Table of Contents
A SYSTEM DYNAMICS MODEL OF THE DEVELOPMENT CYCLE FOR
FUTURE MOBILITY VEHICLES
Taewoo Kang, Choongiap Lim, Daniel Delaurentis, Dimitri N. Mavris
School of Aerospace Engineering
Georgia Institute of Technology
Atlanta, Georgia GA 30332-0150
Phone: (404) 894-3343, Fax: (404) 894-6596
tkang@asdl.gatech.edu, samson.lim@asdl.gatech.edu, dan.delaurentis@aerospace.gatech.edu.
dimitri.mavris@ae.gatech.edu
Abstract
In the development of future air vehicles, a large number of inter-disciplinary areas come
into play. An overall transportation framework is introduced and key agents to a mobility value
network are identified. A simulation of the overall system is needed, but the present paper fo-
cuses on the aspect of product development for a future vehicle manufacturer. A System Dynam-
ics model is developed to investigate the importance of product development lead time and de-
sign failure rate. The effect of uncertainty is explored through Monte Carlo simulation. The case
results show the manufacturer’s net profit to be more sensitive to design and certification lead
time, as compared to a lower development failure rate. In this scenario, a policy that rewards the
early completion of work more than higher design success rate would be effective for the manu-
facturer. This study represents the initial step towards modeling the dynamics within a system-of-
systems mobility environment.
Introduction
The context of the presented research lies in the evaluation of future air vehicles. These
vehicles are expected to have significantly increased performance and affordability characteris-
tics, which are impossible to achieve with current technology capabilities. The goal of these ve-
hicles is to provide enhanced mobility to a large number of the population and revolutionize cur-
rent transportation paradigms. Enhanced mobility refers to the capability to save travel time, take
more trips or even travel to further destinations. The potential impact of disruptive technologies
has been well documented by Christensen [1] and its implications have been realized in many
technology intensive industries. “Disruptive” technologies refer to new innovations that have the
potential to displace mainstream market products in the future, although currently lacking com-
paratively in cost or functionality. The Inter-Urban Vehicles (IUV) concept is one such example,
which is an envisioned personal air mobility alternative that could revolutionize the transporta-
tion industry through affordable on-demand travel. The new paradigm for IUV design is not just
to provide for an efficient vehicle but to provide enhanced mobility. Integration with current
transportation modes has to be considered and issues such as overcapacity and delay, as has
plagued the transportation industry, are also essential in the process of enhancing mobility.
Aerospace designers have long relied on requirements set forth by existing customers.
But for disruptive technologies with no established markets or existing infrastructure such guid-
ance is no longer available. Hence before any kind of design activity takes place, designers must
first determine the minimum requirements needed for the concept to successfully pass through
every facet of the design, development, testing, manufacturing and marketing process. For a de-
sign effort related to improving future transportation, these questions can be answered through a
thorough understanding of the mobility value network as shown in Figure | [1, 2].
Lowest Level of Fidelity Highest Level of Risk
Policy Level / Business Dynamics
Mobility Service Providers
Vehicle Design Level
Aircraft Manufacturers
Disciplinary Level Technologies
Engines, Avionics, Subsystems
Component Level Technologies
Component & Material Vendors
syuawosnbay ubiseq
Testing & |
Evaluation
Standard
i
S
3
o
3
2
<
£
o
o
e
Traditional Scope to Aerospace Design
Highest Level of Fidelity Lowest Level of Risk
Figure 1 — Mobility Value Network Diagram
Design requirements are propagated in a top-down fashion from the policy level to the
component level technologies while the technical feasibility is determined from the bottom up. A
different value criterion exists at each stage as each organizational entity answers to different
customers. For example, airlines would cater to travelers and aircraft manufacturers would be
responsible towards airlines and the FAA. The policy level dynamics, which is typically ne-
glected in the aerospace design phase, usually contributes the highest level of risk to the pro-
gram.
The policy level dynamics of the entire subset of mobility stakeholders is shown through
Figure 2. Unlike a software development firm, for example, where the major interaction only oc-
curs between the development team and consumers [3], the transportation industry is interlinked
by a complicated web of commercial firms, governing agencies and infrastructure providers.
