Académique Documents
Professionnel Documents
Culture Documents
DOI 10.1007/s10846-011-9616-y
Received: 15 February 2011 / Accepted: 15 June 2011 / Published online: 9 September 2011
Springer Science+Business Media B.V. 2011
1 Introduction
The United States Federal Aviation Administration (US FAA) defines an Unmanned Aircraft System (UAS) as a collection of Unmanned
Aerial Vehicles (UAVs), control station(s), command and control algorithms and equipment for
launch, recovery, communication and navigation
[1]. While many efforts have been directed to
the study of these components, there has been
564
The paper develops Petri net models of battery charging and replacement systems that
enable a tight comparison between them.
(Section 2)
The paper develops and compares design options for the functional components of battery
replacement service stations. These attempt to
address the issues of fleet heterogeneity and
UAV landing control robustness. (Section 3)
The paper reviews and conducts operational
testing of key components of a system prototype. (Section 4)
Using estimates of the replacement platform
costs based on the design choices of Section 3,
we compare the cost of various platforms and
systems. (Section 5)
2 Planning UASs
A Petri net can be considered a graphical tool
that may be used to describe distributed, concurrent, parallel, asynchronous, deterministic and/or
stochastic stepwise processes [17]. It is a bipartite graph, in which the nodes are divided into
transitions (T, represented by bars) and places (P,
represented by circles). The connection between
nodes is made by directed arcs, which connect
only a T to a P, or a P to a T, never P to P or T
to T. Tokens (usually represented by dots) travel
through the net. Whenever there is a token at the
input of all arcs leading to a transition, the transition fires, the tokens at the input are consumed
and a token is created at the output of each of the
outgoing arcs. The event of a firing can happen
concurrently and can overlap in time with other
firings so long there are enough tokens to fire
a transition. The location and number of tokens
at the start of the Petri net evolution is called
the initial marking. Times may be associated with
the transitions and places. Tokens entering a
node must wait for this duration before they are
released.
565
566
TF + TR + TI
,
NUAV
(1)
T LPLAT =
TR
,
NPLAT
(2)
T LBATT =
TC + T R
,
NBATT NUAV
(3)
T LCGR =
TC
,
NCGR
(4)
(5)
CSYS =
TF
TC +T R
T F +T R +T I
R
, NTPLAT
, NBATT
, TC
NUAV
NUAV NCGR
t+
N
UAV
.
i=1
1
t
t
0
F
where IUAV
(t) is the indicator that UAVi is in the
i
air at time t (its value is 0 or 1) and we assume the
NUAV UAVs are labelled UAV1 , ..., UAV NUAV .
ti +T
ti
F
IUAV
(t) dt,
i
1
(q T F )
(q NUAV )TCYC
= NUAV
1
T
where ti t0 is the time of the firing of transition T F corresponding to the first launch of
UAVi once the periodic regime has been reached.
Since T = ( p K)TCYC = (q NUAV )TCYC , assuming FIFO consumption of the tokens from the
Ready to Fly place, U AVi will start and complete exactly q flights in the time interval [ti , ti +
T ]. Thus, the integral for each UAV will give
q T F . We have:
i=1
max
N
UAV
i=1
CSYS =
TF
TCYC
CSYS = lim
TF
TF
=
.
NUAV TCYC
TCYC
by dictating required battery storage space, number of charge platforms and overall system speed.
Example 1 Consider a UAS consisting of 3 UAVs
and one exchange (replacement) platform with
T F = 15 min, T R = 1 min and TC = 85 min. Set
T I = 0. Let T S denote the duration of a UAV
operation cycle, that is, T S = T R + T F + T I =
16 min.
Let NUAV = 3, NBATT = 20, NPLAT = 1 and
NCGR = 16. For such a system, the replacement Petri net can be used to obtain T LUAV =
T F +T R +T I
R
= 5.333, T LPLAT = NTPLAT
= 1, T LBATT =
NUAV
TC +T R
C
= 5.059, and T LCGR = NTCGR
= 5.313.
NBATT NUAV
TF
Thus, TCYC = T LUAV and CSYS = TCYC = 2.8125
UAVs/unit time. If we want a lower coverage
with the same system, say 2.5,
Rwe can increase
N
CUAV
1 =2
T I . Setting T I = (T F + T R ) UAV
TGT
CSYS
min, we obtain TCYC = 6 min. We then have
CSYS = 2.5 UAVs/unit time.
3 Design Options
In this section we discuss design options for an
autonomous UAV battery replacement system
ground station. This station is to automatically
swap the depleted batteries of a UAV for a
recharged battery without human intervention.
Here, instead of tackling navigation control algorithm problems, we work with design parameters
to solve issues such as:
567
568
Fig. 2 Communication network topology between chargers and main control unit
569
570
Figure 4 depicts three different skid configurations that can be used to guide the UAV to-
The battery housing has terminals that can couple with the UAV in order to power it and also
has connections to match with the charger in the
ground station. A possible design of a battery
case is depicted in Fig. 5. Projection views of a
battery case are shown on the left and the isometric view of a possible skid assembly can be
seen at the right. At the bottom and at the right
views of the battery case, ferromagnetic plates are
placed with the intent of using an electromagnet to move it in other systems (see Sections 3.6
and 3.7).
