a. INDONESIA :
• EVALUASI PANJANG ANTRIAN KENDARAAN PADA PELAYANAN PINTU KELUAR
PARKIR DENGAN ATAU TANPA PERUBAHAN PINTU KELUAR PARKIR DI SOLO GRAND
MALL
• SISTEM PARKIR MENGGUNAKAN KARTU RFID
• EVALUASI KEBUTUHAN DAN KINERJA PELAYANAN PARKIR PADA RUMAH SAKIT
URIP SUMOHARJO BANDAR LAMPUNG
• EVALUASI KEBUTUHAN DAN KINERJA PELAYANAN PARKIR DAN SISTEM ANTRIAN
PADA PUSAT PERBELANJAAN DI BANDAR LAMPUNG
b. INTERNASIONAL :
• A MULTI-CLASS TIME-DEPENDENT MODEL FOR THE ANALYSIS OF WAITING
PHENOMENA AT A MOTORWAY TOLLGATE
• TOWARDS IMPROVING SUSTAINABILITY OF RAIL TRANSPORT BY REDUCING
TRAFFIC DELAYS AT LEVEL CROSSINGS: A CASE STUDY FOR THE STATE OF FLORIDA
• PERFORMANCE EVALUATION OF TOLL PLAZAS USING QUEUING THEORY
• REGULARIZED TAYLOR ECHO STATE NETWORKS FOR PREDICTIVE CONTROL OF
PARTIALLY OBSERVED SYSTEMS
• A REVIEW ON PERFORMANCE OF TOLL PLAZA BY USING QUEUING THEORY
KETERANGAN WARNA UNTUK ABSTRAK
Abstract
Queuing phenomena occured in public service facilities, such as queues of vehicles at the exit gate of Mall parking lot is affected by the
level of service related with the service of the operator of exit gate, and the vehicle arrival rate. Solo Grand Mall is a Mall that frequently
visited by the people of Solo. This research aims to determine the exit gate service improvement altenative to minimize queue as parking
management reccomendation.
The analysis in this study includes the influence of the queue of veicles and volume of vehicles passing the Solo Grand Mall exit gate
service in terms of vehicle arrival rate and service levels. Queue theorem with First In First Out queue discipline is used to resolve this
case. Statistic analysis calculation in case of arrival rate processed based on poisson distribution with Chi Square testing, and statistic
analysis calculation in case of service level processed based on exponential distribution with Kolmogorov-Smirnov testing.
Calculation performed in this study, resulted a queue length of 2 cars in the car exit gate and 8 motor cycles in the motorcycle exit gate.
There are two counters door services at the exit gate of each type of vehicles, so that the time required for a car parking service is 15
seconds, and requires 12 seconds for a motorcylce parking service. The addition of exit gate service is only necessary for motorcycle at rush
hour by opening alternative gate A1 and A4. The number of exit access in Solo Grand Mall suffied the service of vehicle exit from
offroad parking location, therefore, there is no need changes and addition of exit accesses
Abstrak
Fenomena antrian yang terjadi di fasilitas pelayanan umum, seperti antrian kendaraan di pintu keluar parkir Mal
dipengaruhi oleh tingkat pelayanan yang berkaitan dengan pelayanan operator pintu keluar dan tingkat kedatangan
kendaraan. Solo Grand Mal merupakan salah satu Mal yang sering dikunjungi oleh warga Solo dan sekitarnya.
Penelitian ini dilakukan dengan tujuan mengetahui alternatif perbaikan layanan pintu keluar untuk meminimalisir
antrian sebagai rekomendasi pengelolaan parkir.
Analisa pada penelitian ini meliputi pengaruh antrian kendaraan dan volume kendaraan yang melintasi pelayanan
pintu keluar Solo Grand Mal ditinjau dari segi tingkat kedatangan kendaraan dan tingkat pelayanan. Penyelesaian
kasus ini menggunakan teori antrian dengan disiplin antrian First In First Out (FIFO). Perthitungan analisis statistik
pada tingkat kedatangan berdasarkan distribusi poisson dengan uji Chi Square dan analisis statistik tingkat pelayanan
berdasarkan distribusi eksponensial dengan uji Kolmogorov-Smirnov.