Traditionally, design focus has been placed on the interaction between research, manufacturing
and safety regulations. For future air vehicles where no information regarding the other stake-
holder is available, a “big-picture” perspective to design becomes necessary. This is referred to
as a “System-of-Systems” view, where all the mobility stakeholders are taken into consideration
in the design process. While there has been no shortage of innovative air vehicle concepts in the
past, implementation of new ideas in the transportation industry is becoming increasingly diffi-
cult with more interest groups involved in the decision making. Future innovation in air transpor-
tation would not be solely in radical vehicle designs but in managing the interfaces between
these inter-disciplinary fabrics.
Research & Infrastructure
Development
Consumers
Economic
Stakeholders
Developed
Products
Regulatory
Regulatory Agencies
Guidelines
Figure 2 — Mobility Stakeholder Network
Manufacturer
The mobility stakeholder network can also be viewed as a simple causal relationship dia-
gram for the air transportation industry. Each link between the stakeholders represents a relation-
ship, or a two-way flow of ‘stock’. For the research agencies-to-manufacturer link, this stock
may represent funding for research programs and developed product designs for manufacturers.
Many of the design efforts undertaken by research agencies do not relate directly to the consum-
ers whose lives may be impacted by the advances in technology. Transportation service-
providers on the other hand, are more significantly impacted by consumer trends but do not have
direct control over technology development programs. For a new air vehicle concept to be suc-
cessful, efficiencies in each stakeholder relationship must be maximized and the information
available at later stages has to be predicted during the early design phase to ensure a higher over-
all success rate as well as a lower redesign cost.
A more detailed representation of the relationships between mobility stakeholders is
shown in Figure 3. Stakeholders are listed in the rows and columns, thus the diagonal contains
the modules that represent the function of each stakeholders’ strategic enterprise. For example,
the Aircraft Design & Research module simulates the decision making of research agencies, and
the Production & Financial Dynamics module represents the decision making behavior of aircraft
manufacturers. Complexity in the relationships between the stakeholders exists on many levels.
The primary purpose of the diagram in Figure 3 is to illustrate all the possible relationships to aid
our mental models. In this sense, it is a precursor to a causal loop diagram. However, added
complexity beyond a certain level only serves as a detriment to model validity as compared to
additional insight. Thus, based on the characterization of the strength of the relationship, only the
strong feedback links may be quantitatively modeled. The variables that feed into each module
are categorized into strong, medium and weak levels of relationship, and the strong relationship
variables form the majority of the links shown earlier in Figure 2. The direction of the arrows
also indicates the feed-forward or feedback nature of the flow of resources. For example, ‘Final-
ized designs’ are fed forward from the Aircraft Design & Research module to the Production &
Financial module. ‘Research Funding’ and ‘Vehicle Requirements’ are in turn fed back to the
first module. This block serves as the realm of exploration using System Dynamics and de-
scribed in this paper.
3- Strong Relationship 2 - Medium Relationship 1 - Weak Relationship
; Ezononie
Research Agencies |, Aiteraf Regulatory Agencies! Ifasructure | Transportation Development ‘Conainiats
anutacturers roviders” | Series Providers | Development
Frat
femarn | ennoetona |p] em [7] teeeo |p rcrenina|y gf temo [p.J mse [g] a
Agencies | Rosoarch module 183] Feast Reseach Research Forecasting et], calane NI) late
studies: si ms
Research Vance data Venlo on
heat Production & [alee ret, ox Vehicle
AJruncng, Voice], Productions 1] a] conpiance’ |{] 2] naventon {ff a] wamtenance [9 3 1
tanutactrors |B Purina. Ve nancial Dynamics nie asian ates contoutons |b] atactheness
ae Fase
te gulator Lesion ircraf te gulator infrastructure Crersting confidence in confidence in
pes latory ile ‘guidelines for fe eikeret Regulatory 1k inresituet 1 Guidelines, (ht - lle as
oncies ine cantteaton | standardsmouule [$3] ‘Conscatin 5) ,Sueeines. 11) enssonation 2] tansponation
wl safety safety
Dag rare
TATE Network & Economie
infrastructure guidelines for itt Infrastructure fT data, ATC infreiiichiern: ry Airports, iy jot J Convenience of
Preven” (O02) Steinetue ulna compiance | Wtestuctre 13) murways, are [3] owt ioe Uff 3) st
compatibility needs meee
5 Deagt Fran
frensportation Veto : Economie Cost
femice lffs) SHH Yop] eurchace [ffs oeaton [pra] Sietonata | Serica Network 1 | cone ob [fa] eneshanesso
Providers Peake Demand healer epecily needs | Financial Dynamics creation Travel, Speed
Economic Research Data, Labor supply Emvronmental Tabor Sue Labor supply, | Macroscopic Public
