Electric Multirotor Propulsion System Sizing for

Performance Prediction and Design Optimization

Dmitry Bershadsky∗ and Stephen Haviland∗

Georgia Institute of Technology, Aerospace Engineering, 270 Ferst Dr. NW, Atlanta, GA 30332-0150, USA

Eric N. Johnson†
Georgia Institute of Technology, Aerospace Engineering, 270 Ferst Dr. NW, Atlanta, GA 30332-0150, USA

One of the more daunting tasks of designing a multirotor unmanned aerial vehicle (UAV)
is the selection of a propulsion system that will provide desired performance. Rigorous
methods for selecting these drive components, that is, the motors, propellers, and batter-
ies for electric UAVs are not readily available. Currently, many UAV designs are based
on legacy selections or limited and at times incorrect manufacturer data. These design
methods are either simplistic or lacking in analysis and validation of component selection.
Proper propulsion system design should address the mission requirements for which the
vehicle is being designed. A proper design methodology is the best chance that the designer
has to create a new vehicle that will be mission-capable. This paper attempts to satisfy the
need for more thorough method of propulsion component selection. The paper is written
also to document the popular online drive system analysis tool due to numerous requests.
This tool is one example implementation of the methodologies described by this paper.

Nomenclature
α angle of attack
β blade physical twist
B number of blades per propeller
c wing chord
cR c/R
Cb discharge rating
CD drag coefficient
CL lift coefficient
C energy capacity
dr blade radial section
dt throttle command
dt change in time
dL incremental lift
dQ incremental torque
ηE ESC efficiency
ηM motor efficiency
ηG gearing efficiency
G gear ratio
I0 no load current drawn by 1 motor
Ia current drawn by avionics
Id current drawn by all motors
Im current drawn by 1 motor
∗ Graduate Research Assistant, Georgia Institute of Technology, Aerospace Engineering, 270 Ferst Dr. NW, Atlanta, GA

30332-0150, USA. [email protected].

† Associate Professor, Georgia Institute of Technology School of Aerospace Engineering, [email protected].

1 of 22

American Institute of Aeronautics and Astronautics

Ip current drawn by payload
It current drawn total
Kv motor speed constant
L total blade lift
Mtip blade tip Mach number
mavn avionics mass
mb battery mass
me ESC mass
mm motor mass
mp total propeller mass
mpay payload mass
mt total vehicle mass
mw motor mass
Nm number of motors
Np number of propellers
ω rotations per minute
pp propeller pitch
pmat propeller composition
Pe electrical power
Pm mechanical power
Q total blade torque
r blade radial section location
rR r/R
R rotor radius
Rb battery internal resistance
Re ESC internal resistance, also RDSON
Rm motor internal resistance
Rw wiring harness internal resistance
s battery serial cells
S wing span
θ vehicle pitch w.r.t. local vertical
Vc operational voltage per cell
Vs voltage seen by motor coils
w rotational speed rad/s

I. Introduction
The framework and algorithm presented is meant to be used to select and/or validate the selection of
electric RC aerial vehicle drive components which will allow a vehicle to accomplish specific missions. As
mentioned, one implementation of the methodology is available online, and is named the Electric Multirotor
Sizing Tool (EMST). Missions are defined as hover, climb, or dash segments with associated durations. The
method is capable of considering generic, standard and non-standard vehicle concepts including: airplanes,
multirotors, single main rotor helicopters, coaxial configurations (motors and/or propellers), fixed or tilt
wings/rotors/free wings, as well as generic combinations of the above.

A. Existing propulsion system design methods

The required fuel fraction method or RF method4,5,6 is used to size large, gas, or jet fueled aircraft such as
manned helicopters. The method, when given a mission, provides a minimum engine size and thus attempts
to size other components such as rotor blades and transmissions, as well as providing a minimum GTOW
required after calculating other component weights. However, a similar method to size electric UAV drive
components has not yet been found to be readily available.
Latorre7 documents a design process using BEMT to select motors, propellers, and batteries in an
iterative fashion. This method should yield a reasonable result, but it does not optimize for any of the design

2 of 22

American Institute of Aeronautics and Astronautics

variables. Gur10 presents a multi disciplinary optimization (MDO) approach to design a propulsion system
based on goals such as rate of climb and loiter time, and presents a useful modeling analysis of motors and
batteries. Also presented are sensitivity analysis to certain propeller design elements. Lundström8 describes
an automated approach to designing the propulsion system as well as the body for small airplane UAVs.
Ampatis9 shows a design method using parameterized data from motors, batteries, and ESCs to study the
effect of varying payload on design characteristics. Vehicle diameter, energy, motor length, and battery
weight are described. The authors also consider the effects of varying the number of motors on the design.
Also presented are thoughts on modeling of electrical propulsion systems. Bouabdallah11 describes a method
for iteratively designing a vehicle with a maximum mass and length to achieve a desired thrust-to-weight
ratio. The method requires a database of actuator, battery, and airframe components to calculate the loop
masses. The study presented in this paper, in contrast, attempts to parameterize drive components to rid
of the need to use a database at design and optimization time.
Less rigorous methods in terms of optimality of UAV design also exist. For instance, eCalca has become
one of the go-to tools for some UAV designers, especially hobbyists. Benito12 describes a process of designing
a multirotor using this tool. eCalc allows users to input specific lists of components and will provide a
calculated flight time, in addition to other useful data. Similar to the analysis algorithm described in this
paper, it allows designers to check flight endurance and other characteristics of some types of common
vehicles: heli/multirotors, planes, and ducted fans. eCalc requires that the specific drive components are
provided to the tool. This is one of the major differences between eCalc and the analysis algorithm described
in this paper. That is, the EMST parameterizes and abstracts propulsion system characteristics to allow
users to input only relevant drive system parameters instead of selecting specific components. In addition,
the drive optimizer algorithm proposed in this paper is essentially the analysis methodology or eCalc in
reverse, which the authors believe is at this point a unique capability. The algorithms presented here may be
used in conjunction with or in lieu of existing tools such as this one either to validate each other, or provide
starting points for design.

B. Multirotor electric propulsion system design considerations

There are multiple considerations that must be addressed when designing a multirotor UAV drive system.
The scope of the design space may encompass some or all of the variables described in this section. There
may be other considerations but the ones here are considered as primary design variables in that they greatly
affect flight performance, mainly payload capacity and endurance.