Perhitungan yang dilakukan pada penelitian mendapatkan hasil panjang antrian sejumlah 2 mobil di pintu keluar
mobil dan 8 sepeda motor di pintu keluar sepeda motor. Berdasarkan ketersediaan dua loket pintu keluar parkir di
masing-masing jenis kendaraan membutuhkan waktu pelayanan selama 15 detik untuk melayani mobil dan 12 detik
untuk sepeda motor. Penambahan layanan pintu keluar hanya diperlukan pada sepeda motor pada jam-jam sibuk
dengan membuka pintu cadangan A1 dan A4. Jumlah akses keluar yang berada di Solo Grand Mal sudah mencukupi
pelayanan keluar kendaraan dari lokasi parkir offroad, sehingga tidak diperlukan perubahan dan penambahan akses
keluar.
Kata kunci: antrian, tingkat kedatangan, tingkat pelayanan, parkir.
PENDAHULUAN
Lokasi Solo Grand Mall (SGM) yang terletak di pusat kota Solo yang tepatnya di Jalan Slamet Riyadi menjadikan
SGM mudah diakses oleh semua orang, akibatnya memerlukan kawasan dan manajemen parkir yang baik untuk
menampung kapasitas kendaraan yang masuk dan keluar dari kawasan SGM. Hal ini perlu direncanakan dengan baik
agar para pengunjung nyaman dengan ketersediaan fasilitas ini, sehingga perlu adanya pintu keluar dengan pelayanan
yang baik agar para pengunjung bisa mendapat pelayanan tanpa menunggu lama dalam antrian panjang, salah
satunya adalah menambah loket pintu keluar dan jika masih terjadi antrian panjang maka perlu penambahan
alternatif akses keluar parkir mall. Cara mengurangi masalah yang terjadi pada suatu antrian juga dapat
dilakukan dengan menerapkan teori antrian pada sistem pelayanan pintu keluar parkir. Pertimbangan yang
diambil dalam memilih pintu keluar adalah karena dalam zona inilah yang lebih memiliki tingkat pelayanan yang
tinggi, sehingga memungkinkan terjadi antrian. Berdasarkan penelitian yang dilakukan oleh Anton Maulana (2013),
areal parkir untuk sepeda motor di Solo Grand Mall masih memenuhi standar kebutuhan parkir untuk pusat
perdagangan. Untuk itu dilakukan penelitian untuk melihat kondisi yang sebenarnya dari sistem tersebut
TINJAUAN PUSTAKA
Teori antrian (queueing) sangat perlu dipelajari dalam usaha mengenal perilaku pergerakan arus lalu lintas baik
manusia maupun kendaraan (Morlock, 1978 dan Hobbs, 1979). Hal ini disebabkan sangat banyak kejadian yang
terjadi di sektor transportasi dan permasalahan lalu lintas yang terjadi sehari-hari pada sistem jaringan jalan dapat
dijelaskan dan dipecahkan dengan bantuan analisis teori antrian.
Tamin, 2003. Antrian pada dasarnya terjadi karena proses pergerakan arus lalu lintas (manusia dan/atau kendaraan)
terganggu oleh adanya suatu kegiatan pelayanan yang harus dilalui, seperti misalnya: antrian kendaraan yang
terbentuk di pintu gerbang tol terjadi karena pergerakan arus kendaran tersebut terpaksa harus terganggu oleh
adanya kegiatan pengambilan dan/atau pengembalian (pembayaran) karcis tol. Hal ini hampir sama dengan kasus
yang terjadi di pintu keluar Mal yang juga terganggu dengan adanya kegiatan pembayaran biaya parkir. Kegiatan
tersebut akan menyebabkan gangguan pada proses pergerakan arus kendaraan sehingga mengakibatkan terjadinya
antrian kendaraan di mana pada suatu kondisi, antrian kendaraan tersebut akan dapat mengakibatkan permasalahan
bagi pengguna (dalam bentuk waktu antrian) maupun pelanggan (dalam bentuk panjang antrian).
Dedy Eka K, 2004. Berdasarkan penelitiannya yang dilakukan di Bandung Supermall dalam Evaluasi Panjang
Antrian Kendaraan Pada Layanan Pintu Keluar, menyimpulkan panjang antrian optimal untuk pintu keluar mall
adalah 8 kendaraan untuk sepeda motor dan 2 kendaraan untuk mobil. Sedangkan waktu pelayanan berdasarkan
kenyamanan menunggu untuk sepeda motor adalah 12 detik dan 15 detik untuk mobil.
