ELECTRONICS AND ELECTRICAL ENGINEERING
ISSN 1392 – 1215
ELEKTRONIKA IR ELEKTROTECHNIKA
ELECTRONICS
T170
ELEKTRONIKA
Electronic Parking Control Systems
A. Marma, M. Žilys, D. Eidukas, A. Valinevičius
Department of Electronics Engineering, Kaunas University of Technology,
Studentų str. 50, LT-51368 Kaunas, Lithuania,phone: +370 688 31454, e-mail: arunas.marma@arevita.com,
mindaugas.zilys@arevita.com, danielius.eidukas@ktu.lt , algimantas.valinevicius@ktu.lt
Introduction
Near all big shopping and leisure centres big
concentration of vehicles are allocated. These big vehicle
concentrations can cause traffic jams in the parking area
and in the areas near these objects. All this rising traffic
flows can cause: rise exhaust gases in the parking area,
bigger possibility of accidents such as fires, terror actions
and others. In these areas usually not only cars but also
peoples are constantly visiting, so if we have big vehicle
concentration in one place (especially in the closed, or
multi floor parking areas) we have to take care of area air
ventilation, lighting, heating and others.
One of the ways to help to solve these problems is
electronic control systems. To help solve these problems
we can use many different electronic control subsystems:
• Parking guidance;
• Traffic jams detection;
• Traffic jam control;
• Parking area payment;
• Intelligent parking area administration and
handling;
• Emergency situation control;
• Interface systems to city traffic control systems;
• Parking area ventilation and lighting control.
To create optimal and well running parking control
system all subsystems have to be divided to the
hierarchical levels, all functions and algorithms described
and information flows have to be distributed. Adjusted
parking control system hierarchy is shown in Fig. 1 [1,2].
Parking control system hierarchy consist of 3 layers: 1st
layer is for parking coordination - where parking control
strategy is described, 2nd layer is adaptation – where
parking area control strategy is prepared for control layer
parking area information is processed and 3rd layer is
control – traffic flows are controlled and all data is
collected.
Near recently opened big shopping and leisure centre
“Akropolis” big multi floor parking lot with 2500 parking
places was opened. To ensure best performance of this
57
2007. No. 8(80)
parking (as every day 4000-10000 cars are visiting this
parking lot) – electronic control systems was implemented:
parking guidance, traffic jams detection and control,
intelligent administration and handling, data collection,
emergency situation control and control of lighting and
ventilation. All these systems was developed, installed,
tuned, tested and experimental data collected for modelling
of processes happening in the parking areas and in near
locations of these parking-lots.
Parking guidance systems
Not only physical and architectural solutions (parking
roads width, rise and down ramps structure, parking traffic
structure and others) can optimize traffic flows in the
parking areas, but parking guidance systems also can help
on reaching these results.
For many tasks solution parking guidance system was
developed and installed. Main tasks which parking
guidance system solves is:
• Maximum decrease time of searching free
parking space;
• Reduce possibility of traffic jams in the parking
area;
• Reduce possibility of traffic jams in the roads
and intersections near parking area;
• Gradually distribute incoming cars through the
parking area;
• Give necessary information for traffic lights
systems;
• Inform about too long parked vehicles;
• Collect statistical information about object
attendance;
• Decrease quantity exhaust gas in the parking
area;
• Locate possible fireplaces in the parking;
• Decrease possibility of car accidents;
• Central and distributed parking administration
and handling;
• Control lighting and ventilation systems.
 COORDINATION 1 LAYER
Describing parking area control strategy:
•Permissive parking areas in the parking lot,
•Parking lot filling sequence and priorities,
•Parking times and prices,
•Parking flows analysis,
•And other strategies.
Control strategy data Processed parking area data

ADAPTATION 2 LAYER
Parking area control strategy preparation for control layer and
parking information processing:
•Data base,
•Parking area control strategy preparation for lower level,
•Parking data collection,
•Parking data processing,
•Monitoring of control layer performance,
•Parking data transmission and storage.
Control data Parking data