1. Motor
Brushless DC (BLDC) motors are used for just about all of the vehicles described by this paper. These
are generally preferred over older, brushed designs for their greater efficiency in converting electrical to
mechanical energy. BLDC motor designs have two configurations, outrunner (OR) and inrunner (IR), which
define the component of the motor that rotates and is attached to the output or actuation axle. IR motors
spin an axle with magnets inside of array of windings arranged circularly around it. OR BLDC motors invert
the IR design and wrap a bell of magnets around the winding array. Although IR motors allow for some
tighter installations due to the body of the motor being static, the OR configuration allows those motors to
produce more torque than their IR counterparts. This makes IR BLDC motors a common choice for many
small (< 100 g) multirotor builds, and OR a good choice for anything larger. It is possible to use gearing
(Parrot AR drone is a popular design) as well but due to the inherent complexity, many designs use direct
drive BLDC OR motors. In a direct drive design, the propeller is directly attached to the motor axle. One
benefit however of a geared design is the potential to hide a motor’s axle mostly inside the motor body,
making it more durable in the event of a crash. Of course, the gearing is now accepting the brunt of the
crash energy, but this may be easier to replace than a motor axle.
Arguably the most important parameter of motor selection is the speed constant, or Kv , measured in
RPM/V. It is the manufacturers indication of roughly how fast the motor will spin when unloaded per applied
volt across its wires. This value must be properly matched to the selection of the propeller and battery, as
well as chosen for resulting efficiency and lift capability. For the former, the motor should be able to handle
the mechanical and torque electrical power loads applied by turning the propeller. Too high of a Kv or
voltage, and the motor may be unable to handle the loads at high throttle, or may spin in a lower efficiency
a http://www.ecalc.ch/

3 of 22

American Institute of Aeronautics and Astronautics

regime with too low of an RPM at low throttle. Too low of a Kv and the motor may have to spin too fast to
generate sufficient thrust to fly. Generally a lower Kv motor that spins a larger and/or low-pitched propeller
will be more efficient than a high Kv motor with a smaller and/or high-pitched propeller. One advantage of
the latter however is higher performance for acrobatic flying.

2. Propeller
Propeller composition, radius, pitch, and number of blades must also be chosen to work properly with the
chosen motor. The material from which the propeller is made might affect the efficiency of the propeller at
different RPMs. This may occur for softer propellers due to flexure of the blades changing effective angle
of attack at radial sections away from designed angles. This effect is described by Harrington.3 Increasing
the propeller pitch and number of blades generally generates more thrust, but at a cost of efficiency and
increased electrical and mechanical power requirements on the motor. Increasing the propeller’s radius is
generally more efficient, assuming the rest of the drive system is capable of handling the load. This is because
the larger propeller, with all else being equal, may spin slower to generate the same lift. This allows the
induced velocity to drop, thereby increasing propulsive efficiency.

3. Battery
This study considers common, current, hobby-grade batteries. Newer hobby-grade battery lithium polymer
(LiPo) compositions are capable of specific energy of up to around 250 Wh/kg,15 about an order of magnitude
lower than gunpowder, and two orders of magnitude lower than kerosine. Many COTS batteries are easily
capable of discharging at currents greater than 100 A, some for extended periods of time. These batteries
have all but replaced the previous NiCd and NiMH batteries which were used for hobby-grade vehicles in
the past. Another composition known as lithium ion (LiIon), commonly used in consumer electronics such
as laptops, has a specific energy 50% better than LiPo compositions, although maximum discharge rates are
lower, which makes them better choices for more efficient vehicles.
Other compositions include lithium iron phosphate (LiFe or LFP), which are have a specific energy of
around 100 Wh/kg. While they are heavier than other lithium compositions, they are considered to be more
stable under discharge and when subjected to damage. Unlike LiPo batteries, they will not explode when
punctured and their cells exposed to the air. These batteries are also reported to be able to withstand a
greater number of discharges than other types mentioned above.
Aside from chemical composition, other main parameters must be considered when selecting a battery
or batteries for the vehicle, although they are not necessarily independent. The battery’s cell configuration
and capacity must be considered. The battery must be chosen such that sufficient but not excessive voltage
exists to power the propulsion system and avionics. LiFe cells have a maximum charged voltage of around
3.65 V/cell, whereas LiPo and LiIon cells are charged to 4.2 V/cell. Batteries are built by stacking cells in
series (S) and parallel (P) to achieve endurance and discharge rate goals. Stacking cells in series increases
the pack’s voltage, while adding more cells in parallel increases its capacity. For example, a 4S1P LiPo
battery has four cells in series and one in parallel, giving a fully charged voltage of 16.8 V, while the same
configuration LiFe battery will be fully charged at 14.6 V. This value must correspond with the selection
of motor Kv and propeller to achieve proper RPM and thrust values to lift the vehicle, along with any
regulators that power other onboard electrical systems.
The capacity, normally measured in mAh, determines the energy storage of the battery. A 1500 mAh
battery can provide 1.5 A of current for 1 hour if discharged to 100%. Batteries which are discharged to
around 80% of their capacity per usage also tend to last longer than those which are discharged to a greater
extent. The discharge rate, related to the so-called ”C-rating” of a battery, is a manufacturer’s indication
of the discharge capability of the battery. A 40 C 1500 mAh battery in an operational condition is capable
of sustaining 60 A discharge for some amount of time. The power requirements of the drive system must be
taken into account when selecting the battery.

4. Electronic Speed Controller

At a high level, the electronic speed controller (ESC) takes a PWM control signal, generally 1 ms (off)
to 2 ms (full), and converts it to AC current. The frequency of the switching, done by onboard FETs,
spins the rotor on the BLDC. The main parameter when selecting ESCs to consider is the maximum rated

4 of 22

American Institute of Aeronautics and Astronautics

amperage. This must be sufficiently higher than the required drive current per motor such that the FETs do
not overheat and fail. A secondary decision is whether or not to use ESCs with a battery eleminator circuit
(BEC) onboard. These regulate main battery voltage down to usually 5 V to power other avionics. If this
is not necessary, it is sometimes better to use optically isolated (OPTO) ESCs to save energy and heat, as
the BECs tend to be inefficient and may cause unnecessary heating and energy loss.

II. Parameterization of drive components

A relationship was found between the masses and several key characteristics of all drive components
considered by both the analysis and optimization algorithms. These are necessary to abstract actual com-
ponents into just the key characteristics to enable the algorithms to guess the masses of the components
required to lift the vehicle. They are used in both analysis and optimization algorithms, proposed in later
sections.

A. Motor
The data shown here consider both IR and OR motors, the latter being a more popular choice for many
multirotor configurations for their generally higher torque (per mass) and ease of installation. Figure 1
shows the trend of high Kv motors having low mass and vice versa, also studied by Gur,10 although with
less motors, and with a broader focus on motor types, including heavy duty/high voltage motors. Both IR
and OR motors are plotted in the figure.

104
OR
OR
IR
103 IR
motor mass, g

102

101

100 1
10 102 103 104 105
Kv , RPM/V

Figure 1. OR and IR Kv to mass relationship.

Figure 1 shows that in general, IR motors tend to be heavier per unit Kv than OR motors.
The trend in figure 1 generates eq. (1), providing a mapping between Kv and predicted motor mass mm
for both OR and IR motors.
mm = 10p1 Kvp2 , g (1)
where p1 is 4.0499 and p2 is -0.5329. For IR, p1 is 4.4482 and p2 is -0.5242. Of particular interest for
this study are motors in the range of Kv and mass shown in Figure 2, as they are popular choices for the
hobby-sized vehicles considered below.
Eq. (1) is modified to match some of the lower Kv motors available on the market, as the original fit does
not match market survey data well in this region between 0 and around 500 Kv for hobby-grade motors.
The modified equation for OR motors is seen in figure 2 as OR H. Around 991 OR and 696 IR BLDC motors
were surveyedb with incomplete entries removed.
b http://electrofly.free.fr/download.php?lng=en accessed 03/10/2015

5 of 22

American Institute of Aeronautics and Astronautics

 OR
OR
IR
1,000 IR

motor mass, g
OR H

500

0
0 2,000 4,000 6,000
Kv , RPM/V

Figure 2. OR and IR Kv to mass relationship. OR H is the modified curve used for the analysis and optimizer
tools.