Pada penelitian ini akan mengevaluasi panjang antrian pada pintu keluar parkir Solo Grand Mal Surakarta dengan
menggunakan teori antrian. Kelayakan kinerja pelayanan pintu merujuk pada kondisi eksisting dari offroad parkir Solo
Grand Mal. Batas panjang antrian kendaraan maksimum yang ditentukan dari lokasi eksisting parkir offroad Solo
Grand Mall adalah 8 kendaraan untuk sepeda motor dan 2 kendaraan untuk mobil. Diharapkan penelitian ini dapat
mengevaluasi tingkat pelayanan pada pintu keluar parkir Solo Grand Mal sehingga dapat menjadi alternatif solusi
penanganan parkir.
DASAR TEORI
Komponen Antrian
q = = …………………………………………………………………………………………….. [ 2 ]
d = …………………………………………………………………………………………….. [ 3 ]
w = = d - …………………………………………………………………………………………….. [ 4 ]
dengan :
n = jumlah kendaraan dalam sistem (kendaraan persatuan waktu)
ABSTRAKS
Sensor RFID dapat digunakan pada berbagai macam kebutuhan diantaranya sebagai akses dalam sistem
parkir dan membuka suatu pintu. Dengan menggunakan sensor RFID tempat parkir suatu kampus dapat
mempermudah mahasiswa keluar dan masuk area tempat parkir hanya dengan menggunakan kartu tanda
mahasiswa yang telah didaftarkan sebelumnya ke mikrokontroller. RFID card nanti digunakan sebagai kartu
tanda mahasiswa, setiap RFID card memiliki kode id masing-masing setelah itu kartu tanda mahasiswa
didaftarkan satu-satu ke mikrokontroller dengan menggunakan RFID reader, kode id di RFID card akan diganti
nama dan nbi mahasiswa, jumlah maksimal suatu tempat parkir merupakan informasi untuk memberitahu tempat
parkir kampus tersebut masih ada ruang parkir apa sudah penuh. Penelitian ini telah berhasil mengatasi
permasalahan informasi ketersediaan tempat dengan menggunakan kartu tanda mahasiswa ber RFID, dan
mampu mengurangi antrian kendaraan pada pintu masuk dan keluar.
Kata Kunci: kendaraan, rfid, mikrokontroller, atmega 328, lcd, motor servo, seven segmen, infrared
Oleh
ARYA NUGRAHA
Kata Kunci : Akumulasi Parkir, Satuan Ruang Parkir, Tingkat Kedatangan, Waktu
Pelayanan
BAB I. PENDAHULUAN
khususnya di daerah pasar dan pusat perbelanjaan, dan terjadinya kemacetan atau
antrian menjadi hal yang tidak terelakkan karena di sini akan terjadi bangkitan dan
Kejadian ini biasa terjadi jika kebutuhan pada suatu pelayanan melebihi
pengguna fasilitas yang tiba tidak bisa segera mendapatkan layanan itu. Hal ini
dapat dilihat pada saat terjadi baris tunggu dari pelanggan atau pengguna jasa
yang menunggu layanan, karena pada saat itu bagian pelayanan sedang melayani
pelanggan yang.
6
antrian pada tempat-tempat tertentu, baik skala kecil maupun skala besar yang
Tjutju (2004: 349) teori antrian adalah teori yang menyangkut studi matematis
dari antrian - antrian atau baris - baris penungguan. Pengguna fasilitas atau
pelanggan datang dengan waktu yang acak, tidak teratur dan tidak dapat segera
memakan sebagian badan jalan. Pada banyak hal tambahan fasilitas pelayanan
berikut :
pelayanan.
penelitian.
3. Pengamatan dilakukan pada pintu masuk parkir yang menuju areal parkir
Bagi perusahaan:
lalulintas.