Povelopment |{] 1) Gowenment {{} 1] economic [J 2} compiance [ff 2! wanscetanon M3} ecenemic Economic |ff 3} Transporation
[Stakeholders Funding benefits needs ieculrernants benefits Development modes
Dag K
Snaumers ‘quidelines for if Market If Public opinion Traveler ft Traveler it Taxes, Human | Consumer Trends
e fe) oNens™ (2) eect 4] ons [O05] eaters [APS] ect MP3) “sacar” | commer na
contigrain
Figure 3 — Relationship Matrix between Mobility Stakeholders
The research challenge in the field of mobility dynamics is the creation of a modeling
framework where each of the stakeholder relationships is quantified through simulation methods
such as System Dynamics. A System Dynamics formulation for part of the mobility framework
appears appropriate and useful for many reasons. It facilitates the simulation of the feedback
links within a complex system as well as the effect of policy over a prescribed time period. Sys-
tem Dynamics is also well suited to the level of modeling detail necessary for the aforemen-
tioned needs. At a system policy level, the emphasis does not lie in simulating and validating a
detailed model but rather on assessing a problem through policy simulation exercises.
The simulation work discussed in this paper focuses on the relationship that an aircraft
manufacturer has with the product development process (highlighted portions in Figures 2 & 3).
Results discussed pertain to sensitivities of management policies that the manufacturer may
adopt within the company or with other entities. An example would be shortening the product
development lead time by hiring more research staff or by reducing the redesign workload
through quality control. The primary value of the aircraft development cycle model is in the crea-
tion of a module which can now be linked to other stakeholder modules in the future.
Problem Definition — [UV Manufacturer’s Policy Analysis
A vehicle manufacturer plans to develop and launch a new generation of IUVs on the
market. Demand for mobility solutions is practically limitless since it addresses a fundamental
societal need. However, growth of the industry is limited by instabilities of the economy as well
as instabilities in the development cycle. One of primary goals of company management is to
expand its market share of mobility products while maintaining a positive balance sheet. Many
IUV concepts developed in the research laboratories are often abandoned before reaching the end
of the development cycle. This is due to the large uncertainties that exist due to insufficient in-
formation at the developmental stage. These uncertainties are propagated from other stakeholders
and causes scheduling and quality problems for the research team.
The development of a new IUV takes an average of 4 years from initial concept defini-
tion to product release. The introduction of a new IUV into the market reduces the average travel
time and hence increases the customer base for the IUV. This is due to travelers now being able
to travel to further destinations, take more number of trips, or make trips not possible before. The
cost of the new IUV is comparatively higher than other transportation modes but decreases over
time due to larger production quantities. [UV concepts typically have a market life of 5 years,
after which the product models are retired and taken off the market. The development of each
IUV prototype represents a significant capital investment for the company and a huge concern
for management is the ability of the company to forecast future demand for [UVs and manage
risk in the development process. A design prototype that is not certifiable or misses the market
window for product launch provides no positive return for initial research investment. Thus the
specific problem to be solved is determining the significance of the development delay time and
design failure rate for the manufacturer.
Implementation of a System Dynamics Approach
The simulation model traces the flow of IUV designs, from when the idea was first con-
ceived to when the product is finally retired from the market. An initial investment is specified
and an increase in cash flow occurs only after the first batch of products is introduced into the
market. The flow of revenues to the design inception phase acts as a reinforcing feedback and the
challenge for the manufacturer is in overcoming the design, certification and production lead
times. The primary model variables of interest are listed as in Table 1, including a “Connected
to” column that relates to Figure 3.
Table | — Model Variable List
Variable Name Description Connected to
Fiaenen rane Fraction of Projects that fail design feasibility test or FAA _ |Research groups,
certification test Regulatory agencies
Fraction of Projects that fail design feasibility test even affer |Research groups,
Fraction Falled Redesign redesign or after delaying FAA certification test Regulatory agencies
Revenue eamed by Manufacturer per product, after taxes, |Service Providers,
Revenue per product
production and miscellaneous expenses ($ millions) Consumers
Initial Capital Initial capital investment for product launch Esonomic stakeholders.