In addition to parametrizing the mass as a function of Kv , we are interested in the resistance of the
motor, as generally, larger motors have larger electrical resistances. Gur10 shows that the internal resistance
of the motor Rm is a function of Kv or mass.

B. ESC
Data from around 20 ESCs rated for up to 100 A are taken to study the relationship between ESC mass and
current throughput capacity. These data are shown in figure 3. Eq. (2) is the function found that provides

survey
80 fit

60
ESC mass, g

40

20

0
0 20 40 60 80 100
maximum rated amperage, A

Figure 3. Masses of ESCs and their respective maximum amperage rating.

a mapping between an ESC’s maximum rated amperage and its mass:

me = p1 Amax , g (2)
where p1 is found to be 0.8421. In addition to ESC mass, it is useful to know the resistance of an ESC when
it is operating. Ampatis9 shows that the resistance of an ESC Re is between 9 and 45 mΩ. This value is

6 of 22

American Institute of Aeronautics and Astronautics

3RDSON , where RDSON is the resistance of one path from the input to the output of the ESC when the
path is shorted.
Wiring, including all signal and power lines, has been found to weigh around 5% of the GTOW of all
UAVRF multirotor vehicles. This fraction is used for all calculations by the tool.

mw = 0.05mt (3)

Wiring resistance Rw is small and is scaled with the amperage required Id .

C. Battery
A parameterization of battery mass and related capacity is also required for the algorithm to function. Data
found are shown in figure 4.

2S
3S
4S
1,000 5S
battery mass, g

6S

500

0
0 2,000 4,000 6,000 8,000
battery capacity, mAh

Figure 4. Battery masses for configurations between 2S and 6S. All batteries surveyed have 1P configuration.

Eq. (4) is the map between battery capacity and mass, assuming 1P configuration, calculated by consid-
ering approximately 30 common LiPo batteriesc . For the same chemistry, the mass of the battery is more
or less linear with number of cells contained within it:

mb = (p1 s + p2 )C, g (4)

where p1 is 0.026373 and p2 is 2.0499e-05. Gur10 also studies the relationship between battery capacity and
mass.
As with the other electrical components, the battery’s total internal resistance must be estimated. Am-
patis9 shows the calculation of resistance of the battery Rb . This will also be used in the electrical model
described in this paper.

D. Propeller
Propellers of varying compositions including nylon-plastic, carbon fiber, wood, and other plastic have also
been characterized, providing a mapping between material, radius, and mass. Data from about 30 propellers
are shown in figure 5.
The mass of the propeller is found by

mp = p1 (2R)2 + p2 2R + p3 , g (5)

where p1 is 0.08884 and p2 is 0 for wooden propellers, 0.05555 and 0.2216 for plastic, 0.1178 and -0.3887 for
nylon reinforced plastic, and 0.1207 and -0.5122 for carbon fiber.
c http://www.hobbyking.com/hobbyking/store/index.asp accessed 02/07/2015

7 of 22

American Institute of Aeronautics and Astronautics

 100 CF
Nylon
Plastic
80 Wood

propeller mass, g
60

40

20

0
0 10 20 30
propeller diameter, in

Figure 5. Masses of propellers of various diameters composed of four materials.

0.25

0.2

0.15
c/R

0.1

5 · 10−2

0
0.2 0.4 0.6 0.8 1
r/R

Figure 6. Non-dimensionalized propeller blade chord at radial section r/R. Data shown from GWS Slowfly
and DirectDrive propellers.2

8 of 22

American Institute of Aeronautics and Astronautics

 25

20

15

β0.6
◦
10

0
0 2 4 6 8
pitch, in

Figure 7. Twist of GWS blades at r/R = 0.6 compared to indicated blade pitch.

In addition to mass information, geometrical data are parameterized for propellers. This allows the use
of a generic, averaged propeller if a specific geometry is not available or selected. Geometrical propeller
data are taken from Brandt.2 Specifically for this paper, we will consider GWS Slowfly and DirectDrive
propellers, although other available data are also compiled and stored in separate, user selectable functions
in the algorithm. This is necessary to feed the thrust calculation loop described below, as the lift per radial
segment of the propeller is calculated based on these data.

30

20
β◦

10

0
0.2 0.4 0.6 0.8 1
r/R

Figure 8. Blade twist at radial sections r/R for nine GWS Slowfly and DirectDrive propellers.2

Eq. 6 provides the physical twist angle β at the radial segment r/R for an existing set of GWS propellers
and data from the above propellers are shown in figure 8.
3 2
β = p1 rR + p2 rR + p3 rR + p4 , ◦ (6)
where, for this subset of propellers, p1 is 30.322 p2 is -64.731, p3 is 23.008, p4 is 20.558.
These data are averaged to provide a basis function for β(rR ). This function is scaled by a function
considering the pitch of the blade. The scaling is found by investigating the effect of pitch rating on the

9 of 22

American Institute of Aeronautics and Astronautics

actual twist β(rR ) of the blade at r/R = 0.6. Pitch may be qualitatively defined as the linear, axial distance
traveled by the propeller (at rR = 0.6) should it be rotated by one revolution. This relationship between a
pitch label and the physical twist β of a propeller may be seen in figure 7 for the GWS propellers.
pp , β, and inflow velocity are calculated by the blade element and momentum theory (BEMT) algorithm
and form the angle of attack of the blade section at rR . If the flight condition includes a non-zero airspeed,
the vertical component is added to V0 to achieve the final angle of attack of the blade segment. Lastly, eq. 7
provides the chord at r/R for all nine GWS propellers described, also needed by the BEMT loop to calculate
thrust. The distinction between the Slowfly and DirectDrive propellers is evident in the figure. For this
subset of GWS propellers,
3 2
cR = p1 rR + p2 rR + p3 rR + p4 (7)
where p1 is -0.2872 p2 is -0.1637, p3 is 0.4551, p4 is 0.05648.

E. Structural
A survey of a few UAVRF vehicles was completed to parameterize the proportion of structural to GTOW.
The algorithm assumes a very light-weight vehicle to have a structure of 8% of GTOW (GTQ Mini17 ), and
a heavy-weight vehicle to have a ratio of 40% (GTQ218 ). The ratio used for the discussion below is 19%.
That is,
mc = 0.19mt (8)

III. Overview of proposed design algorithm

The presented methodology is composed of two general components. The flight time calculator, or analysis
toold , is used when vehicle parameters are the inputs, and one is interested in determining capabilities of the
vehicle. This might be used to analyze the selection of component parameters. If both the mission and vehicle
design parameters are specified, the analyzer also allows for a sensitivity analysis of design parameters, as
well as a best range and endurance analysis around the specifications. This is shown in the bottom part of
figure 9.