Bagi Peneliti:
khususnya.
adalah hal yang penting dalam pusat kegiatan dapat menimbulkan masalah
lintas jika ketersediaan kapasitas area parkir di tempat tersebut tidak bisa
Rumah sakit Urip Sumoharjo merupakan salah satu rumah sakit swasta yang
terdapat di kota Bandar Lampung. Rumah sakit ini terletak di Jalan Urip
September 2001 dengan bangunan yang berdiri pada lahan seluas 5.000 m2.
lahan menjadi ± 22.911 m2, ± 6.666 m2 luas areal parkir, dengan 425 tempat
tidur. Selain itu saat ini Rumah Sakit Urip Sumoharjo terdapat beberapa
2
kinerja parkir pada rumah sakit Urip Sumoharjo terutama pada waktu jam
besuk pasien dan jadwal-jadwal praktek dokter spesialis pada poli kesehatan
Urip Sumoharjo.
Urip Sumoharjo.
3. Hasil dari penelitian ini juga dapat dijadikan sebagai bahan referensi
ScienceDirect
highlights
Article history: The planning, design and operational management of motorway toll booths are of great in-
Received 16 September 2019 terest in traffic engineering, as these facilities directly influence the quality of the service
Received in revised form offered to users. This paper focuses on a time-dependent queue model based on the co-
20 March 2020 ordinates transformation criterion for operations assessment at a motorway tollgate. This
Accepted 10 April 2020 model allows to face the whole spectrum of situations that may characterize a toll booth,
Available online 18 September 2020 some of which often fall outside the boundaries of the probabilistic theory for stationary
queues.
Keywords: The paper proposes an M=G=1 multi-class queue model for the evaluation of evolutionary
Traffic engineering profiles of waiting times and queue lengths by closed-form equations. The results obtained
Time-dependent queue model for three numerical test cases show a good approximation level, compared with the mean
Shared queue values of queue parameters obtained reiterating a discrete-state simulation model.
Coordinate transformation method The proposed time-dependent equations will be useful in technical cases, allowing to
Tollgate operate quickly and compactly even when probabilistic queue theory is not applicable or
produce unrealistic results, and the burden of complexity of the simulation approach is not
conveniently absorbable. The discussion highlights a significant flexibility of the model
proposed in addressing situations with conventional vehicles, i.e., with total human con-
trol and proposes some considerations for application in future scenarios with the pres-
ence of connected vehicles (CVs).
© 2020 Periodical Offices of Chang’an University. Publishing services by Elsevier B.V. on
behalf of KeAi Communications Co. Ltd. This is an open access article under the CC BY-NC-
ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).
The paper is structured as follows. Section 2 reviews and with the aim of studying the costs minimization, Boronico
discusses the main researches appeared in the international and Siegel (1998) applied an approach derived from the
literature since the 1960s of the last century, going through probabilistic theory of queues. The arrivals are determined
the different approaches related to the performance analysis under the assumption of uniformly distributed demand on
of toll stations and barriers. Section 3 shows an example of a gates of the same type, with different service times
motorway tollgate characterization, examining the nature of depending on whether the collection system is automatic or
the arrival and service processes, and presents the closed- manual. These authors used exponentially distributed
form equations for a multi-class M=G=1 stationary queue service times, while noting that the variance in real-life data
model. Section 4 deals with the time-dependent modeling is lower than that of the chosen distribution. In light of
for the non-stationary queue. Starting from an introduction these considerations, and focusing the analysis on a New
to the coordinate transformation method, the time- Jersey Parkway toll station, an M=M=1 queue model was
dependent solutions for the multi-class M=G=1 queue model developed in order to obtain the higher limits of the
are deducted for a tollgate. The closed-form equations for investigated MOEs.
the average waiting time and the average number of Also Cherng et al. (2005) used the probabilistic theory for
vehicles in the system during a certain observation interval steady-state queues to evaluate the waiting times at the
are provided, which are also applicable sequentially allowing gates of a motorway toll station and at the convergence
to evaluate any time profile for demand and service. Section points downstream. These authors formulated some basic
5 presents the SDES model used to compare time-dependent assumptions to define the model: a constant flow uniformly
model results. A stochastic simulation model with Poisson distributed on the gates; exponential arrival and service
arrivals and multi-class Gamma-distributed service times, times; total waiting time obtained as the sum of time to
corresponding to the already mentioned multi-class M= G= 1 receive and complete the service plus the lost time at the
queue model, is presented. Section 6 discusses some confluence to reach the exit segment. An M=M=1 model for
numerical results, comparing time-dependent model waiting at the gate and an M=G=1 model for the downstream
outputs with a sufficiently high number of iterations of the confluence were used, in order to determine the optimal
SDES model by applying the Monte Carlo method. Section 7 number of gates for minimizing the total time spent. Also
offers an insight for a future scenario, which considers an Wang (2017) presented an M=M=1 model for the estimation
increase in automatic payment technologies and a growing of the waiting time spent at a motorway tollgate,
popularity for connected vehicles (CVs). Section 8 contains considering different types of payment services. The queue