571
572
Magnets are very versatile, but they have disengagement problems against shearing forces and
impact. Also, if they are used as terminals, they
may attract metallic bodies that may cause shortcircuits. Adding magnets to all batteries in a system is a good idea to ensure terminal connection
and locking. However, it may cause problems in
battery transportation because the batteries may
attract each other and cause unwanted terminal
contact (which can be very undesirable for Li-
573
574
The battery capturing system has only one function: prevent the battery from movement in any
direction relative to the capture device or grabber. We considered a mechanical claw, servoassisted magnets and electromagnets. Mechanical
devices require guidance and positioning in order
to clamp the battery and pull it out of the UAV.
This sort of system demands use of physical space
to open and close arms, has many moving parts
(is thus more prone to failure) and may require
tight tolerances to ensure assembly and guide
movements. Electromagnets could also be used as
a capturing system. They are easy to control and
have high repeatability, although a ferromagnetic
plate must be attached to the battery case, increasing on board weight. A servo-assisted permanent
magnet was used in [4] to extract the battery from
the UAV. It is also an option to use magnets on
both sides of the UAV/battery interface. Such a
system does not use a servo-motor at all times,
and so, small amounts of current are required.
However, it has moving parts and magnets may
attract foreign bodies, causing short-circuits in the
system.
We propose a capturing system with one electromagnet to remove the battery case from the
UAV. Such a system must be tested to ensure no
interaction with UAV guidance systems.
In Fig. 9a, the UAV is assumed to have reached
a specific position. At the bottom of the UAV,
there is a battery case, which is being held by
either mechanical coupling or by magnets, as de-
Fig. 9 In a, an electromagnet is moved up. In b, the electromagnet reaches the battery case and is turned on, capturing the
battery from the UAV. In c, the electromagnet is pulled downward and the battery case is extracted from the UAV
575
576
Fig. 11 Magazine of batteries in a circular array of chargers actuated by rack and pinion system
577
4 Prototype
3.9 Elevator Design Option for UAV Vertical
Transportation
There are other options to deliver the battery
case to the swapping site, other than moving the
battery after extracting it from the UAV. One
is to lower the entire UAV, and at the swapping site, exchange batteries with the UAV, as
depicted in Fig. 13. Such a system may reduce
578
Fig. 13 In a, UAV is lowered to the swapping site. In b, the depleted battery is extracted straight out of the UAV. In c, a
new battery is actuated in the UAV after magazine actuation
579
No
Center UAV
Bring battery to
exchange zone
Extract battery
from UAV
Lock UAV
Insert battery
back in UAV
Unlock UAV
Inform control
That UAV is ready
Start
UAV has
arrived?
Place depleted
battery into charger
Yes
No
Place next
charged battery in
position
Yes
End
Prepare system
for next UAV
580
Start
Lift UAV
Send IR burst
No
Count received
pulses
Yes
Lower UAV
End
locking system ensures that the process of extracting the old battery and inserting a new one will not
affect the position and orientation of the landed
UAV. This is required since the battery swapping system assumes the UAV to be in a known
position to perform its tasks. The module was
implemented as a pair of servomotors each with a
Assuming that the battery was successfully removed from the UAV, brought to the exchange
zone, and that there is a charged battery already
in position to be exchanged, the battery swap
system exchanges the old battery for a new one.
The battery swap module is also responsible for
581
Fig. 17 Prototype for the lock module, orientation-fixing module and battery extraction module. In a it is possible to see the
elements of each module. b shows a UAV with the orientation already fixed and locked into the platform
582
the new battery was carried to the UAV and the battery
actuator reaches the old battery while in e the old battery
is brought to the charger
583
Time (s)
Position fixing
Orientation fixing
UAV lock
Elevator
Battery swap
Total
15.0*
13.0
1.0
6.5*
4.5
47.5
Notes
Worst case (full turn)
To be counted twice
To be counted twice
To determine the cost of a UAS requires an estimate of the number of components in the system.
From Theorem 1 we can directly obtain the number of components to achieve a target coverage
TGT
.
CSYS
Corollary 1 For a recharge platform, the minimum
number of components that will guarantee a target
TGT
coverage of CSYS
is:
TGT
CSYS
(T F + T R + T I ) ,
(6)
NUAV =
TF
NPLAT =
TGT
CSYS
TR .
TF
(7)
TGT
Proof By Theorem 1, CSYS
CSYS
= T F /TCYC.
TGT
Equivalently, T F / CSYS max T LUAV , T LPLAT
TGT
TGT
T LUAV and T F / CSYS
T LPLAT .
or T F / CSYS
Since TLUAV = (TR + TF + TI ) /NUAV and TLPLAT =
T R /NUAV , it is equivalent to require NUAV
TGT
TGT
CSYS
(T R + T F + T I ) /T F and NPLAT CSYS
T R /T F . The result follows.