CONTROL 3 LAYER
Traffic flows control and data collection:
•Parking data transmission communication networks (Ethernet, RS485, RS232),
•Data collection devices (inductive road loops, parking spaces sensors, road barriers end swishes, etc.)
•Control devices (parking spaces indicators, parking rows displays, floor displays, road barriers, controlled
electronic road information boards)
•System control and auxiliary devices
Fig. 1. Parking control system hierarchy

Implemented parking guidance system architecture concentrators, traffic controllers, servers with databases
with hierarchy layers is shown in Fig. 2. Lowest layer: communicating to each other and to other layers through
parking space sensors and indicators, parking places Ethernet network. Highest layer is management PC’s with
displays and concentrators are communicating through application software’s .
RS485 bus. Middle layer equipment: parking

1 LAYER

Parking management PC Parking management PC

2 LAYER

Ethernet Traffic flows

Parking controller
concentrator
Server with SQL database

1st. Floor 2nd Floor 3rd Floor RS485

3 LAYER
RS485 concentrator concentrator concentrator
RS485 PX 123 > PX 123
RS485
RS485
RS485 RS485
ZC1. Zone ZC2. Zone ZC3. Zone
concentrator concentrator concentrator PX 123 > PX 123
RS485 RS48 RS4 20 parking 24 parking
RS485 row panels row panels PX 123 > PX 123
P12 > RS485 RS485
PX 123 > PX 123
P12 > 6 ZC Zone 7 ZCZone
concentrators concentrators
P12 >
RS485 RS485
P12 >
12 pcs. 12 pcs. Floor
600 park 700 park
Parking area panels
space sensors space sensors
Parking Parking Parking
Parking rows panels
panels(12pcs.)
Parking spaces spaces Parking spaces Parking spaces
sensors indicators spaces indicators spaces indicators
(127pcs ) (127pcs.) sensors (127pcs.) sensors (127pcs.)

Fig. 2. Parking guidance system architecture

58
 One of the main target for parking guidance system is
to decrease vehicle time of searching free parking space
and decrease traffic flows in the parking area. Principle of 90
parking guidance system is: in each parking space 80
ultrasonic vehicle detection sensor and LED indicators 70
60
(green/red) are installed, on each parking row, entry roads

Even ts
50 No system
and on raise and down ramps free parking places displays
40 With system
with direction arrows are installed also different counters,
30
controllers and other equipment is integrated. So when 20
vehicle arrives to the parking area driver by the parking 10
spaces displays is leaded to the parking row where is 0
enough free parking spaces and then he is directed to the <30 60 90 120 150 180 210 >240
free parking space by parking spaces indicators. In this way
Parking time, s
time of searching time and traffic flows are decreased as
drivers do not going rounds through parking searching free
parking space. An example of car cruising and searching Fig. 5. Time histogram of searching free parking space with and
without electronic parking guidance system
free parking space with and without parking guidance is
showed in Fig. 3–4.
Traffic jams control system

Next no less important parking control system target is

Fig. 3. Example of car cruising through parking without
electronic guidance system
to detect and control possible traffic jams. Parking control
systems part is traffic jams recognition and control
systems. Few types of traffic jams in the parking-lots can
be described: traffic jams on entry/exit roads, on raise and
down ramps and on roads to parking lines. Best way to
avoid such a jams is to detect not only traffic jam which is
already formed, but to detect raised possibility of forming
traffic jam. If we can detect the raised possibility of traffic
jam formation we can take actions to avoid this traffic jam
by: closing some entries exit roads, parking areas, inform
drivers to avoid some suspicious places and in this way
avoid traffic jams. So main target is to find phenomenon
appearing before traffic jam is formed and inform drivers
one step ahead.
To find such a phenomenon appearing before traffic
jam is formed in the parking-lot 12 inductive road loops
was installed. Inductive road loops are installed on the
entry/exit roads and near each rise and down ramps.
Automated measurements and registration system was
created to collect traffic data, see this system structure in
the Fig. 6. Over 500000 measurements were done during
this experiment.