the concluding remarks. model was used within an optimization method for the
identification of the number and the configuration of the
gates, according to a non-dominated sorting genetic
2. Literature background algorithm II (NSGA-II).
As an alternative to the M=M=1 model, Kim (2009) proposed
Edie (1954) was the first to deal with the analysis of a toll booth an M=G=1 model with service times generically distributed
using the analytical theory of the queues, analyzing its with known mean and variance, for estimating the average
performance through the probabilistic M=M=s and M= D= s waiting time required by an integer and non-linear
queue models with multiple serving channels, assuming programming algorithm for a toll station optimization. The
Poisson arrivals (i.e., M=/=/) and exponentially (i.e., /= M= model considers the hypothesis that the arrivals by payment
/) or deterministically (i.e., /=D=/) distributed service type are uniformly distributed on the gates. The assumption
times. The model was used to estimate the time taken by a regards a real-life evidence: an incoming vehicle tries to
vehicle to carry out the operations and to clear the gate, in reach the shorter queue among those that offer the
the hypothesis of homogeneity in vehicle types and gates requested service and, once in the waiting lane, the vehicle
characteristics. Compared to the discussion of Edie (1954), does not abandon it for another.
which appears not properly in accordance with the actual Mehri and Djemel (2011) provided further specifications for
operations of a station with a sufficiently sized toll plaza, probabilistic models regarding service times and gates
Haight (1958) proposed a model applicable to a double-entry number. The authors identify a Gamma (G) distribution for
toll station. With respect to this model, the incoming vehicle service times starting from the analysis of experimental data
chooses time to time the shortest between the two parallel collected for a Tunisian motorway and using a stationary
queues. The model considers Poisson arrivals and M=G=s for performance analysis. The analysis considers an
exponential service times with possibly different average approximation of the Pollaczek-Kintchine (PK) formula for
values for the two gates, in the double hypothesis that each the M=G=s model. Using a multi-agent simulation model,
vehicle must necessarily keep the gate chosen on the arrival they confirm the lower adequacy of an M=M=s model.
or that it can change queue while waiting. Karsaman et al. (2014) used an M=M=s model for the
To address queue lengths and waiting times for a toll sta- stationary analysis of the differences in capacity between
tion with a variable number of gates and different types of toll cash and electronic payment systems on Indonesian
collection systems (e.g., manual, automatic, electronic, even- motorways. Also Liu et al. (2017) proposed an M=M=s model
tually mixed and with distinction for vehicular class), Zarrillo for cost analysis and optimization of a toll station, including
et al. (1997) used a deterministic model based on the a cellular automata model to evaluate some traffic indicators.
cumulative curves of arrivals and service processes. In Regarding the aforementioned single or multiple channel
presence of multiple gates with different types of service, models, Chakroborty et al. (2016) proposed a model called
Cleaner Logistics and Supply Chain 1 (2021) 100001
A R T I C L E I N F O A B S T R A C T
Keywords: Rail transport makes a significant contribution to the economy of many geographical locations around the
Railroad transport globe. Along with numerous merits, rail transport has some demerits as well, such as traffic delays and safety
Level crossings issues at level crossings. Traffic delays at level crossings have various negative impacts, including economic
Traffic delays losses, increasing number of violations because of delays, queue spillbacks at the adjacent intersections, and
Crossing accidents
environmental impacts. This study reviews a number of level crossing delay estimation approaches found in
Safety countermeasures
Crossing upgrades
the literature. Based on the state‐of‐the‐art review and practical considerations, a new approach for estimation
of the overall traffic delay per day at level crossings with the existing warning devices is proposed to improve
the sustainability of rail transport. Moreover, in order to assist with the selection of countermeasures for level
crossing upgrades, the proposed approach can be also used to estimate the overall traffic delay per day at level
crossings after implementation of countermeasures. A case study is conducted for the public and private level
crossings in the State of Florida, considering the countermeasures commonly used in the United States. It was
revealed that traffic delays were higher at the public level crossings than at the private level crossings, both
before and after implementation of the candidate countermeasures. The case study also indicated that low‐
cost countermeasures could be more advantageous in terms of safety effectiveness and traffic delays as com-
pared to many high‐cost conventional countermeasures. Furthermore, certain attributes of the level crossings
in Florida with the highest traffic delays were investigated.