RECH
REP
, VPLAT
, VUAV , VBATT and VCGR are
where VPLAT
the costs of a recharge platform, a replace platform, UAV, battery, and charger, respectively.
The UASs described here are composed of several modules that result in the final cost estimation
of this ground automation system. The cost analysis is a rough estimate. We focus our attention
on parameters specifically related to the systems
designed and shown throughout the development
of this work, which is the energy replenishment
of the UAV itself. Parameters such as the cost of
the navigation system, the external power and the
human resources are not used in the estimation.
We consider
584
Table 2 Data used to
plot the cost comparison
graph
Cost of (USD)
Battery
UAV
Charger
Refill platform (raw costsame for all)
Replace platform (raw costsame for all)
Time of (min)
Flight
Charge
Replace
Maximum values per replacement platform
Maximum coverage
Minimum chargers
same coverage, because of the relatively fast energy replenishment time (See Section 4.5).
The information obtained from the Petri nets
(see Section 2) and the experiments mentioned in
previous sections are summarized in Table 2, for
three types of UAVs.
The Cheap UAV is in the micro aerial vehicle (MAV) range, which can be a very small
UAV with the size of an insect or a UAV for
hobby purposes with estimated flight time of 10
min and battery recharge time of 25 min. Their
battery has fewer cells, thus less cost (battery and
charger costs are estimated as 5 and 30 USD
Cheap
Mid-range
Expensive
5
30
10
20
500
15
20
1,500
100
1,500
50
10
25
1
15
85
1
30
85
1
10
27
15
87
30
87
585
(10)
Positive values of Adv(C) imply that the replacement system is economically more viable
than a refill system. Figure 19 shows that the
points where replacement systems become more
efficient are C = 1.23 and C = 1.33 for midrange and expensive UAVs, respectively. On the
other hand, a replacement UAS never becomes
economically better than a refill UAS for cheap
UAVs. However, it is important to notice that the
number of UAVs required to fulfill a given target
coverage is usually greater in a refill UAS than in a
replacement UAS. This analysis does not include
control algorithm complexity nor the availability
of control channels, therefore replacement UAS
may still be better.
6 Concluding Remarks
In order to plan a new UAS or to analyze an existing UAS, we developed a Petri net model which
allows us to calculate the system coverage based
on the component parameters. UAS expansion
or resource management can be guided and/or
optimized with the aid of this model.
We also developed several module prototypes,
both conceptually and physically, which together
comprise one service station to swap batteries
with high success rates. The complete station
is able to compensate for orientation and positioning errors. It addresses navigation imprecision, weather conditions and/or UAV damage,
which are common issues in outdoor missions and
combat environments. The ground station is also
capable of handling heterogeneous UAV fleets
with not only different shapes and sizes, but also
different number of battery cells per battery pack.
Further development of such replacement stations
and replenishment stations is essential to reduce
the limitations caused by having humans in the
loop.
7 Future Work
As future work, we will conduct further testing with prototype modules of UAV positioning
systems (see Section 3.2), new skid designs (see
Section 4) and battery transportation systems
(see Section 3.7). After the prototype modules
are fully operational and reliable, the next intended step is to make an entire integrated system that can replace UAV depleted batteries for
new recharged ones autonomously upon UAV
landing.
Other future work is related to gasoline refueling ground stations, where studies will be performed with the goal of having fully autonomous
ground service stations for heterogeneous fleets.
References
1. U.S. Department of Transportation - Federal Aviation Administration: Order 8130.34 Airworthiness Certification of Unmanned Aircraft Systems (2008)
2. Dale, D., How, J.P.: Automated ground maintenance and health management for autonomous unmanned aerial vehicles. Thesis (M. Eng.), Department of Electrical Engineering and Computer Science,
Massachusetts Institute of Technology (MIT) (2007)
3. Kemper, F.P., Suzuki, K.A.O., Morrison, J.R.: UAV
consumable replenishment: design concepts for automated service stations. J. Intell. Robot. Syst. 61, 369
397 (2011)
4. Swieringa, K.A., et al.: Autonomous battery swapping
system for small-scale helicopters. In: 2010 IEEE International Conference on Robotics and Automation
(ICRA), pp. 33353340 (2010)
5. Mohammad, H.M., et al.: Vision-based landing of
light weight unmanned helicopters on a smart landing platform. J. Intell. Robot. Syst. 61, 251265
(2011)
6. Cassinis, R., Bartolini, P., Tampalini, F., Fedrigotti, R.:
Docking and charging system for autonomous mobile
robots. Tech. rep., Universit degli Studi di Brescia
(2005)
7. Kim, K.H., et al.: Development of docking system for
mobile robots using cheap infrared sensors. In: Proceedings of the 1st International Conference on Sensing Technology, pp. 287291 (2005)
8. Parker, G., Georgescu, R., Northcutt, K.: Continuous power supply for a robot colony. In: Automation Congress, 2004. Proceedings. World, pp. 279286
(2004)
9. Parker, G., Zbeda, R.: Controlled use of a robot colony
power supply. In: IEEE International Conference on
586
10.
11.
12.
13.
14.
15.
16.
17.
18.
19.
20.
21.
22.
23.