Inductive road Inductive road

loop IK11 loop IK12
4th FLOOR
PLC Programmable
WAN
Inductive road Inductive road logic controller
loop IK9 loop IK10
Internet

Inductive road Inductive road

loop IK6 loop IK8
3rd FLOOR

Fig. 4. Example of car cruising through parking with electronic Inductive road Inductive road Random Access
loop IK5 loop IK7 Memory - RAM
guidance system (Buffer)
Inductive road Inductive road
loop IK2 loop IK4
Before and after parking guidance system was 2nd FLOOR
Work station with SQL
implemented experiment was done and vehicle searching Inductive road Inductive road data base
free parking space times was measured. During experiment loop IK1 loop IK3

150 measurements was done and results are shown in 1st FLOOR

Fig. 5. Results showed that average parking space
searching time without electronic parking guidance system
t p.vid .o = 84 s and with electronic parking guidance
t p.vid . p = 55s , so we have time decrease of 65%.
Fig. 6. Traffic jams detection and automated measurements and
registration system structure
Two main parameters was registering during this
experiment: time between cars passing inductive road loop

59
and time of cars passing inductive road loop (cars speed).
Two characteristic days was selected for analysis. One day Histogram
is one of the shopping and leisure centre opening days
6000
(07.04.10) when parking lot had 10200 incoming cars per
day also had formed traffic jams time to time and common 5000
working day (07.04.24) when parking lot had 4500

E vents (cars)
4000
incoming cars and no traffic jams. These two days (opening 10.Bal
day low attendance day) results histograms of two entry 3000
24.Bal
roads are shown in Fig. 7–10. 2000

1000
Histogram 0

,0

,0

,0

,0

,0

e
1200

0,

0,

0,

0,

or
10

30

50

70

90

11

13

15

17

M
1000 Time between cars, s

Fig. 10. Histogram of times between incoming cars from 2nd

E vents (cars)

800

600
10.Bal entrance
24.Bal
400 Analyzing two characteristic days collected data and
200
histograms we see that times of incoming cars passing
inductive road loop (speed) times is much bigger on the
0 day when parking lot had big quantity of incoming cars and
0,2
0,6
1,0
1,4
1,8
2,2
2,6
3,0
3,4
3,8
4,2
4,6
5,0
5,4
5,8
6,2
6,6
7,0

some traffic jams formed. So incoming cars speed when

Detection time, s
traffic jams occurring is much slower and histograms are
Fig. 7. Histogram of times cars passing inductive road loop showing it. Times between incoming cars are also different
incoming from 1st entrance on these two characteristic days: on the day when jams was
formed and big vehicle attendance was times between
incoming cars was shorter comparing to the common
Histogram
working day, so the cars are going much closer to each
800 other. Made an analysis we see that using this data we can
700
forecast possibility of traffic jam formation.
600
Process simulation
Events (cars)

500
10.Bal
400
24.Bal To recognize and forecast traffic jams in the parking-
300
lot model with neural network using Levenberg-Marquardt
200
optimization was done. For the modelling Matlab nftool
100 library was used. Structure of neural network is shown in
0 the figure 11. For training and testing this neural model 510
0,2
0,6
1,0
1,4
1,8
2,2
2,6
3,0
3,4
3,8
4,2
4,6
5,0
5,4
5,8
6,2
6,6
7,0

data sets was prepared. Each data set includes events of

Detection time, s times between incoming cars and times of car passing
Fig. 8. Histogram of times cars passing inductive road loop
inductive road loop. Half of the data sets represent events
incoming from 2nd entrance which are occurring before traffic jam is formed and half
data sets represent simple events when no traffic jam is
formed. Output data set was also prepared.
Histogram

4500
Input Hidden layer Output layer Outpu
4000
3500
W* W*
E ven ts (cars)

3000
2500 10.Bal
+ +
2000 24.Bal
1500 b b
1000
500 20 20 1 1
0
Fig. 11. Structure of neural network
,0