1. Introduction use of highway vehicles (Pucher et al., 2007). Thus, a sustainable rail
transport can help to achieve a sustainable highway transport by
The economy of many geographical locations around the globe is reducing congestion on highways. Rail transport may have some
significantly influenced by rail transport. For example, the Florida adverse effects at places where railways intersect highways (i.e., level
Department of Transportation (FDOT) reported that the economic out- crossings). Level crossings may negatively affect system‐wide highway
put of freight rail transport and passenger rail transport in the State of traffic and influence the livelihood of nearby societies (e.g., increased
Florida (United States – U.S.) was measured by $121.0 billion and noise due to highway and railway traffic, increased emissions,
$589.3 million in 2018, respectively (FDOT, 2018). Rail transport increased risk of accidents). Highway traffic delays are typical for
can comprise a notable portion of urban transport and reduce the many level crossings around the globe (Wullems, 2011; Beanland
⇑ Corresponding author.
E-mail addresses: jp17j@my.fsu.edu (J. Pasha), mdulebenets@eng.famu.fsu.edu (M.A. Dulebenets), prashant1.singh@famu.edu (P. Singh), moses@eng.famu.fsu.edu (R. Moses),
sobanjo@eng.famu.fsu.edu (J. Sobanjo), eozguven@eng.famu.fsu.edu (E.E. Ozguven).
https://doi.org/10.1016/j.clscn.2021.100001
Received 25 April 2021; Revised 13 June 2021; Accepted 16 June 2021
et al., 2018). The Federal Highway Administration (FHWA) of the U.S. sures with minimal delays and installation costs along with the
expects that the highway vehicles in the vicinity of level crossings will adequate effectiveness to provide a sufficient safety level for highway
experience an additional delay of 123.22 million hours annually by and rail users.
2022 as compared to 2003 (FHWA, 2002). On the other hand, the Since the presence of countermeasures can extend delays and result
trucks in the vicinity of level crossings are likely to experience an addi- in various negative externalities, this study specifically focuses on traf-
tional delay of 6.66 million hours annually by 2022 as compared to fic delay estimations at level crossings. In particular, a new approach is
2003 (FHWA, 2002). The present value of delay for all the vehicles proposed for estimation of traffic delays at level crossings before and
due to highway and rail traffic at level crossings in the U.S. comprises after implementation of countermeasures, taking into account the find-
approximately $7.9 billion for the 20‐year period between 2003 and ings from previous state‐of‐the‐art efforts and practical considerations.
2022 based on the 50% confidence interval (FHWA, 2002). The ultimate goal of the proposed approach is to facilitate a sustain-
Traffic delays can increase even further with the frequency of trains able rail transport by selecting the appropriate countermeasures and
that pass and, hence, obstruct level crossings (Nguyen‐Phuoc et al., reducing traffic delays at level crossings after implementation of these
2017). There are a number of reasons behind traffic delays at level countermeasures. The remainder of this manuscript begins with a
crossings. Features of level crossings vary significantly from highway review of the relevant literature in section 2. Section 3 illustrates the
intersections. For example, in case of gated level crossings (i.e., the proposed approach for estimation of traffic delays at level crossings,
level crossings that have either two‐quad gates or four‐quad gates while section 4 presents a case study for the State of Florida. Finally,
installed to prevent the vehicles entering the crossing in the event of section 5 discusses the outcomes of this study and outlines some direc-
an approaching train), the frequency, timing, and duration of gate clo- tions for the future research.