Figure 9. EMST overview and workflow.

The drive optimizer algorithm is so called as it assists in mapping desired mission parameters to needed
vehicle design parameters. That is, given a desired mission capability, such as a cruise for a certain du-
ration, the drive optimizer will attempt to design a vehicle’s propulsion system (as well as some chassis
d http://controls.ae.gatech.edu/dbershad/EMSTAirTimeCalculator.html

10 of 22

American Institute of Aeronautics and Astronautics

parameters such as wingspan and chord, if desired) such that the vehicle’s performance will be adequate
to accomplish the mission. Both the analysis and optimization algorithms consider aspects of the following
drive components: BLDC motors, electronic speed controllers (ESCs), propellers, and batteries. Non-drive
components considered include payload, structure, and avionics, which are used for weight budget and power
draw calculations.

A. EMST Segment analysis algorithm

The analyzer may be used to substantiate the selection of components for a UAV design. This includes
validation of the outputs of the optimizer algorithm described below. Given a propulsion system design, the
algorithm allows a user to confirm that the vehicle will be capable of the mission for which it is designed. This
is done in two ways. First, the electrical power requirements that the drive components will be subjected
to in the mission flight condition are calculated so that relevant data of the selected components may
be compared to manufacturer specified limits. This attempts to ensure that the selected components are
capable of handling loads such as current, torque, etc. Second, the time endurance of the vehicle will also
be calculated so that the user may confirm that the vehicle will be capable of completing the mission should
the drive components handle the other requirements calculated.
The calculation method also provides a maximum range analysis. The analysis will provide the best
airspeed for maximum range, and since the method is geared toward VTOL vehicles, the pitch angle from
vertical required to attain it. In addition to this, the method provides a sensitivity analysis to design
parameters, including motor Kv , propeller diameter and pitch, GTOW, battery capacity, altitude, number
of motors, wingspan, etc. This helps identify key design parameters around the desired flight condition.
The analysis algorithm performs all calculations based on the input of given or estimated characteristics
of drive system components and their masses. Outputs of the method are described below and include hover
time, design space sensitivities, drive system loading requirements in terms of RPM, torque, mechanical
and electrical power, etc. Other data are also available as outputs of the analyzer. These are shown for an
example vehicle below in table 3.

1. Analyzer hover time calculation

The analyzer algorithm first uses a loop to arrive at the linear throttle input dt required to allow the vehicle
to hover, climb, or dash as specified by the mission scenario inputs. The hover case is described first and
cases with airspeed are described below. The throttle required to hover must be sufficient such that the total
thrust calculated based on this throttle balances the weight force of the vehicle. The throttle required, along
with the inputs described below, are passed to a BEMT calculator to calculate thrust generated by each
propeller. The BEMT algorithm is also fed propeller characteristics, all of which are derived from models
described in the parameterization and mission inputs. Throttle dt is incremented until it is sufficient to
accomplish the mission segment.
The vehicle’s total mass must either be provided or estimated. The estimation of component weights is
described in detail in the parameterization section. Eqns. 1, 2, 4, 5, and 8 are used to find the GTOW
mt . The mt , number of motors Nm , motor Kv , number of propellers Np , number of blades per propeller
B, propeller radius R, propeller pitch pp , and current analyzer loop throttle command dt are passed to the
BEMT algorithm to calculate the thrust generated by all propellers. Other inputs include results of eqns.
6, 7, the no-load current of the motor I0 , current draw of other power sinks Ia and Ip , propeller material,
battery configuration, composition, and qualitative quality, flight altitude and associated air densitye .
Together with the data above, a section lift and torque are calculated using the BEMT equations. For
this study, the CL (α) and CD (α) are taken from the NACA 0015 profile.13 The entire blade section is
integrated and which provides the total lift and torque applied by each propeller blade. This lift and torque
is then scaled by the number of blades per propeller and the number of propellers on the vehicle.
The algorithm also allows for pitch-controlled propeller hubs. When the user selects such a hub, all hubs
on the vehicle are assumed to be pitch-controlled. The throttle required sent to the BEMT by the outer
loop is turned into a blade pitch command, subject to the maximum allowable pitch specified by the hub
geometry. The blade planform is also selectable to be of a more traditional helicopter blade style, where
β(rR ) and c(rR ) are assumed constant. Also in such a case, a governor may be enabled which controls the
e calculated using the ISA standard atmosphere

11 of 22

American Institute of Aeronautics and Astronautics

PWM command to the motor, thereby ensuring that the RPM at the head does not exceed a specified value.
Otherwise, a headspeed may be specified for the BEMT calculations.
The algorithm has limited capacity for lifting bodies. Three types of lifting bodies are considered. First,
the body of the vehicle has a selectable CL (α) and CD (α). If data or an estimate are available, this may be
adjusted or ignored for the purposes of all calculations. The second and third types are wings. For these,
the algorithm accepts CL (α) and CD (α). Wings may be either fixed to the body via an incidence angle, or
free/actuated and controlled to a specified α.

2. Electrical modeling
The algorithm calculates the power and current required to hover. It does so by summing the torque
components on all of the propeller blades on the vehicle. Standard equations of power are applied to all
components shown in figure 10.

Figure 10. EMST propulsion system electrical model.

Assuming a steady hover, the mean voltage seen by the motor coils Vs of one motor is

Vs = dt sVc − Id (Rb + Rw + Re + Rm ) (9)

Drive current Id is the sum of current for all motors, and Im includes the no load current I0 for each
motor. It is also important to account for the change in internal battery resistance Rb as current demand
changes. Current pulled through this resistor will generate heat and this power must be accounted for in the
analysis. The effect is described by Stepaniak.20 For all analyses, ESC efficiency ηE is assumed to be 97%f .
Wiring resistance Rw is taken from standard copper wire, assumed to be of a particular gauge and length,
scaled with the current requirement Id . Internal resistance of the battery Rb is scaled with the qualitative
condition of the battery specified by the user and as described by Ampatis.9
Motor RPM is calculated based on the voltage seen by the motor coils. This is sent to the BEMT for
the torque calculation.
RP M = Kv Vs /G (10)
This no-load RPM is for directly driven or geared drive trains. It is important to note that the actual
RPM will be lower as load is applied to the motor. The reduction in RPM depends on amount of torque
applied, and generally for the hobby class of propellers discussed here, RPM has been found to decay at a
rate of approximately 15-20% per blade added to the propeller. Current increases by the same amount given
the same dt . Gear efficiency ηG is also considered at this point, i.g. about 3% increase in power required for
a spur gear. Helicopter governed head speeds may also be specified as applicable. The motor current Im is
calculated using the BEMT output of torque:

Id = Nm (Qw/(ηM Vs ) + I0 )/ηE (11)

Drive, payload, and avionics currents are summed to arrive at the total current requirement in the given
condition.
It = Id + Ip + Ia (12)
f Scorpion Calc v3.37: http://www.scorpionsystem.com/downloads/ accessed 01/13/2015

12 of 22

American Institute of Aeronautics and Astronautics

 This should include other power sinks such as tail rotors for helicopters (around 5%1 ), coaxial propeller
losses (around 15%1 depending on separation), etc. To calculate the flight time, the algorithm compares the
requirement It and the battery capacity. A discharge percentage less than 100% is generally used and may
be specified by the user to elongate the life of the battery.