Consumers
Research groups,
ICost per Project Development Cost to design, test and certify a vehicle ($ millions) Regulatory agandee
Fraction Payout Fraction of product revenue used to pay debt and dividends {Manufacturer Policy
Interest Rate Interest rate Economic stakeholders
ICost Inflation Rate Infiation rate Economic stakeholders
Sroduct Merten [Average number of years that a product stays in the market [Manufacturer Policy,
before being retired Service Providers
Design & Certification Lead Time [Average number of years need to design, test and cerlifya Research groups,
lvehicle concept Regulatory agencies
Redesign and Certification Lead Time [Average number of years needed to redesign or added |Research groups,
certification time Regulatory agencies
Production & Marketing Lead Time _[A°e"@ge number of years needed to prepare for Manufacturer policy,
maufacturing and product introduction Economic stakeholders
The causal loop diagram of the [UV Manufacturer’s development cycle is shown in Fig-
ure 4. Much of the information is relevant to and hence referenced from Coyle’s example of a
pharmaceutical company [4]. The solid lines refer to a physical flow of material (stock and
flow) while the dashed lines only represent a flow of information (influencing variables). The
primary feedback presented in this example is the flow of Disposable Revenue from current
products in market back to product research and the key aspect is the two delay modules of
product research and product introduction, as well as the delay module of product retirement.
+ Products in
a Design ee
Cost of Product -
pe Product Research
Research ——* product Research a Completion Rate
~_y Start Rate
t+ +
i * Products Available
re H - Products Availa
# H Product «- ___ to Marketing
i ' ‘Abandonment Rate»
i H Be Cost of Product
i aha 5 ‘. y e—_ Introduction
i H fe " g-—— Rate of Starting Product q_._
{ H Products in + Introductions + Sy
: " a introduction }
Disposable {+ phase 2 +|/D i
Roveme row: A. rate of Contpleting Product i
H tt +4 Introductions, :
‘ H Products it D
‘ H Market 7
‘ o_o F + H
isposable ; : }
iRavertie Flow per 4 ProductWitdrawal
\ Product g.
Fraction of Disposable _----.-------
Revenue Flow spent on
product introductions
Figure 4 — Causal Loop Diagram [4]
The stock and flow model developed using STELLA®, is shown in Figure 5. The stock
referenced as “Product Ideas” represent a pool of research development projects in the early de-
velopment stage. The rate at which this increases is determined by the average costs associated
with each developmental project and the resources pooled from the company cash flow. “Com-
pleted Designs” represent projects at the end of the development stage that has passed the feasi-
bility criterion. “Rejected Designs” represent product ideas that have failed the feasibility crite-
rion. These failed concepts go through a redesign process, at the end of which the products are
tested again. The average design and certification lead time is 3 years and the projects that fail
the first feasibility test undergo an extra year of redesign and testing to achieve the required stan-
dard. A lead time of a year for production scheduling and marketing is then required before the
new products can be first introduced to the market. The number of products in the market deter-
mines the company’s revenues for the year, based on the average revenue achieved per product.
Due to decreasing revenues with product age, products 5 years and older are retired from the
market to make way for newer configurations. The “Cash Flow” stock is initially representative
of the capital investment and subsequently represents the revenues from product introductions.
Before revenues are earned all expenditures involved go towards bring the first batch of vehicle
concepts to market, and following that, a percentage of the revenues go toward repaying debt and
payment of dividends to investors.
The model is simplified by aggregating the delay times of the multiple design and testing
phases into one delay process (refer to DP Jogic in Figures 5 & 6). This level of abstraction may
be insufficient to address a different problem of identifying the congested segment of the devel-
opment cycle that increases delay time. However, for the prescribed problem of measuring the
significance of the development delay time and design failure rate, the level of detail is adequate.