,0

,0

,0

,0

e
0

0
0,

0,

0,

0,

or
10

30

50

70

90

17
11

13

15

Time between cars, s During modelling was tested how many data sets
neural network needs to recognize and forecast traffic jam.
Fig. 9. Histogram of times between incoming cars from 1st In 12th figure we can see mean square error pivot to the
entrance

60
data sets quantity. And in the 13th figure regression probability p=0.82 we need minimum 8 events and it is
(correlation) coefficient pivot to the data sets quantity we enough 400 data sets to train this network. As it is very
can see. important to recognize traffic jam before it is formed so
modeling showed that we can forecast it before
approximately 5-10 seconds before it is formed.
0,7
0,6
M e a n S q u are E rro r

0,830
0,5 0,820

Regression (correlation)
Training MSE
0,4 0,810
Validation MSE 0,800
0,3
Testing MSE 0,790
0,2 0,780
0,1 0,770
0,760
0
0,750
63 100 150 200 250 300 350 400 450 500
0,740
Training Data Set 0,730
2 4 6 8 10
Fig. 12. Neural network mean square error pivot to the data sets Number of events before traffic jam
quantity
Fig. 15. Neural network regression (correlation) coefficient pivot
to the number of events
1,2
R eg re ss io n (c o rrelatio n )

1
Conclusions
0,8 Training R
0,6 Validation R 1. Hierarchy structure of parking control systems
Testing R was given.
0,4
2. Parking control system was developed and
0,2 implemented to the real parking –lot.
0 3. Investigation of parking control system was done
63 100 150 200 250 300 350 400 450 500
and experimental data was collected.
4. Experiment showed that Parking guidance and
Training Data Set
Traffic control systems decreased parking time up to 65%
5. Maximum reached traffic jam forecast with neural
Fig. 13. Neural network regression (correlation) coefficient pivot network probability p=0.82
to the data sets quantity
6. Modelling with neural network showed that for
traffic jam forecast (with probability p=0.82) we need at
Mean square errors and regression (correlation)
least 8 traffic events before it starts
coefficient pivots to the number of events taken to
7. Modelling with neural network showed that we
recognize and forecast traffic jams. These pivots are shown
can forecast traffic jam in approximately. 5-10sec. before it
in the Fig. 14–15.
starts
8. Modelling with neural network showed that it is
0,450 enough 400 data sets to train neural network for traffic jam
0,400 forecasting with probability p=0.82
0,350
M ean S quare E rror

0,300 References
0,250
1. NCHRP Sythesis 307 Systems Engineering processes for
0,200
developing traffic signal systems. – Washington: D.C., 2003. – 88
0,150
p.
0,100 2. Andrzej Adamski. Polyoptimal intelligent and integrated
0,050 traffic control method // Telematics and road traffic safety
0,000 IV scientific conference issues CD, Katowice, 2004. – P. 84–92.
2 4 6 8 10 3. Alec Lik Chuen Woo. A holistic approach to multihop
Number of events before traffic jam routing in sensor networks. – University of California, Berkeley ,
phd. dissertation, 2004. – 224 P.
4. Kim D. K. A new mobile environment: Mobile Ad Hoc
Fig. 14. Neural network mean square error pivot to the number of
Networks (MANET) // IEEE Intelligent Transportation Systems
events
Conference, 2004. Washington DC, USA.
5. Shun Miura, Yi Zhan, Tadahiro Kuroda. Evaluation of
Analyzing the results we see that with neural network parking search using sensor network // Department of electronics
we have maximum reached successful traffic jam forecast and electrical engineering, Keio University Press, Yokohama,
with probability p=0.82. Also to reach forecasting Japan, 2005.

61
6. Basu P. Little T. D. C. Networked Parking Spaces:
Architecture and applications // MCL Technical Report No. 07- Submitted for publication 2007 06 13
01-2002, Department of electrical and computer engineering,
Boston University Press, Boston, USA, 2002.