3. Flight time calculation, non-hover

Other forces are also considered for non-hover cases such as climbing or cruising. These forces include lift
and drag of fuselage and lifting body components. Starting at a vertical thrust axis aligned with gravity, the
algorithm uses an iterative loop to calculate the angle required to achieve a desired airspeed. Constraints
are set so that the vehicle does not lose altitude during the cruise. A climb rate may also be specified. The
method considers lifting bodies and control, their incidence angles, as well as controlled wings that keep a
constant angle of attack with the free stream. For each α, the vehicle’s total L and D is calculated. The
loop quits when the total lift is greater or equal to the weight force of the vehicle.
In the outer, throttle required loop, the angle calculated by the above loop is used to add axial flow into
the propellers. V∞ , the axial component of the TAS, is added to the V0 component in the BEMT calculator.

4. Range calculation
The analyzer algorithm attempts to calculate the vehicle’s time endurance for a range of airspeeds. Using
these with the angular iteration above, a maximum range is available for each condition, providing roughly
”optimal” airspeeds for maximum range and maximum endurance. Of course, this is range in the wind
reference frame, so any wind would have to be taken into account for planning purposes.

5. Sensitivity analyses
The analysis algorithm calculates hover time sensitivities to design parameters. The sensitivities allow a user
to incrementally move around the design space and evaluate the effect of the parameter shift in question on
the hover time capability of the vehicle. These parameters include commonly changed ones such as motor
Kv , propeller pitch and radius, GTOW, battery capacity, and hover altitude. Hover altitudes use both MSL
and AGL, the latter for hover in ground effect (HIGE).
For each sensitivity, a linearisation of values, for example, dt/d Kv , are derived around the set point,
where dt is the change in hover time and d Kv is the desired Kv increment. The analyzer is rerun changing
Kv +/- d Kv . The two resultant values are then used to calculate a linear dt/d Kv at the set point. Of
course, as the increment becomes larger, the linear assumption deteriorates, but should be sufficient for most
conditions with ”small” increments. For battery capacity and propeller radius, the algorithm will calculate
additional masses resulting from adding/subtracting increments of both parameters. Sensitivity analyses for
several parameters are shown below for two vehicles at the UAVRF.

6. LiPo cold degradation

Battery performance is affected by the temperature in which the battery is operating. Operation in cold or
hot environments drastically affect15,16 the performance of LiPo and LiIon batteries. Some manufacturers
claim that the optimal temperature for LiPo operation is above 68-70 ◦ F g . LiIon data suggests 77-100 ◦ F is
even more ideal. h . While extensive thermal range testing of batteries has not been studied by the authors,
they were able to test several batteries in around -20 ◦ F temperatures. Using the performance data that
were collected during these flights, and anecdotal data acquired on online user groups, a simple performance
degradation factor minimized at 70 ◦ F is created as a crude way to account for extreme temperatures. Both
the capacity and voltage of the battery are reduced exponentially at different rates far away from this optimal
temperature. These trends may be observed in the LiIon data sheet.
g http://batteryuniversity.com/learn/article/discharging_at_high_and_low_temperatures accessed 11/01/2015
h http://industrial.panasonic.com/lecs/www-data/pdf2/ACA4000/ACA4000CE417.pdf accessed 11/01/2015

13 of 22

American Institute of Aeronautics and Astronautics

7. Ground effect
A ground effect correction1 is applied to all flight calculations. The power required to hover is scaled by the
HIGE correction described, which is a function of the AGL altitude of the vehicle, which is assumed to be
the same as separation of the propeller disc and the ground.

8. BEMT inflow model

The BEMT model runs in two modes; one considering inflow and one without. Inflow velocity is calculated
by iterative means. For general, hobby-sized vehicles in the range of 0-2 kg, the flight times only vary about
1-2% between the two BEMT solutions whereas the computation time may vary up to +20% for each loop.
As it is not readily clear which is more accurate considering other sources of error, it is recommended that
the inflow model not be considered for most calculations unless forward flight is required.

9. Major error sources

One of the greater sources of error for this methodology implementation is the propeller airfoil assumption
used when a specific propeller’s geometry is not used. The best case is of course using the geometry of
the propeller to be flown. For example, in all analyses described in this paper, the NACA 0015 airfoil is
for all BEMT calculations. This airfoil, of course, is generally a poor assumption, especially for propellers
with higher pitch values. This is due to the zero camber of the symmetric NACA 0015 as compared to the
non-zero cambers of most propellers available in the hobby market. Bohorquez14 shows that Cp /Ct of a
propeller can vary greatly when a non-zero camber is used as opposed to when the airfoil is symmetric. This
may greatly change the thrust and power calculations produced by the BEMT.

B. Drive optimizer
The drive optimizer is a proposed method to design and optimize the propulsion system of generic sin-
gle/multirotor vehicles. The output of the methodology is generally a propulsion system which will allow a
vehicle to accomplish a specific mission. Depending on the objective of the optimizer, the vehicle is sized for
either lightest weight, smallest wing, highest efficiency, highest climb rates, or other performance objectives.
The propulsion system which is sized consists of a motor, propeller, battery, and ESC. The work thus far
focuses on electric vehicles, although the process is valid for other energy sources with some modifications.
The methodology is developed to allow a designer to generate a vehicle that will be capable of completing a
specific mission.
As seen in figure 11, until triggering a termination condition, the algorithm will loop through potential
propulsion and other parameters (e.g., Kv , pp , S, etc.) to find acceptable solutions to the mission problem
requirements. The algorithm will then select the best configuration depending on desired optimization
parameters. For example, in the default mode (seen in the figure), if the lightest possible vehicle is desired,
then the configuration with the lightest GTOW will be selected. The algorithm estimates the masses of
all subsystems per iteration using the characterizations of the drive, structural, and electrical components
discussed in previous sections. This is done for all components for which mass is not provided, i.e., for
components that are not specifically selected by the user. For each iteration, the sum of all drive and
structural components is used arrive at a total gross takeoff weight (GTOW).
To save time, the loop termination condition will be set to exit the loop when a feasible configuration is
found while increasing incrementally the allowable, available mass. This ensures the lightest configuration
is found, within the range of the mass increment, and of course, other calculation errors.