PM logic
cost per project development DP logic a RP logic
‘expenditures
Product Ideas Completed Designs Products in market
design inception design production & retiring
process marketing products
inflation rate Revenue per product
Rejected Designs| Cast Flow
expenditures
redesigning incoming
revenue
fail rate
fraction failed fraction payout
profit pay out
fail again rate payout
logic
fraction failed redesign ai
interest rate (© initial capital
Net Profit
Figure 5 — Stock and Flow model of the product development cycle
Design Detay
finishing
design
‘starting
design
fraction failed
design inception dicen
designs
Redesign Delay
starting finishing
redesign redesign
successful
re designs
fail rate
fraction failed again
Figure 6 - Stock and Flow model of the Design Process delay (DP logic)
Model Validation and Testing
In System Dynamics literature, several guidelines are offered for model validation and
testing. These include validating the influence diagram with the statement of the problem, di-
mensional validity throughout the model, and simulation results that stay within reasonable
boundaries. These methods were employed to test the validity of the manufacturer’s model de-
scribed earlier, and the baseline results display the reality of the model to a reasonable extent.
Coyle suggests that a more suitable definition to model validity is one that is “well suited to its
purpose and soundly constructed” [4]. The primary purpose of the presented model is to examine
the importance of design & certification lead time and design failure rate for an [UV manufac-
turer, and to establish links to other mobility stakeholder modules in a system of system con-
struct. In terms of meeting the problem objective with sufficient simplicity, the model is believed
to be adequate. The baseline simulation analysis is performed with the assumptions as shown in
Table 2.
Table 2 — Variable settings for baseline simulation run
Variable Name Value Units
Fraction Failed 0.5
Fraction Failed Redesign 0.5
Revenue per product 65 $ million
Initial Capital 80 $ million
Cost per Project Development 10 $ million
Fraction Payout 0.5
Interest Rate 8%
Cost Inflation Rate 8%
Product Market Life 5 years
Design & Certification Lead Time 3 years
Redesign and Certification Lead Time 4 years
Production & Marketing Lead Time 1 years
The simulation results for the baseline settings are shown in Figure 7. “Net Profit” refers
to the cumulative total net present value of the investment thus far, and it can be seen that break-
even occurs at year 8, the 4" year after first product launch. The revenues generated from the
products in market from years 4 onwards contribute to the cash flow of the company sufficiently
to provide funding for the next generation of products. The baseline scenario results accurately
represent a typical situation for aircraft manufacturers, with long development lead times and the
financial viability of the company being highly dependent on the success of its initial product
offerings.
Financial Cash Flow Analysis ve Time Products in Development vs Time
1200 2
Wet Proft Products in Market
++ Profit Payout | 1B Product Ideas
1o00}{— cash Flow | Completed Desians Fs
18 :
f }
20} i] i
Millions (8)
Number of Products:
‘ _ x
o1 23 4 6 6 7 8 9 0 1 12 13 14 16 o 123 4 6 6 7 8 9 0 1 12 13 14 16
Time (Years) Time (Years)
Figure 7 — Results from Baseline analysis
The simulation results for a scenario with higher development costs and lower revenues
generated per product is shown in Figure 8. With 30 million dollars of revenue generated per
year per product, 100 million dollars of startup capital and a development cost per project of 20
million dollars, the manufacturing company is not viable beyond its first range of product entries.
This scenario is not highly realistic since it does not give consideration to learning curve effects
or future anticipatory reaction by management to increase revenue or reduce cost. However, un-
der extreme conditions, the results do correlate with expected outcome and reinforces confidence
in the validity of the model before an additional layer of detail is built into the model.
Low revenues & High Cost Scenario for Net Profit Low revenues & High Cost Scenario for Products in Market
; — Net Proit
eof! +--+ Profit Payout
4 =~ Cash Flaw
— Products in Market
= Product Ideas
=~ Completed Designs
§ millions
6 :
042 3 4 5 6 7 8 9 10 11 12 13 14 15 O42 3 4 5 6 7 8 9 10 11 12 13 14 15
Time (Years) Time (Years)
Figure 8 — Lower revenues with higher development costs per product
Simulation Results and Discussion
The problem under study is the link between research and manufacturing and determina-
tion of preferred policy. In the link between research and manufacturing, the key barriers to suc-
cess are the design failure rates as well as the design & certification delay time. Design projects
that do not progress to the expected level of maturity do not represent any value to the manufac-
turer despite significant initial investments [5]. Hence the importance of the design failure rate
and the design and certification delay time cannot be over emphasized. The primary results ob-
served for the simulation is the manufacturer’s sensitivity to product design and certification lead
time as well as design and certification failure rate. It is a widely accepted fact that these two
factors are critical to a manufacturer’s financial profit, but to an unknown extent. The two fac-
tors are also not entirely independent of each other. Rushed design work would likely be lacking
in terms of quality compared to one that has gone through a normal cycle time, and vice versa.
