to an app called "Mentor," which rates drivers based on
multiple factors, including driving speeds, braking power, and texting or making phone calls when
driving. The tiring of the system is to identify if a driver is being unsafe when operating the vehicle.
Additionally, drivers are required to download the app on personal devices, and many claim that the
app does "not turn off." Current technology also requires the delivery staff to take a photo of the
customers' items after placing them at their doorstep.

Amazon's direct competition includes UPS (United Parcel Service), FedEx, and companies of
similarity to DHL. On the other hand, Amazon has many indirect competitors, including companies
such as Walmart, Best Buy, and Costco. These companies are categorized as indirect competitors
given that purchases are not commonly delivered through these companies; however, items of similar
nature are sold through these companies, making them a threat to Amazon. Amazon has surpassed
its competitors by $232 billion in net sales annually (BigCommerce 2022). A further investigation of
E-Commerce revenue for Amazon and its competitors is depicted in Figure A.

Key marketing factors are strictly affected by company offerings. Companies that offer the
same products will pose a threat to Amazon. The list of these companies is immensely based on the
landscape of items Amazon promotes for sale. In terms of price, amazon tends to have a reputation
for retailing cheaper items due to wholesale storage in warehouses. Additionally, using flex drivers
and other companies like USPS and UPS have allowed Amazon to deliver packages in record time,
given various sources.

The proposed UI and delivery interface are designed to work in conjunction with the current
fleet of delivery vehicles and drones. Currently for drone delivery, Amazon uses the MK27 drone
which has a 5-lb max package weight, and an 89-lb maximum liftoff weight. The battery life is 20-25
minutes. Because of these constraints, the drones will be used to deliver light weight packages that
fit in the size of a shoebox or less, and heavier and bulkier packages will still be delivered on foot.

4
Additionally, Amazon has purchased a fleet of 3000 Rivian vans. While the specific specifications of
the vans were unavailable, the interior was assessed based on images shown in Figure B.

Regulatory Considerations
Federal Aviation Administration (FAA) regulations will greatly direct the design of a potential
UI. Anyone who flies a drone for commercial purposes requires some form of licensing. For licensing
purposes, the operator role is "the person performing remote control or … input expected at certain
states during normal operations" (NIST). This means that semi-autonomous drone operation is
subject to equally stringent regulation as its completely non-autonomous counterpart. The minimum
license, a Part 107 license, can usually be obtained with 15-20 hours of work (Dartdrones, 2017). A
Part 107 license allows the operator to pilot a single drone at a time that weighs up to 55 pounds
(National Archives). Other pertinent stipulations of this code state that an operator may not operate
drones from another moving vehicle and must be a visual observer of the drone's position throughout
the operation.

Package delivery by drone as well as piloting more than one drone at a time, requires a license
under part 135 of the Air Traffic and General Operating Rules (FAA). These are much more rigorous
to obtain because they extend to flying any commercial aircraft beyond drones. A part 135 Basic
license allows the operator to pilot up to 5 drones at a time and pilot drones that weigh more than 55
lbs. This license can be obtained with 3-4 weeks of dedicated work. A 135 basic license costs $4,200
(Air Carrier Compliance Group Inc.). This is a non-negligible investment that would need to be made
by either Amazon or their independent contractors. The system needs to increase efficiency to an
extent that makes this investment cost-effective. Flying more than 5 drones at a time, which is
required for piloting a fleet of 20 drones, requires a Part 135 Standard license, which is a full
commercial pilot's license. In addition to 150-200 hours of anticipated work, the licensing fee for this
is $6,500. This requires a substantial investment, which, when the average delivery rate is only 40
packages per hour, is unrealistic to assume the majority of contractors will adopt.

There are some restrictions on where and when drones can fly, and in addition to these
national regulations, drone flight is subject to state and local ordinances. Most notably, these
regulations restrict flights over national parks, near stadiums, and around airports. Drones may not
fly over national parks and may not fly within three nautical miles of FAA-designated stadiums from
1 hour before to 1 hour after a sporting event (B4UFLY). Additionally, flying within a 5-mile radius of
airports requires authorization. The degree of authorization required varies based on the proximity
to the airport and the other planned flight activity (FAA). Generally, automated approval is available
for drone flights up to 200 ft; however, for some areas in the immediate radius of airports, direct
approval is required for any drone flights. Airport regulations present an additional challenge to
implementation as large swaths of cities, such as Atlanta, are subject to them, as seen in Figure C.