1. Feasibility of optimizer solutions

For all calculations, the method abstracts the drive components; that is, it does not need to know the specific
motor, propeller, ESC, and battery used, and instead relies on their parameters. To illustrate this, consider
a motor. The method does not need to select a specific motor from a database for calculation of capabilities
and instead uses standard motor parameters (ie, Kv , I0 , etc). However, because of this abstraction, motors,
propellers, ESCs, and batteries of either the analyzer or the optimizer’s resulting configuration need to be
checked before implementation. For instance, the motors are assumed to be able to handle the required
current for hover or the specified flight condition for a particular mission. The manufacturer’s specifications

14 of 22

American Institute of Aeronautics and Astronautics

 Optimizing sufficient propulsion system design for minimum mass

g
ρ
γ
c
mr > ma Exit function
Vc No
R
NM
S

Yes

increment
find θ,Tr
ma
P
mr = (mb , mm , mw , mc , 2mp , mpay , mavn )
Subsystem

θ find required calculate

Tr RPM component
masses

Yes
increment
pp
No design
R
feasible
Nm
S

Figure 11. Optimizer overall flowchart. This particular optimizer is set to find the lightest possible vehicle to
accomplish the specified mission segment.

15 of 22

American Institute of Aeronautics and Astronautics

of maximum power of motors must be confirmed to be higher than the power required for the model. In
addition to this, propellers and structural components are assumed to be able to handle dynamic and static
forces experienced in flight without yielding.

IV. Results
A. Validation of analyzer algorithm
A survey was done in an attempt to validate the modules (BEMT, component parameterization functions,
etc.) that are used in both the analyzer and optimizer. The survey compares reported maximum hover or
flight times and those calculated by the analyzer. Table 1 shows results from vehicles that were tested. For
each vehicle to be tested, the vehicle’s parts characteristics must be available. All vehicles in the table have
corresponding characteristics including, at a minimum, motor Kv and number of motors, propeller diameter
and pitch, battery configuration and capacity, and GTOW, all available from each respective source. Of
course, the nature of the survey brings with its data a certain anecdotal aspect, although with a large
enough sample size, results should be valuable. Several of the vehicles surveyed are part of the UAVRF fleet
and the authors have averaged endurance data from hundreds of flights. Of the vehicles surveyed, the mean
absolute error between reported and analyzer-calculated flight times is 5.7%.

1. Application of the EMST algorithm

The optimizer and analyzer tools were used in the design of several vehicles for the UAVRF. This discussion
will focus on one in particular. The GTQ-Mini17 ”Frobenius” is the winner of the 2015 AHS MAV Challenge
held in Virginia. GTQ Mini was built specifically for the competition by the Georgia Tech Aerial Robotics
(GTAR) team. The vehicle was to maneuver in a GPS-denied environment with no external navigation
aides and was to comply with a 500 g maximum GTOW. GTAR decided to employ a vision algorithm which
required a heavy, i7 processor to deploy successfully. The computer used requires an average 3 A draw at
around 16 V (4S). To select the drive system, the drive optimizer was used. The results of the optimizer are
seen in table 2. R was constrained to 2.5 inches due to physical size limitations imposed by the competition
and design. The other inputs of relevance are a 10 minute desired hover endurance and a maximum weight
of 500 g.
The Multistar 1704, 1900 Kv motors fit the weight and power budget and although they do not match
closely with the optimizer output, they at least come the closest of the motors which were available to
GTAR at the time. Note the increase in Id when switching from 1383 to 1900 Kv . These motors were
tested to ensure that they are an appropriate choice. Because the EMST tool was largely untested at the
time of design of this vehicle, combinations of 3S and 4S battery, 5030 (shorthand for 5 in diameter, 3 in
pitch), 5030x3 (3-bladed), and 6030 propeller configurations were tested. The combination with the best
balance of thrust, flight time, and motor temperature was 4S 5030 for this motor. In contrast to what is
shown by Mulgaonkar,19 this vehicle draws about 65% of the total power for propulsion in hover due to the
power-hungry i7 computer and small motors.
Table 3 shows the analyzer outputs for the vehicle. The 850 mAh batteries keep the vehicle in the air
for about 6.5 minutes and in weight for the competition, although the 1300 mAh battery used for testing
keeps it in the air for about 9-10 minutes, depending on allowed level of battery discharge.
A maximum range analysis is also performed for the vehicle assuming perfect 850 mAh batteries and
HOGE (hover out of ground effect), shown in table 7. These data have not been validated in flight test.
Table 3 also predicts the propeller RPM during hover to within the resolution of the tachometer used to
measure the RPM during an actual flight.
As seen in table 3, the analysis method predicts an Id of 4.49 A for hover. When added to the i7 Ip
of about 3 A during heavy calculations, the total current required to hover It becomes about 7.5 A. This
matches well to what is observed when a current meter was used during a hover flight. Table 4 shows other
compared values from the actual vehicle.
Note the algorithm’s estimates for all components but payload. In this case, the payload is the i7 computer
and related avionics, which has an installed weight of about 175 g including a RAM module, USB serial
devices, and WiFi module with antennae. Table 4 shows the estimates as compared to actual measured
weights used on the competition vehicle. Other than that and the structural component, the predictions
match closely to what is observed. The structural component is off because eq. 8 assumes that the structure

16 of 22

American Institute of Aeronautics and Astronautics

 Table 1. Flight time calculation validation survey

Type Vehicle Reported endurance Calculated endurance ∆

min min %
QR GTQ Mini 5.5a 5.8 -5.5
QR Quadshot (hover) 7b 9.2 -31.4
QR HS 250 7.5a 6.9 8.0
COAX Helibaby 8c 7.8 2.5
QR BlckBd QR 8.5a 8.4 1.2
QR Blade Nano QX 10d 9.5 5
QR KMEL500 11e 11.6 -5.5
QR DIYD QR 1 (Gstv) 12f 12.1 -0.8
SMR T-REX 600e 13g 12.8 1.5
OR 3DR X8 15h 14.9 0.7
BWB Quadshot (cruise) 15i 14.9 0.7
QR DJI Inspire 15j 14.8 1.3
HR MB Hexa 15.5a 16.7 -7.7
QR Hummingbird 18.3k 18.9 -3.3
QR 3DR Iris+ 19l 18.2 4.2
QR DJI Phantom 1 25m 25.5 -2
QR QR 4 73.5n 72.3 1.6
QR QR 3 87n 73.5 15.5
QR QR 1 97.1o 101.7 -4.7
QR QR 5 109.7n 107.5 2
QR QR 2 129.2n 102.5 20.7
mean err, %: 0.4
mean abs err, %: 5.4
a http://www.uavrf.gatech.edu/ accessed 01/22/2015
b http://thequadshot.com/w/index.php?title=Quadshot_Technical_Specifications accessed 02/19/2015
c http://www.tech-sov.com/news_13.htm accessed 02/22/2015
d http://www.bladehelis.com/Products/Default.aspx?ProdID=BLH7600 accessed 02/18/2015
e http://kmelrobotics.com/ accessed 02/19/2015
f http://diydrones.com/group/arducopterusergroup/forum/topics/new-vehicle-design-tool?xg_source=activity ac-
cessed 09/17/2015
g http://rc.runryder.com/helicopter/t669735p1/ accessed 02/19/2015
h http://3drobotics.com/x8/ accessed 02/19/2015
i http://thequadshot.com/w/index.php?title=Quadshot_Technical_Specifications accessed 02/19/2015
j https://www.dji.com/product/inspire-1 accessed 06/06/2015
k http://www.asctec.de/en/uav-uas-drone-products/asctec-hummingbird/ accessed 03/14/2015
l https://store.3drobotics.com/products/iris accessed 02/19/2015
m http://www.dji.com/product/phantom accessed 07/22/2015
n http://www.rcgroups.com/forums/showthread.php?t=1880665 accessed 01/22/2015
o http://diydrones.com/profiles/blogs/my-97minute-06second-record-quadcopter-flight accessed 01/22/2015

17 of 22

American Institute of Aeronautics and Astronautics

 Table 2. Drive optimizer outputs for GTQ Mini design.