Should management policy direct effort to better quality control or shorter development lead
time? Several scenario results are shown in Figure 9.
Effect of Developmental Lead Time on Net Profit
Effect of Developmental Lead Time on Products in Market,
2500 60
— Baseline - 3 years — Baseline - 3 years
ws Lead Time - 33% Decrease ---- Lead Time - 33% Decrease
2000 [= Lead Time - 33% Increase so Lead Time - 33% Increase a
g 10 240
= 1000 23
cS . a
a S
3 3
€ 20
3
10
we a
oD1462 3 4 5 6 7 8 9 10 11 12 13 14 1°23 4 5 6 7 8 9 10 11 12 13 14 15
Time (Years) Time (Years)
Effect of Design Failure Rate on Net Profit Effect of Design Failure Rate on Products in Market
800 30
— Baseline -0.5 — Baseline -0.5
zon /| ---> Fraction Failed - 33% Increase ~--+ Fraction Failed - 33% Increase /
— _ Fraction Failed - 33% Decrease 25 {= Fraction Failed - 33% Decrease /
z £2
é 3
= gt
€ a
a 3.
2 4
e” ;
5
123 4 6 6 7 8 9
10 11 12 13 14 15 1 23 4 5 6 7 8 9 10 11 12 13 14 15
Time (Years) Time (Years)
Improving Design Fail Rate ve Decreasing Development Lead Time Impreving Design Fail Rate vs Decreasing Development Lead Time
1400 %
— Baseline — Baseline
Lead time - 39% Decrease, Fraction Feiled - 39% Increase Lead time -23% Decroase, Fraction Failed 33% Increase |,
1200)! Lead time - 33% Increase, Fraction Failed - 33% Decrease 30}|— Lead time - 33% Increase, Fraction Failed - 33% Decrease |.”
1000 s
Ea
3 =
2 @ 20
Fs el
= é
a Gib
2 a
510
2
5
Q
o1234 567 8 9 0 1 12 19 4 18 o12 3 4 6 6 7 8 9 0 1 12 13 4 15
Time (Years)
Time (Yeats)
Figure 9 — Effect of Development Lead Time and Fail Rate
The sensitivity of the total Net Profit to a shorter design & certification lead time of the
manufacturer appears to be much more significant compared to a lower development failure rate.
A significant advantage of a shorter development time is also being able to return a profit much
earlier. The positive gain from better design quality is also important but significant gains are
only realizable further in the future. One applicable policy the manufacturer can learn from this
exercise is the distribution of design work and research funding to research agencies. With a
priority on shortening development time, individual technology portfolios may be integrated and
its development supervised as a whole. Aerospace technology development typically takes place
at multiple locations, each with different areas of expertise. NASA conducts research also in a
distributed network structure, with independent contractors filling in many of the gaps. While the
quality of the individual work may be higher, the time it takes to integrate the pieces signifi-
cantly increases development time. For a manufacturing company depending on profits and not
government subsidies to survive in the marketplace, an efficiently integrated development proc-
ess may be the most critical aspect.
Reinertsen describes that research developmental projects are almost never completed
ahead of schedule [5]. They either finish on time or are delayed by an uncertain amount of time.
There is not significant incentive for early completion of work, and emphasis in the research labs
is always more driven towards successful demonstration of the technology. Based on the pre-
sented scenario results, it appears that a research policy that rewards the early completion of
work more than higher design success rate may be more effective for the manufacturer in the
long term.
Another important question constantly on the mind of any policy maker is the issue of
uncertainty. Even with the best of planning, an unexpected downturn in the economy may force
the company towards financial instabilities. To simulate the effects of uncertainty, a Monte Carlo
Simulation was performed with a uniform distribution placed on the variables shown in Table 3.
The uncertainty ranges were selected with the baseline values as mid-points except for the “Cost
per Project Development” (research spending is never less than the initial budgeted amount but
often goes over the budget). Uncertain certification requirements are reflected on the “Fraction
Failed” variable. A Monte Carlo Simulation is simply a random number generator that selects a
variable setting from its distribution and performs the analysis a large number of times repeat-
edly. With a sufficient number of analysis runs, an accurate assessment of the probability of an
event occurring can be predicted. A uniform distribution was used to reflect the lack of knowl-
edge on the uncertainty of the variables used. With a more specific context applied, more refined
distributions can be applied to add confidence to the simulation.