The implementation of the solution additionally needs to fully conform to the Federal Motor
Vehicle Safety Standards and Regulations and should additionally meet the SAE Ground Vehicle
Standards (National Archives). The implemented HUD must still allow for complete visibility as a

5
windshield. Standard 103 requires that each windshield has a defrosting and defogging system
covering the critical area, as shown in Figure D. Standard 104 requires vehicles to have a powered
windshield wiping and washing system. Additionally, the implementation needs to protect driver
safety during a crash. Standard 212 details windshield retention requirements and states that in a
forward collision up to 48 km/h (30 mph), the windshield must retain 50 or 75 percent of the outside
of the windshield corresponding to the use or lack of use of passive restraints respectively.

Value Proposition
For a ground delivery service that wants to use drones to deliver in a friendly and efficient
way, FWGSTW's product is a user interface that provides delivery drivers with a user-friendly way to
deliver more packages per stop using drones. Unlike current delivery user interfaces, we use haptic
knobs as controls and a unique algorithm for point finding to provide information in a human-friendly
way about multiple locations of drop-offs and the best location for the driver to park.

Design Process Overview

Stakeholder Engagement
This project had three main groups of stakeholders whose interests needed to be considered.
The first group, who also recommend the vulnerable group, are the Amazon Delivery Service
Providers (DSPs) in addition to reducing the physical demands and long hours of their day. The
second category of interest is that of Amazon and its customers, primarily interested in reducing time
and costs. The third stakeholder is the HMI-Lab which wants a user interface to show at a conference
that can be controlled using two haptic knobs.

The two primary demographics whose needs are being explored are the consumers, those
receiving deliveries from Amazon, and the delivery drivers. Consumer demographics are widely
available online; Amazon consumers tend to be male, between the ages of 19 and 44 (with an average
of 37), have middle-class socioeconomic status, and are likely to be subscribed to Amazon Prime
delivery service (65% of consumers are subscribed) (Radic, 2022). Interviews and surveys with this
demographic demonstrate that many order more than 10 packages per month, and approximately
80% would be willing to interact with and distribute signatures directly to drone interfaces (see
figures E, F, and G).

Amazon Delivery Service Providers (DSPs) were a much more challenging demographic to
collect information about. There exist no sources of statistical data surrounding their age, sex,
socioeconomic status, etc., likely by the intention of protecting their privacy rights at Amazon
themselves. However, from online forums, we can deduce that delivery drivers generally feel stressed
and heavily encumbered by their highly demanding daily workloads (to the extent that they'll often
elect to use bottles rather than proper restrooms in order to save time) and are wary of injuries
occurring due to said frantic nature of the job. Additionally, Amazon provides some insurance
coverage for their DSPs, but due to the third-party nature of their employment, Amazon typically

6
avoids accepting liability for these injuries, leaving DSPs to either purchase independent insurance
or run a higher risk of impairment as a result of injury (Scalfane 2022). One further insight, collected
from three in-person interviews, was that DSPs could not afford to talk to consumers much while out
on delivery but otherwise would like to, and each driver recounted several fond stories of social
interaction that they encountered while en route.

Based on initial stakeholder engagements, it was concluded that amongst others the main
priorities of this project were to save time, save money, and create a user experience that was
comfortable and safe. After using the House of Quality model, shown in Figure 10, it was additionally
revealed that there were synergistic relationships between minimizing hidden components, reducing
the number of parts, maximizing strength, and reducing manufacturing costs.

Mounting Concept Generation and Selection

For initial prototyping for the physical mounting, there were three areas of the inside of the
Rivian Van where there was space to mount the knobs: the center console, the floor next to the
driver's seat, and around the steering wheel. The steering wheel mounts, shown in Figures H, I, and J
were abandoned because it was determined that they would interfere with driving. The center
console idea, shown in Figure K, was abandoned because it was an uncomfortable location from
which to operate the knobs. At this phase of prototyping, it was concluded that the best solution
would be retractable, such as in Figure L, so that it could be deployed when in use and put away when
not needed.

The second phase of prototyping narrowed in on retractable solutions with arm support that
could be operated in a seated position. The prototypes and their corresponding qualitative feedback
are shown in Table M. The first option represents armrests which would be added to the left and right
of the driver's seat, where the knobs are placed at the end and can flip up when needed and flip down
when not in use. A big piece of positive feedback about the prototype was that there was a negative
wrist angle because the knobs sloped downwards. While this was unintended in this prototype, a
similar feature was added on purpose to the final model. The second option featured armrests that
can flip up and down to allow for ease of entry. Users found this height to be the most comfortable
overall, and most people said they didn't mind the motion of deploying and retracting the arms.
However, numerous people mentioned that they would have preferred to work with their arms
slanted inwards. The third option, focused on more of a table concept than armrests, featured a two-
axis hinge, allowing the table to flip up and down. While people appreciated the potential for versatile
knob placements, this design had other drawbacks. This model was the most sensitive to different
heights, where people found it too tall or too short. Several people also found it to be claustrophobic
and hard to maneuver. Prior to testing, it was anticipated that this solution would be fully realized,
and the corresponding CAD model assembly is shown in Figure N.