Parameter Value
s 4
Kv , RPM/V 1383
C, mAh 936
R, in 2.5
pp , in 3.0
Ωhover , RPM 11600
Id , A 3.52
mt , g 496.3
mb , g 104.7
mm , g 116.1
me , g 37.9
mp , g 11.6
mpay , g 165
mavn , g 20
mw , g 13.5

will be less efficient in terms of weight than what was produced in order to be in weight for the competition.
A sensitivity analysis is also performed at the design point. These parameters are varied with all other
parameters held constant. It is possible to run this with an MDO wrapper to provide an indication of where
design efforts should be concentrated. Table 5 presents the inputs taken by the algorithm, whose outputs
are shown in table 6.
According to the algorithm, the vehicle, around the setpoint described, would benefit most in terms of
flight endurance by increasing the battery capacity. Increasing any other value would reduce the flight time
and vice versa. Although this has not been rigorously validated and is, of course, a linearization, the change
of battery from 850 to 1300 mAh increases flight time by about four minutes, roughly equivalent to the
algorithm’s prediction of 0.84 min/100 mAh. That is,

(0.84 min/100 mAh)(1300 − 850 mAh) = +3.78 min predicted (13)

which is approximately correlates to what is observed: approximately a four minute increase in endurance.
Physically changing the other parameters has not yet been tested. Flights are terminated when any battery
cell reaches 3.6 V (loaded) to extend battery life.

V. Conclusion
A method for design and validation of design for a generic UAV electric drive system has been presented.
The method is based on designing a vehicle for a particular mission. The method considers geometric char-
acteristics of propellers, aerodynamic, and electrical aspects of the system to be designed. Two algorithms
are discussed to map bidirectionally between a mission profile and a set of UAV drive system’s parameters.
While flavors of the mission analyzer algorithm exist, the algorithm presented is more general than most in
terms of vehicle configurations. Also, a sensitivity and range analysis are provided. The sensitivity analysis
allows a designer to study the effect of nudging the design space on flight performance. The range analysis
provides insight into the maximum range and best calculated flight speed/attitude to achieve it. Other than
the work presented here, algorithms for the reverse direction, from mission to components, are not readily
available.
Other optimization design methods, described above, require a database of components to calculate
performance characteristics of a vehicle. The presented algorithm abstracts specific component databases
into relevant parameters, eliminating the need for database upkeep and data mining. In addition to this, the
algorithm is directly set up for an MDO approach to enhance the optimizer’s performance. The algorithm

18 of 22

American Institute of Aeronautics and Astronautics

 Table 3. Analyzer general system outputs for GTQ Mini at hover

Battery C required 6 C Component total: motors 64 g

Throttle / collective required 41.25 % ESCs 62.29 g
Drive current required 4.49 A Batteries (drive) 95.39 g
Motor electric power / motor 17.93 W Payload 0 g
Motor mechanical power / motor 12.91 W Structural 95.41 g
Efficiency: specific thrust / motor 7.01 g/W Wiring 24.95 g
Battery temperature coef. V 100 % Avionics 50 g
Battery temperature coef. C 100 % Propellers 4.68 g
Vehicle pitch angle 0 ◦ GTOW 499 g
Propeller speed 11599.5 RPM 80%, 100% dt
Motor speed 11599.5 RPM Thrust/weight 3.79, 5.92 -
Tip speed 77.13 m/s Specific thrust 2.64, 2.83 g/W
Tip Mach number 0.23 Mach Current/motor 2.80, 4.09 A
Motor torque / motor 0.00352 Nm Motor electric power 44.73, 65.26 W
Disc loading 9.85 kg/m2
Power loading 28.07 kg/kW
Max governor 0 %
Blade pitch 3 ◦,in
Mixed flight time 4.08 min Hover time 6.81 min

then may easily be used to find propulsion systems with high loiter times for endurance or high thrust-to-
weight ratio for adeptness in acrobatics and high rate of climb. Allowing this departure from component
databases is the key to a truly optimized drive system, and potentially a faster and less limited path to it.
That is, for the former, the algorithm does not need to cycle through the data from thousands of motors in
the database to find an appropriate one. For the latter point, a database-based optimizer is limited to a finite
number of solutions, directly proportional to the number of components stored. The presented algorithm
should thus be able to traverse a greater design parameter space than any that relies on a database.
The new design algorithms are applied to several vehicles, and one such application has been described.
The GTQ Mini is the winner and only finisher of the autonomous category in the 2015 AHS MAV Challenge.
The main outcomes related to this work were described in the results section. The algorithm was able
to accurately predict required RPM, throttle, drive current, flight time, component weights, and battery
capacity sensitivity for the vehicle. Other data have been outputted as well but have not yet been validated.
The system has also been run against a multitude of existing vehicles and found to predict their reported
flight times to within 5% on average. Vehicles which were studied include helicopters, multirotors, and a
blended wing-body VTOL vehicle.
To allow these algorithms to function, a parameterization of motors, ESCs, batteries, and propellers
has been presented. A trend between motor’s speed constant and its mass was found. This allows for the
prediction of a motor’s mass based on knowlege of Kv . Similarly, prediction of a battery’s mass is possible
based on knowlege of its composition and configuration. The drive current required allows for predicting the
mass of the ESC and wiring, and the diameter and material of a propeller blade is an indicator of its mass.
A simple electric propulsion system model has been presented. One area for improvement is in the motor
model. More work needs to be done to model the RPM performance of a loaded motor, especially when
B is greater than 2. Also, current limitations of motors should be parameterized to throw away impossible
configurations when optimizing the drive system. Another such area is the battery model which does not
include Vc sag dynamics. When a LiPo battery is stressed, the cell voltage may drop abruptly, especially
under 3.5 V for the batteriesi used by the UAVRF.
The aerodynamics model of the propellers would benefit from knowlege of a more accurate airfoil of
i Various 3S Thunderpower, Zippy, and Turnigy Nano batteries between 20-90 C

19 of 22

American Institute of Aeronautics and Astronautics

 Table 4. Analyzer predictions from table 3 compared to actual measurements for GTQ Mini in HOGE.