Table 3 — Uncertainty range of variables
Variable Name Range Units
Fraction Failed 0.4 -0.8
Fraction Failed Redesign 0.4-0.8
Revenue per product 30 - 80 $ million
Cost per Project Development 10 - 20 $ million
Interest Rate 3-12 %
Cost Inflation Rate 3-12 %
With current software modeling capabilities in STELLA®, a significantly large number of
runs could not be performed easily. Hence, 500 simulation runs were performed and the results
for Net Profit are displayed in Figure 10. The histogram displays the extent of risk present, as-
suming the range and shape of the uncertainties expected from the variables. Assuming a success
criterion of a Net Profit of greater than 100 million dollars after year 15, the probability of suc-
cess is measured to be approximately 35%. Furthermore, nearly 24% of the simulation runs re-
turn a negative net profit after year 15. The spread of the Net Profit distribution in Figure 10 is
observed to be much less than the spread in Net Profit due to changed lead time, from Figure 9.
Similar results exhibited from a more through simulation exercise, would support a clear policy
recommendation to reduce lead time, despite uncertainties present due to missing information.
An important aspect for future work is the incorporation of more accurate implementation
of Monte Carlo Simulation, perhaps through the use of MATLAB’ or Crystal Ball”. The meas-
urement of risk in the mobility stakeholder is important also in another aspect. Typically in a ro-
bust design environment, designers use optimized control variable settings to minimize overall
sensitivity to noise variables. Control variables refer to ones that are within the designers’ control
and noise variables to ones which are not. In a mobility network, one entity’s noise variable
could very well be another’s control variable. With the links established between each of the
stakeholders, the number of noise variables within the entire system becomes much smaller than
the sum of the parts. For a revolutionary concept to maximize its probability of making a suc-
cessful impact in the marketplace, the mobility network has to be calibrated for robustness as a
whole system and not as individual entities.
% Frequency
% Frequency
Soa ca] a a0 700 mo io 2 0 8 1 1 om DAO
Net Profit (§ millions) afer 10 years Net Profit (§ millions) after 15 years
Figure 10 — Histogram of Net Profit after 10 and 15 years
Current Limitations and Future Work
The work described in this paper has been presented in two levels. First, a qualitative de-
scription of the mobility stakeholder network is discussed and then a quantitative analysis of the
manufacturer’s development cycle is presented. With the adoption of a top-down approach,
depth in analytical modeling detail has not been achieved and is a future task at hand. Several
future tasks, for the further development of the mobility dynamics environment, are listed below
and in no order of importance.
e Use of Real Options theory to enhance the analytical capability of the manufacturer module.
When an economic analysis is being undertaken to assess future viability of a design concept,
a Net Present Value (NPV) approach, based on an assumed cash flow, is typically used to
represent the value of the project. While such an approach is easily manageable with the aid
of an excel spreadsheet and a dose of common sense, it is a deterministic method of calculat-
ing project value that is neither accurate nor realistic. A probabilistic enhancement to the
NPV approach includes ranges in uncertainty and provides valuable estimates of the prob-
ability of success. The value of a probabilistic approach, however, only lies in the ability to
quantify the uncertainties involved. In order to account for the actions the project planner
takes over time, in reaction to the fluctuations in the market, a stochastic model has to be
employed. Real Options theory is derived from financial options theory and it is a stochastic
method aimed at reducing the project’s exposure to uncertainty.
e Use of Agent-Based methods and utility theory to formulate a consumer demand module. It
would not be an understatement to designate the consumer demand module as the most im-
portant of the mobility stakeholder dynamics. Disruptive technologies are only viable on the
assumption that consumer trends change with the introduction of new products, and predic-
tion of future demand cannot be extrapolated from past product research data. With agent
based methods, guidelines can be created that form the basis of consumer decisions and fu-
ture behavior of consumer groups can be predicted.