The ergonomics relevant to this project are relatively straightforward. The ergonomics
surrounding seated operations of dashboards are heavily researched, particularly for motor vehicles,
to ensure consumer comfort and operation. Several handbooks exist that contain compiled

7
anthropometric data for limb lengths, including the all-steel reference guide for anthropometric
design, published by Scott Openshaw at Oregon State University. In conjunction with user testing, we
used their outlines to define parameters and quantify dimensions for the mounting mechanism of the
two haptic knobs. The dimensions relevant to these ergonomic considerations is shown in Figures O,
P, Q.

UI Concept Generation and Selection

The first round of drafting for the User Interface involved looking at pre-existing strategies
for tackling positioning, direction, and selection within UI. An exploration of several predicate
examples indicated that Bird's Eye views of maps were the most efficient for communicating
positioning and direction information. Operating under the assumption that the majority of the
process is designed to be automated, we required only a single input for selection – launching the
drone navigation protocol itself. Examples of these two interface layouts can be found in Figure R.

The second round of drafting involved a variety of changes. Initial user testing and need-
finding indicated that more selection was necessary to determine logistic factors. Additional
exploration and research uncovered a series of design standards and platform requisites for Amazon
software. These requisites also provided a layout for how information should be organized, and
selection be conducted. These modifications and an overhaul of the design language were introduced
to align with the associations and perceptions predicated by other Amazon UIs. Examples of these
modifications can be found in Figure S.

User testing for the second draft of the UI involved monitoring the speed at which individuals
could select and navigate through the interface, as well as direct interviews over specific pain points.
It was determined that requiring input to launch the selection process was unnecessary and took
more time than necessary to the process, so the series of flows was removed. The selection screen
was also modified to allow independent scrolling between options and selection of said options via
one of each of the knobs, respectively. This change was shown to improve selection time. Finally, a
series of small aesthetic changes were introduced to improve the appearance and "comfort of use" of
the final design draft. Although the sum total of these three changes was small, the additional layer of
polish they provided helped improve the UI's efficiency of use and the users' emotional response.
Although the User Testing procedure will be elaborated upon in greater detail later, please find an
overview of the final U.I design and Flows in Figure T.

Following the construction of this U.I draft, two animations illustrating the intended sequence
for error messages and for the “marching ant line” tracking of the drone’s locations were developed.
Please find stills illustrating those animations in Figures U and V.

Positioning Algorithm
The UI will position the van in specific locations minimizing the number of stops and
decreasing the stopping time. The locations are governed by an algorithm which takes in the GPS
coordinates of all the stops, weights of packages, and dimensions of packages. The algorithm groups

8
together each individual street and sorts the street addresses in order. Drones are assigned in groups
of five per stop shall it meet the conditions of lifting a package of no greater than five pounds, and is
10 x 14 x 5 inches in the X,Y,Z dimension respectively. The algorithm then plots the street using the
line function, as well as it plots the van stopping locations, the locations of drone delivery, and the
deliveries on foot. The algorithm conditionals are shown in Figure Z followed by simulation data in
table 1 along with the resulting plot of the map, Figure 2.

9
Specifications
Requirements Source
Functionality
Assist delivery drivers in delivering packages Project Specs
Stop Optimization: Stop every 5 drone deliveries FAA Regulations
Geometry
Footprint Space Requirement: <1 ft2 Team
Time
Average U.I operation time (specific instruction): 0:20 > Team; ANSI
Average operation time for the U.I (before testing): 1:30> Team; ANSI
Energy
Current draw: < 1 A Hapticore Knob
Signals
Inputs: Knob velocity, Position, Torque, Clicks Project Specs
Physical Parameters (Bushing, Bore Diameter, etc.): Predetermined by Hapticore Knob
the provided knob.
Knob Sensitivity: Inputs measured at 22.5-degree intervals (1/16th of a GlobalSpec
full rotation).
Dead Zones: One 22.5-degree rotation to be completed within a 50 to Tool Module: The
150 ms time frame. Hand
Outputs: Driver location (GPS Coordinates), Package Locations (GPS Project Specs
Coordinates), Drone information (battery life, altitude, speed, trouble
light)
Display: OLED Project Specs
Safety
Lock knobs when not in use Team
Drone Speed <= 10 mph Team
Drone Height: 12 ft height delivery, 25 ft height cruising Amazon Practices
Ergonomics

10
 10" clearance between seat and table Grand Rapids Chair
Co (Reynolds)
Arm Length: 22.4” - 27.9” Anthropometry
Summary Table
Seated Eye Height: 30-33 cm; 15 degrees above parallel Anthropometry
Summary Table
Materials
High Strength Carbon Fiber Granta Materials
Selection

Prototype Evaluation
Overall, the project initially had three main goals. The first goal was to reduce the time needed
to deliver packages; the second goal was to create a User Interface that enabled drivers to interact
and manage a fleet of five drones at a time, and the third was to create a comfortable experience for
drivers. This project achieved a degree of success on all three of these fronts.