Parameter Predicted Measured

RPM 11600 11560 mean
It , A 7.49 7.8 mean
Motors, g 64 68
ESC, g 62.29 38
Battery, g 95.39 96
Structural, g 95.41 41
Wiring, g 24.95 20
Avionics, g 50 47
Propellers, g 4.68 6.8
Payload, g - 175

Table 5. Sensitivity analysis inputs for GTQ Mini at hover.

Parameter +/ − ∆
Kv , RPM/V 100
Prop diameter, in 0.5
Prop pitch, in 0.5
GTOW, g 50
Capacity, mAh 100
Altitude, m 3000
TAS range, m/s 0-15.24

the propellers modeled instead of the NACA 0015 used, which is symmetric. The ground effect correction
applied is only valid in a hover case and needs to be updated to include downwash from wings and forward
flight effects. Tip losses are also currently not modeled.

Acknowledgments
The authors would like to thank the members of the UAVRF and the RC flight community for their help
with this project.

20 of 22

American Institute of Aeronautics and Astronautics

 Table 6. Sensitivity analysis outputs for GTQ Mini at hover

Sensitivity parameter Specific sensitivity Total sensitivity

dt/d Kv -0.0018 min/Kv -0.18 min/100Kv
dt/d D -0.66 min/in -0.33 min/0.5in
dt/d P -1.29 min/in -0.64 min/0.5in
dt/d m -0.01 min/g -0.33 min/50g
dt/d C 0.01 min/mAh 0.84 min/100mAh
dt/d Alt 0.000393 min/m -1.18 min/3000m

Table 7. Maximum range analysis outputs for GTQ Mini at various TAS.

TAS Pitch dt Id Endurance Range

◦
m/s % A min km
0 0 39 4.24 7.18 0
1.52 9.1 40 4.37 6.99 0.64
3.05 14.6 41.3 4.58 6.7 1.22
4.57 18.6 43.4 4.97 6.22 1.7
6.1 22.1 46.2 5.54 5.63 2.06
7.62 25.6 49.7 6.28 5.01 2.29
9.14 28.1 53.4 7.12 4.46 2.45
10.67 30.6 57.7 8.13 3.94 2.52
12.19 33.1 62.7 9.22 3.5 2.56
13.72 35.1 68 10.33 3.14 2.58
15.24 37.1 73.7 11.21 2.9 2.65

21 of 22

American Institute of Aeronautics and Astronautics

 References
1 Leishman, J.G., Principles of Helicopter Aerodynamics, Cambridge University Press, New York, 2006.
2 Brandt, J., Selig, ., ”Propeller Performance Data at Low Reynolds Numbers”, 49th AIAA Aerospace Sciences Meeting,
2011, AIAA 2011-1255. http://www.ae.illinois.edu/m-selig/props/propDB.html. accessed 01/30/2015.
3 Harrington, A. M., ”Optimal propulsion system design for a micro quad rotor,” M.S. Thesis, Department of Aerospace

Engineering, University of Maryland, College Park, MD, 2011.

4 P.R. Bates, D.P. Schrage, The Configuration and Conceptual Design for Rotary Wing Aircraft”, Proceedings of the

AIAA/AHS/ASEE Aircraft Design Systems and Operations Meeting, St. Louis, Missouri, USA, September 14-16, 1987, pp. 10
5 Schrage, D.P., 2013, Lecture Notes on Rotorcraft Systems Design, Vehicle Synthesis for Advanced VTOL Aircraft, School

of Aerospace Engineering, Georgia Institute of Technology.

6 Khalid, A. ”Development And Implementation Of Rotorcraft Preliminary Design Methodology Using Multidisciplinary

Design Optimization,” Ph.D. Dissertation, Department of Aerospace Engineering, Georgia Institute of Technology, Atlanta,
GA, 2006.
7 Latorre, A. M., ”Propulsion system optimization for an unmanned lightweight quadrotor,” M.S. Thesis, Department of

Aerospace Engineering, Universitat Politcnica de Catalunya, Catalonia, Spain, 2011.

8 Lundström, D., Amadori, K., Krus, P., ”Automation of Design and Prototyping of Micro Aerial Vehicle”, AIAA-2009-629,

47th AIAA Aerospace Sciences Meeting, Orlando, FL, USA, Jan. 2009.
9 Ampatis, C., Papadopoulos, E. ”Parametric Design and Optimization of Multi-Rotor Aerial Vehicles.” In Applications of

Mathematics and Informatics in Science and Engineering, pp. 1-25. Springer International Publishing, 2014.
10 Gur, O., Rosen, A., 10-12 September 2008, ”Optimizing Electric Propulsion Systems for UAVs,” AIAA 2008-5916. 12th

AIAA/ISSMO Multidisciplinary Analysis and Optimization Conference 1em plus 0.5em minus 0.4em Victoria, British Columbia
Canada.
11 Bouabdallah, S., Siegwart, R., ”Design and Control of a Miniature Quadrotor,” Advances in Unmanned Aerial Vehicles,

Springer Netherlands, 2007.

12 Benito, J.A., et al. ”Design considerations of a small UAV platform carrying medium payloads.” Design of Circuits and

Integrated Circuits (DCIS), 2014.

13 Sheldahl, R E, and P C Klimas. 1981. ”Aerodynamic Characteristics of Seven Symmetrical Airfoil Sections through 180-

Degree Angle of Attack for Use in Aerodynamic Analysis of Vertical Axis Wind Turbines”. United States. doi:10.2172/6548367.
14 Bohorquez, F., Pines, D., Samuel, P.D., ”Small Rotor Design Optimization Using Blade Element Momentum Theory and

Hover Tests”, Journal of Aircraft, 2010 Vol. 47, No. 1, 268-283. DOI: 10.2514/1.45301
15 Tarascon, J.M., ”Issues and challenges facing rechargeable lithium batteries,” Nature Publishing Group, 2001.
16 Pesaran, A.A., Vlahinos, A., Burch, S.D., ”Thermal Performance of EV and HEV Battery Modules and Packs,” Proceed-

ings of the 14th International Electric Vehicle Symposium, Orlando, Florida, December 1517, 1997.
17 Haviland, S., Bershadsky, D., Magree, D., and Johnson, E.N., ”Development of a 500 gram Vision-based Autonomous

Quadrotor Vehicle Capable of Indoor Navigation,” Proceedings of the AHS International 71st Annual Forum and Technology
Display, May 2015.
18 Magree, D., Bershadsky, D., Haviland, S., et al, ”Georgia Tech Team Entry for the 2012 AUVSI International Aerial

Robotics Competition,” Fourth Symposium on Indoor Flight Issues, August 2012.

19 Mulgaonkar, Y, et. al, ”Power and weight considerations in small, agile quadrotors,” Proc. SPIE 9083, Micro- and

Nanotechnology Sensors, Systems, and Applications VI, 90831Q 2014.

20 Stepaniak, M. J., Van Graas, F.,Uijt De Haag, M., ”Design of an Electric Propulsion System for a Quadrotor Unmanned

Aerial Vehicle”, Journal of Aircraft, 2009 Vol. 46, No. 3, 1050-1058. DOI: 10.2514/1.38409.

22 of 22

American Institute of Aeronautics and Astronautics