¢ Creation of meta-models to represent other stakeholder entities. Through the use of response
surface methodology, meta-models of the aircraft design module have been developed in the
past [11]. The creation of similar meta-models for other stakeholder entities such as regula-
tory agencies and infrastructure providers remain a future modeling challenge. They will
have to be based on existing high fidelity analysis tools and with sufficient flexibility to link
with other modules. These meta-models can then be used to analyze the solution space of the
mobility network and optimize policy solutions.
e Exploration of the noise/control dichotomy and policies to coordinate overall system robust-
ness. Prior studies relating to risk reduction have been conducted piece-wise, with robustness
of an individual stakeholder or a project as the focal point. An important aspect of future re-
search would be the expansion of these boundaries to examine robust system-of-system sce-
narios.
Conclusion
The concept of future vehicles for mobility improvement was first introduced along with
the need for paradigm change in aircraft design. Instead of focusing on static vehicle require-
ments generated from case studies, there exists a need for designers to generate potential re-
quirements for future transportation from a dynamic viewpoint. Based on the underlying concept
of mobility, a network of mobility stakeholders with the most significant effect on the transporta-
tion system was identified. The coupling of relationships between the stakeholders, which are
inherently causal effects, provides an effective platform for the implementation of System Dy-
namics methods. Due to the large scope of the overall mobility problem, focus on the develop-
ment cycle was necessary for initial quantitative analysis. The stock and flow model describes
the flow of products through its design and certification stage as well as the revenue flow for the
manufacturer, which is dependent on available products in the market. The simulation runs ex-
amined the effect of design & certification lead time as well as the failure rates and found the
lead time to have much more significant impact on the long term financial stability of the manu-
facturer. A simple Monte Carlo simulation exercise was also performed to simulate the effects of
uncertainty within the model and only 44% (based on assumed uncertainty ranges) of the cases
were found to produce a net profit of greater than $100 million dollars over a 15 year period.
The System Dynamics model of the development cycle is relevant to the overall study of the
mobility environment, by providing policy guidelines in the relationship between research agen-
cies and manufacturers.
Pa
13.
14,
References
Christensen, C. M. The Innovator’s Dilemma. Boston: Harvard Business School Press, 1997
Holmes, B.J., “Alternatives for Air Mobility,” Proceedings of the 2002 World Aviation Congress,
Society of Automotive Engineers, Warrendale, PA, Nov.
Himola, O., Helo, P., Ojala, L. “The Value of Product Development Lead Time in Software
Startup,” System Dynamics Review 19 (2003): 75-82
Coyle, R.G. System Dynamics Modelling. A Practical Approach. London: Chapman & Hall, 1996
Reinertsen, D.G. Managing the Design Factory. New York: The Free Press, 1997
Starr, P.J. “Modeling Issues and Decisions in System Dynamics,” TIMS Studies in the Manage-
ment Sciences 14 (1980): 45-59
Sterman, J.D. Business Dynamics. Irwin Mcgraw-Hill, 2000
Forrester, J.W. Urban Dynamics. Cambridge: The M.I.T. Press, 1969
Mavris, D.N. and DeLaurentis, D.A., "Methodology for Examining the Simultaneous Impact of
Requirements, Vehicle Characteristics, and Technologies on Military Aircraft Design," Proceed-
ings of the 22nd Congress of the International Council on the Aeronautical Sciences (ICAS), Har-
rogate, England, August 27-31, 2000. Paper ICAS-2000-1.4.5.
. Myers, R.H., and Montgomery, D.C., Response Surface Methodology: Process and Product Op-
timization Using Designed Experiments, John Wiley & Sons Inc., Indianapolis, IN, 1995.
. DeLaurentis, D.A., Mavris, D.N., "Uncertainty Modeling and Management in Multidisciplinary
Analysis and Synthesis," AIAA Paper 2000-0422, Jan. 2000.
. Du, X. and Chen, W., "Efficient Uncertainty Analysis Methods for Multidisciplinary Robust De-
sign", AIAA Journal, Vol. 40, No. 3, 2002, pp. 545-552.
Meyer, M.D. and Miller E.J., Urban Transportation Planning: A Decision Oriented Approach,
2"4 ed., McGraw-Hill, New York, 2001.
Donohue, G.L. and R. Shaver, "United States Air Transportation Capacity: Limits to Growth Part
I (modeling) and Part II (policy),” National Research Council Transportation Research Board,
National Academy Press, Report Nos. 00-0582 and 00-0583, Washington D.C., Jan. 2000.
Back to the
CC BY-NC-SA 4.0