GPS Simulation Time Calculations

Given there are two mutually exclusive outcomes, a package is delivered either with or
without the use of a drone, binomial distribution is applied. According to Forbes, Bezos reports 86%
of Amazons packages weight five pounds or less warranting drone delivery (Guglielmo, 2013).
Illustrated in EQUATION 3, the binomial distribution formula evaluates the likelihood of drone
delivery. According to FAA standards, five drones are to fly at a time, making a total of 55 groups,
cycles of drones to be sent out daily. According to the binomial distribution computation in table 4,
drones will confidently deliver at least 215 packages per day.

In addition to relieving the driver, drones will deliver packages in record time. The drone has been
selected to operate at a horizontal speed of 10mph maximum. Medical News Today claims the average
human walks 3mph (Fletcher 2022). This speed ratio suggests a delivery time decrease to 3.33 times
the current delivery time. To further illustrate the impact, table 5 illustrates the time difference for
two general extremes of delivery: a short distance delivery and a long-distance delivery.

UI Testing
Testing for the U.I involved two rounds. The first round of testing was comprised of casual
interviews with users and Amazon DSP’s as they observed its operation and toyed with its functions.
Although little quantitative data was collected, this procedure revealed several pain points in the
design’s previous execution, aesthetics, and flows.

11
 The second round of testing involved three steps. First, the users were introduced to the U.I
and allowed to navigate/explore it freely to determine all of its functions. The time required for them
to feel satisfied with its operation was secretly recorded and summarized in a table. Next, users were
requested to follow the flows of the U.I, make a specific series of selections, and to do so as efficiently
as possible. These times were also recorded and summarized in the same table. Finally, a brief
interview was conducted where users were asked to come up with any pain points, they experienced
during their time operating the U.I. Their comments were also briefly summarized in the same table.
You can find the said table in Figure 6.

Cost Analysis
There are multiple solutions to a HUD (Heads up Display) that can be linked with software providing
an interactive experience to the user. The solutions range from a $300 windshield projector to a
$25,000 OLED 1080p transparent screen. The average Amazon driver makes $14.83 per hour. Table
7 illustrates the cost savings if drivers had the ability to reduce their hours. The main takeaway is
workers will not need to put in as much time each day into deliveries; however, efforts can be focused
on making more deliveries and working the same hours, which is a profit to the company.
Additionally, Amazon’s savings on drone delivery are depicted in Figure 8.

Ergonomic Testing
Based on initial user testing data and qualitative recommendations, a final hybrid design is
created that incorporates lessons learned from the successes and failures of initial designs. This
prototype allowed for an adjustable wrist angle of the knobs to allow for either a neutral or negative
wrist angle preventing strain.

Initial anthropometric data was gathered from a study conducted by the University of North
Carolina in order to take educated guesses at the correct ergonomic proportions. However, the first
two prototypes (which were tested simultaneously) revealed that these proportions were slightly off,
and that users had a variety of pain points and qualitative feedback notes (which you can find in
Figure O). These pain points were explored and elaborated upon in a second round of ideation and
prototyping.

During the second round of prototyping, we began by cross referencing anthropometric data
from the North Caroling study with a second study from the University of Ohio. This data was used
to create a range within two standard deviations of the populations average for men and women.
Following this, a series of sketches were developed to illustrate the overall view of the prototype, in
addition to some of its features and ergonomic measurements (see Figure P and Q). Following
construction of the final prototype, detailed in Figure LAST, a second round of user feedback was
gathered from our final prototype: please see Table 9 for more details. It is also theoretically
compared to the earlier iterations in an evaluation matrix in Figure 11.

12
Final Design Specifications and Details
The primary subsystems developed over this project were the user interface, control knobs,
positioning algorithm, and the physical mount for the knobs. The most obvious and straightforward
piece of integration is between the physical mount and the knobs. If the mount (shown in Figure Y)
is dimensioned correctly and space is left for the cords, there should be no integration problems,
which was the case. Integration between the knobs and the user interface was a challenge, however
due to lack of documentation, it was difficult to receive and accurately interpret inputs from the knob.
The primary task was making the inputs from the knob control the flow of the UI by mapping the
knob to keyboard presses in an Adobe XD UI. The last obvious form of integration was using the
algorithm to instruct the UI and driver. The algorithm is meant to show the best place for the driver
to park and which packages can be flown as opposed to being delivered on foot. Since the user
interface is communicating with the driver, the algorithm must be in the backend to optimize
deliveries. The systems in this project do not directly affect each other necessarily, but there is a flow
where a broken piece affects overall function. The mounting of the knobs must be ergonomic and
user friendly, so tradeoffs such as comfort must be considered. This affects the use of the knobs which
directly affects how the driver can operate the UI. The UI must be human friendly and have an
intuitive flow as a confused driver will decrease efficiency regardless the controls operation. Lastly,
the algorithm is the main driver in optimizing the deliveries in real time. If it is faulty, then the
operation of the other systems would be for naught.