A. Marma, M. Žilys, D. Eidukas, A. Valinevičius. Electronic Parking Control Systems // Electronics and Electrical Engineering.
– Kaunas: Technologija, 2007. – No. 8(80). – P. 57–62.
Problems in the big parking lots near big people concentration places were researched. To control processes in parking lots
electronic control systems are needed. These systems was analysed, systems architecture and their hierarchy structured. To ensure best
performance in the multi floor parking lot with 2500 parking places electronic systems were implemented: parking guidance, traffic
jams detection and control, intelligent administration and handling and emergency situation control. All these systems were developed,
installed, tuned, tested and experimental data collected for modelling. To model these systems over 500000 measurements were done
during experiment. To find a phenomenon appearing before traffic jam is formed in the parking-lot, recognize and forecast traffic jams
in the parking-lot model with neural network using Levenberg-Marquardt optimization was done. Experimental and modelling results
showed advantages of the Electronic Parking Control Systems. Ill. 15, bibl. 6 (in English; summaries in English, Russian and
Lithuanian).

А. Марма, М. Жилис, Д. Эидукас, А. Валиневичюс. Электронные системы контроля автостоянок // Электроника и

электротехника. – Каунас: Технология, 2007. – № 8(80). С. 57–62.
Исследованы проблемы в больших автостоянках, находящихся в близости к местам массового собрания людей. Для
управления процессов в больших автостоянках нужны электронные системы контроля. Исследованы электронные системы
контроля автостоянки, созданы структуры систем, разделены слои иерархии. Чтобы гарантировать хорошую роботу 2500-
местной автостоянки были инсталлированы системы руководства по поискам свободного места, детектирования и контроля
транспортных пробок, администрирования и система контроля экстремальных ситуаций. Все эти системы были разработаны,
инсталлированы, настроены и попробованы, собраны результаты экспериментов. В эксперименте для моделирования системы
контроли автостоянки, было собрано больше чем 500000 данных. Методом Левендбера-Маргуардта была создана модель с
нейронными сетями для установления процессов, предшествующих транспортным пробкам, обнаружения транспортных
пробок, их прогнозирования и контроля. Данные экспериментов и моделирования показали хорошие свойства электронных
систем контроля автостоянок. Ил. 15, библ. 6 (на английском языке; рефераты на английском, русском и литовском яз.).

A. Marma, M. Žilys, D. Eidukas, A. Valinevičius. Elektroninės automobilių aikštelių valdymo sistemos // Elektronika ir
elektrotechnika. – Kaunas: Technologija, 2007. – Nr. 8(80). – P. 57–62.
Išnagrinėtos problemos, kylančios automobilių stovėjimo aikštelėse, esančiose prie masinio susibūrimo vietų. Didelėse aikštelėse
vykstantiems procesams valdyti reikalingos elektroninės sistemos. Šios sistemos buvo išanalizuotos, sudarytos jų struktūros, suskirstyti
hierarchiniai lygiai. Sklandžiam 2500 vietų aikštelės veikimui užtikrinti buvo sukurtos, įdiegtos, suderintos ir išbandytos automobilių
statymo palydos, transporto spūsčių aptikimo ir valdymo, administravimo ir ekstremalių situacijų valdymo sistemos. Kontrolės sistemai
sumodeliuoti surinkta daugiau nei 500000 eksperimentinių matavimų duomenų. Reiškiniams, vykstantiems prieš susidarant transporto
spūsčiai aikštelės prieigose nustatyti ir transporto spūsčiai aptikti ir prognozuoti buvo atliktas modeliavimas neuroniniu tinklu, naudojant
Levenbergo ir Marquardto metodą. Eksperimentų ir modeliavimo duomenys parodė elektroninių automobilių statymo aikštelių valdymo
sistemų pranašumus. Il. 15, bibl. 6 (anglų kalba; santraukos anglų, rusų ir lietuvių k.).

62