Realization and Deployment

Amazon is currently in a partnership with Rivian, which plans to produce a fleet of
approximately 100,000 delivery trucks by the year 2030 (Baldwin 2020). Therefore, there are two
options for implementing the system into the Amazon workforce: only modify existing vans once they
have been produced or work with Rivian and insert modifications to the vans that will be
manufactured later. The former is difficult, as a lot of the modification is required to disassemble the
truck somewhat; an efficient way to do that in mass is absent. However, the latter involves
interrupting the manufacturing flow of Rivian to insert the design, which could be time-consuming
and potentially interrupt the jobs of those on the assembly line, requiring overtime, which takes away
from the workers' lives. Further communication with both Amazon and Rivian is required before a
conclusive answer can be given.

Design For Manufacturing and Assembly

A way to manufacture the design would be to use extruded stainless for the structural and
prismoid parts and use injection molded plastic on the armrests and other irregular parts Stainless
steel is the material with the lowest cost which would allow the armrest to fully support the full
weight of someone in the 95 percentile and then half again that much (see Figure W for Granta
report). An extruded profile would allow for wires to be threaded through cavities in the steel and

13
would be easily sourced. There are likely similar armrests already available for a lower cost that
amazon could integrate into the final design as well.

To produce 100,000 of these, bar stock will be used so cost estimates can be pulled from a
steel manufacturer. See Figure X for the cost estimates for each extruded part, each unit (excluding
electronics, see cost analysis in UI prototype evaluation section for technology cost analysis), and all
the units combined, which come out to be about 14.6 million dollars. Keep in mind that this doesn’t
account for the cost of integrating with Amazon’s fleet or Rivian's manufacturing. The cost of the
physical unit excluding the electronic

Conclusion
The project overall served as an opportunity to advance the delivery service system. As
drones become more of a significant part of industry improvement, jobs will shift from menial tasks
to drone piloting and monitoring. As it’s far too early, inefficient, and impractical to only use drones
for delivery, we instead advanced the driver’s role to be one of a drone pilot as well as a delivery driver,
though our system ensures that the delivery part will be much less demanding. The UI/UX interface
helps the driver monitor, and in some cases fly, drones delivering light packages. This, along with an
algorithm developed since the situation of a dual driver/drone delivery is novel, significantly reduced
the time it took to deliver packages in a suburban environment. The chair we designed was highly
rated by user testing for comfort, and designed with ergonomic focuses in mind, such as carpal tunnel
syndrome, and joint strain caused by height. Though this is just a first iteration for these concepts,
this provides a solid baseline for future endeavors to build upon.

14
 References
Air Carrier Compliance Group Inc. (n.d.). Service descriptions and fees. ACCG - Services and Fees.
Retrieved February 5, 2023, from https://www.air-compliance.com/prices.php
“Anthropometric Data Men Women - Ohio.” Anthropometric Data for Men and Women, University of
Ohio, https://www.bwc.ohio.gov/downloads/blankpdf/ErgoAnthropometricData.pdf.

B4UFLY App | Federal Aviation Administration. (2022). Faa.gov.

https://www.faa.gov/uas/getting_started/b4ufly
Baldwin, R. (2020, October 8). Amazon Shows Prototype of Its Rivian-Built Electric Delivery Van.
Car and Driver. https://www.caranddriver.com/news/a34315562/amazon-rivian-electric-
van-prototype-revealed/
Control Knobs Specifications." GlobalSpec,
https://www.globalspec.com/specsearch/searchform/mechanical_components/industrial_
hardware/control_knobs.

Dartdrones. (2017). Dartdrones: The Nation's leader in drone training. Part 107 Study Guide.
Retrieved February 5, 2023, from https://www.dartdrones.com/wp-
content/uploads/2018/01/Part107_StudyGuide.pdf
Fletcher, J. (2022, November 16). Average walking speed: Comparisons by age, sex, and walking for
health. Www.medicalnewstoday.com.
https://www.medicalnewstoday.com/articles/average-walking-speed

Flying Near Airports | Federal Aviation Administration. (2020). Faa.gov.

https://www.faa.gov/uas/getting_started/where_can_i_fly/airspace_restrictions/flying_nea
r_airports
Fracis, Taylor. "Anthropometric Data for US Adults (All Dimensions in Inches)." Anthropometry-
Table-Summary, University of North Carolina,
https://www.ergocenter.ncsu.edu/wp-
content/uploads/sites/18/2020/07/Anthropometry-Summary-Table-2020.pdf

Gebel, M. (n.d.). I’m a 55-year-old Amazon driver. I risk rolled ankles, blown knees, and dog bites daily
— but I still enjoy the job. Business Insider. Retrieved February 1, 2023, from
https://www.businessinsider.com/amazon-delivery-driver-day-in-the-life-2020-10
Gebel, M. (n.d.). I’m an Amazon delivery driver who’s had to pee in water bottles and eat lunch in my
van. I hate the new surveillance cameras and feel like I’m always being watched. Business
Insider. https://www.businessinsider.com/amazon-delivery-driver-warehouse-employee-
working-conditions-van-cameras-2021-3
Guglielmo, C. (n.d.). “Turns Out Amazon, Touting Drone Delivery, Does Sell Lots of Products That
Weigh Less Than 5 Pounds.” Forbes. Retrieved April 30, 2023, from
https://www.forbes.com/sites/connieguglielmo/2013/12/02/turns-out-amazon-touting-
15
 drone-delivery-does-sell-lots-of-products-that-weigh-less-than-5-
pounds/?sh=41b5d698455e

National Archives. (2023). The Federal Register. Code of Federal Regulations. Retrieved February 4,
2023, from https://www.ecfr.gov/current/title-14/chapter-I/subchapter-F
National Archives. (n.d.). The Federal Register. Code of Federal Regulations. Retrieved February 5,
2023, from https://www.ecfr.gov/current/title-49/subtitle-B/chapter-V/part-571/subpart-
B?toc=1
National Institute of Standards and Technology. (H.-M. Huang, Ed.), Volume 1: Terminology
Autonomy levels for unmanned systems (ALFUS) framework (n.d.). National Institute of
Standards and Technology.
Package Delivery by Drone (Part 135) | Federal Aviation Administration. (2017). Faa.gov.
https://www.faa.gov/uas/advanced_operations/package_delivery_drone
R/amazondspdrivers. reddit. (n.d.). Retrieved February 3, 2023, from
https://www.reddit.com/r/AmazonDSPDrivers/
Reynolds, A. (n.d.). Things designers know: Restaurant chair height, table and chair distance & more.
Grand Rapids Chair Blog. Retrieved March 20, 2023, from
https://blog.grandrapidschair.com/things-designers-know-restaurant-chair-height-table-
and-chair-distance-
more#:~:text=How%20much%20space%20between%20your,the%20underside%20of%2
0the%20table.

Sclafane, S. (2022, July 13). How Amazon thinks about auto insurance for Delivery Service Partners.
Insurance Journal. Retrieved February 3, 2023, from
https://www.insurancejournal.com/news/national/2022/07/14/675735.htm#:~:text=Am
azon%20runs%20an%20auto%20insurance,Amazon%20is%20fine%20with%20that.
"Tool Module: the Hand." Tool Module: The Hand,
https://thebrain.mcgill.ca/flash/capsules/outil_bleu15.html.

Top 9 Amazon competitors and why they are successful (2022). The BigCommerce Blog. (2022,
December 19). Retrieved February 4, 2023, from
https://www.bigcommerce.com/blog/amazon-competitors/

16
Appendices

Figure A. Amazon’s Competition Landscape, Surpassing Competitors 1

1 Top 9 Amazon competitors and why they are successful (2022)

17
Figure B: Pictures of the inside of a Rivian Van

18
Figure C: Screenshot from B4UFly app showing areas in Atlanta which are subject to flight
regulations. The circles represent radii around airports where flight requires approval.

19
Figure D: Critical Area of a windshield (CVA) that must have a defrosting or defogging system

Figure E: A Consumer survey reporting their preference for interaction/signature distribution.

20
Figure F: A survey of consumers reporting how many packages they order per month.

Figure G: A Survey of consumers reporting their willingness to receive drone deliveries.

21
Figure H: CAD model of Steering Wheel Mount

Figure I: Steering Wheel Mounted Prototype

22
Figure J: Alternative Steering Wheel Mounted Prototype

23
Figure K: Console Mount

24
Figure L: Initial Prototype of a retractable knob mount

25
 Table M. Qualatative User Testing Data for mounting prototypes

Flip down armrests Flip-up armrests Airplane Table

Too low. Feel like you have It’s a perfect height where the Knobs in the middle feel better
to slouch. elbow rests If there was some than knobs at the ends
way that the knobs could come
Want the knobs to be flat or It feels like a desk more; it feels
together based on people’s
maybe more of a downward- like you want to put stuff on it
preferences
facing angle
Very close to it; never sit this
Would have the entire armrest
The prototype felt flimsy, but close to a desk, so there may not
angle in instead of a part of it
that wasn’t. be enough space for a desk-type
Want the forearm supported. thing in a car
It would be helpful for it to
be able to slide forward for Moving it is kind of Clunky
longer arms because it is larger than the
other 2. It’s not great for
If the chair can move relative
maneuverability.
to the armrests, can you
adjust yourself relative to it – The space could also be
yeah, probably obnoxious for a larger person
6’3”
Liked the angle and found it Want full arm support Just sitting at the table isn’t
comfortable because it is comfortable either inwards or
Your wrist is flat so that the
only for drone operation outwards
knobs are angled slightly lower
instead of 24/7
5’7”
Forearm-to-wrist support is
The hands can be rested.
important.
Ergonomic mice are
sideways, and it feels like a
sideways ergonomic mouse

26
Because the height isn’t too
high, you aren’t resting your
arms as much as setting your
hands down

The downward is good Many cars have systems like an Would there be other rests there
airplane/trucks
It’s a little too high to rest
Since this kind of mechanism comfortably anywhere; it’s just
exists, it feels familiar to use too high. The height is not in a
place where you can rest
If it was lower, like the height of
a table, it might be better
If it was curved and had an
opening in the center. Like if it
was further
The distance from the chair may
be impacting that as well
The wrist goes below the neutral
position
It was fine to put it away from
where it was
5’8”
The height is a little low Normal armrests and angled 5’5”
again knobs feel comfortable. By
If it were like 5 or 6 inches
tilting the knobs outward and
It would be better if the lower, it wouldn’t be bad
down
angle were a little better
If it were the height of the actual
Lifting the armrest is too far
The tilt of the knob is fine. tabletop, that would be good
back; it’s not that different
Unless the knob itself is
from a regular armrest, but the It could be hard to maneuver
angled, angling the knob
length makes it easy to
down would make it out of
overextend your shoulder
sorts
Slightly to the side would be
optimal
Would like it if it was higher What if it was half an armrest In planes
and slightly more angled when it dropped down halfway
You are only putting your wrist
down through
down regardless of the height
Does it feel intuitive to
The table circumvents the height
activate and deactivate
problem because you don’t have
to maneuver anything,

27
 It’s easy to maneuver
Like the ability to move arms
around because your arm will
compensate for the angle
“I like it.”
This one is comfortable the
other one is more functional
In the middle one, you are
resting your arms, but the one
on the left isn’t high enough
If there is something that is
already there, it would be fine,
but then there aren’t the
armrests for driving
I do drive at the bottom of the
wheel and use the armrests to
rest on
The process of lifting them is
fine
Pushed them to lift them
It’s slightly too high can feel it
is pushing shoulders up even if
it's slight
Holding the buttons from the
side is nice, but if you’re
holding from above, it’s a little
awkward. Like the arm rests
better than the table
It’s not too hard to lift up, but if
you were doing it every 5
minutes, it could be tired

28
Figure N: Flip Table Assembly and BOM

29
Figure O: Qualitative Feedback provided from the first
round of ergonomic user testing.

30
Figure P: An Anthropometric overview of the individual in question for
user testing and prototyping purposes.

Figure Q: Sketches illustrating the features and overall design

criteria of the second prototype.

31
Figure R: The first draft of the UI, involving a simple Bird's Eye View of the map
and a singular launch input, with little selection capacity.

32
Figure S: The second draft of the UI, involving design language and standards
predicated by existing Amazon User Interface designs.

33
 Figure T: The Final U.I Design flows. Please find a link to explore the actual
operation of the above U.I here: https://xd.adobe.com/view/57399d88-3131-43d2-a992-
8f7d452dcf9a-026a/?fullscreen

Figure U: The intended error warning sequence for the U.I.

34
Figure V: The intended “marching ant line” animations for drone locations.

Figure W: Granta Materials Analysis

35
Figure X: Extruded Production cost analysis

Figure Y: Angled Arm CAD Model

36
Figure Z: Conditionals for Drone Optimization Code

37
Table 1: Sample Data for Simulation

38
 Figure 2: Plotted Map – Results from Data

Equation 3.

𝑛 Where:
𝑃𝑥 = ( ) 𝑃 𝑥 𝑔𝑛−𝑥
𝑥 P = 86% Probability of Drone Delivery
g = 14%
n = 275 (average of 250 – 300
pkgs/day)
x = number of drones

39
 Table 4. Drone Delivery Probability

# of Drones Computation Probability of Drone

Delivery
(55 groups)
275 275 5%
𝑃𝑥 = ( ) 0.86275 0.140
275
250 275 0.9073%
𝑃𝑥 = ( ) 0.86250 0.1425
250
215 275 99.98%
𝑃𝑥 = ( ) 0.86215 0.1460
215
(86% of 275)
150 275 100%
𝑃𝑥 = ( ) 0.86150 0.14125
150
100 275 100%
𝑃𝑥 = ( ) 0.86100 0.14175
100
5 275 100%
𝑃𝑥 = ( ) 0.865 0.14270
5

Table 5. Delivery Time Estimates

Distance (one way) Time Drone Delivery (Round Trip) Time On-Foot (Round Trip)

0.522 mi 6:16 20:53

0.069 mi 0:50 2:46

40
 Figure 6: A table summarizing the second round of user testing
for the U.I specifically.

Table 7. Baseline human costs associated with delivery

Hours 10 8 6 5 4 0 (Reduced
Staff)

$ to worker 148.3 118.64 88.98 74.15 59.32 0

$ Saved 0 29.66 59.32 74.15 88.98 148.3

41
Figure 8: Amazon’s Cost Drone Delivery vs. Competitors Cost

42
 Table 9: User feedback from the final prototype.

On a scale of 1-5:
1 being uncomfortable, 5 being the
Comments
most comfortable you can
imagine; how comfortable is it?
“I think maybe the one improvement is a chair that
5’11” 3.5
can recline slightly
“One has clearance over the knee, which is nice. The
other one doesn’t.’
Having something to set your wrist on would be
helpful.”
Admit it; you want to work on the Lazy Boy.”
5’10” 4
“Didn’t realize they were buttons.”
“Like the tactileness, it takes longer to come back up
than you would think.”
“Make it a little more rigid.”
“I can sit on there for a long period of time.”
“I like the dials. I feel like you need a little too much
pressure to make them click.”
6’1” If you had dials with a button on the side for your 4
thumb, would you be able to press it with less of your
wrist”
“Yeah, it’s pretty good. I could definitely be in this
position for a long time. The buttons would be better
5’9” 4
if they were buttons on the face instead of the whole
knob being the button.’
“It’s fun to press the buttons.”
“I’d say the height of the padding is good.”
5’11” 3.5
With the inwards configuration: This is good” This is
how I would use them.”
“The knob pressing is a little too much force” If I have
5’5” to do this for 12 hours straight, it will get annoying.” 4
For me, it is the perfect height
“It’s a little lower than I’d like.” “The tactile feeling is
6’1” 4
really nice.” “I like the button press.”
“I’m quite comfortable.”
“They are a little difficult to push.”
5’9” The chair is very comfortable, but where my arms are 4
may take some getting used to. More used to having
my arms in my lap. I don’t use armrests much.”
“I’m cozy.”
6’5” “I’m pretty comfortable.’ 4
‘My forearms hurt. These are hard to click.”

Table 10: Project Criteria and Specifications

Requirements Source
Functionality

43
Assist delivery drivers in delivering packages Project Specs
Stop Optimization: Stop every 5 drone deliveries FAA Regulations
Geometry
Footprint Space Requirement: <1 ft2 Team
Time
Average U.I operation time (specific instruction): 0:20 > Team; ANSI
Average operation time for the U.I (before testing): 1:30> Team; ANSI
Energy
Current draw: < 1 A Hapticore Knob
Signals
Inputs: Knob velocity, Position, Torque, Clicks Project Specs
Physical Parameters (Bushing, Bore Diameter, etc.): Predetermined by Hapticore Knob
the provided knob.
Knob Sensitivity: Inputs measured at 22.5-degree intervals (1/16th of a GlobalSpec
full rotation).
Dead Zones: One 22.5-degree rotation to be completed within a 50 to Tool Module: The
150 ms time frame. Hand
Outputs: Driver location (GPS Coordinates), Package Locations (GPS Project Specs
Coordinates), Drone information (battery life, altitude, speed, trouble
light)
Display: OLED Project Specs
Safety
Lock knobs when not in use Team
Drone Speed <= 10 mph Team
Drone Height: 12 ft height delivery, 25 ft height cruising Amazon Practices
Ergonomics
10" clearance between seat and table Grand Rapids Chair
Co (Reynolds)
Arm Length: 22.4” - 27.9” Anthropometry
Summary Table

44
Seated Eye Height: 30-33 cm; 15 degrees above parallel Anthropometry
Summary Table
Materials
High Strength Carbon Fiber Granta Materials
Selection

Figure 10: House of Quality

45
Figure 11: Evaluation Matrix

46
Figure LAST. Final Assembly and Bill of Materials

47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69