sures could vary from one level crossing to another. Such variability
makes travel decisions (e.g., decisions regarding departure time and
2. Review of the relevant literature
route) harder for motorists, which could lead to vehicle impedance
at level crossings (Kawamura and Allahyari, 2020). Moreover, the
A review of the relevant literature on level crossing delays is pre-
down times of gates at level crossings could be significantly longer
sented in this section of the manuscript. Specifically, this section con-
than the red times of highway intersections, which could result in
centrates on the studies discussing the effects of traffic delays at level
longer queues at level crossings. Such long queues could also require
crossings as well as the existing approaches for estimation of traffic
a considerable amount of time for dissipation and, therefore, increase
delays at level crossings. Moreover, a literature summary and contribu-
delays for the vehicles along the queues.
tions of this study are outlined in this section as well.
Estimation of traffic delays at level crossings could be associated
with several parameters, such as the length of a passing train, speed
of a passing train, time‐wise distribution of railway traffic, time‐wise 2.1. Effects of traffic delays at level crossings
distribution of highway traffic, level crossing obstruction time, and
others (Okitsu et al., 2010; Port of Los Angeles, 2011; CUTR, 2014; Traffic delays at level crossings have various negative impacts,
STB, 2020). If the other factors remain unchanged, longer trains will including economic losses, increasing number of violations because
lead to longer delays at a given level crossing. Several characteristics of delays, queue spillbacks at the adjacent intersections, and environ-
related to the speed of a passing train could be associated with delay mental impacts. A number of research efforts have been carried out to
estimations, such as the typical maximum train speed, maximum time- analyze the economic impacts of delays at level crossings (Nguyen‐
table train speed, typical minimum train speed, and so on. Railway Phuoc et al., 2017; Gruyter and Currie, 2016). Dodgson (1984) stated
traffic and highway traffic could be high during specific times of the that delays to highway vehicles could be caused by various reasons,
day. If both railway traffic and highway traffic are significant for the such as short‐term speed restrictions imposed at level crossings after
same time period, notable traffic delays at level crossings are likely incidents or long‐term speed restrictions. Moreover, delays to highway
to occur. On the contrary, highway traffic can be relatively low during vehicles might impose added time costs to railway passengers or incre-
off‐peak hours, which will reduce the associated delay. Many level ments in the railway operating costs due to acceleration/deceleration.
crossings are equipped with different types of protection (e.g., flashing Ryan (1996) computed delay costs accrued by highway vehicles at
lights, wigwags, two‐quad gates, and four‐quad gates), and increasing level crossings. According to the study, the total delay cost at each
warning time of these devices before an approaching train would fur- level crossing could be disintegrated into several cost components,
ther lead to additional traffic delays at these crossings. A precise delay including the user cost for stopping conditions, user cost for speed‐
estimation approach should incorporate all of these considerations. change cycles, and travel time cost. It was revealed that the mean
Along with traffic delays, level crossings may pose major safety annual delay cost at the level crossings in Maryland (U.S.) was
concerns, such as possibilities for train‐highway vehicle collisions, $4,180; however, it could potentially reach a maximum of $407,441
vehicle‐warning device collisions, train‐pedestrian collisions, property (Ryan, 1996).
damage, etc. (Abioye et al., 2020; Keramati et al., 2020; Lam and Tai, Hakkert and Gitelman (1997) indicated that traffic delays at level
2020; Zhou et al., 2020). The Federal Railroad Administration (FRA) crossings have a number of demerits, such as time loss due to velocity
of the U.S. reported that a total of 1,649 level crossing accidents, cycles, added fuel consumption, and additional vehicular expenses.
which led to 268 fatalities as well as 755 injuries, were recorded in These demerits could be reflected by an economic loss. The study pre-
the State of Florida between January 2000 and December 2018 sented a linear regression formula to estimate economic losses due to
(Pasha et al., 2020). These issues could further intensify with incre- delays at level crossings. According to the developed formula, the eco-
ments in highway traffic and railway traffic. Safety issues at level nomic loss at a level crossing due to traffic delays may be estimated
crossings could be offset with the implementation of countermeasures from the traffic volume, number of level crossing obstruction occur-
(e.g., two‐quad gates, four‐quad gates, barrier curbs, median separa- rences per day, duration of level crossing obstructions per day, average
tors, and flashing lights). However, such countermeasures may extend free speed of vehicles, and average slowdown of vehicles that
traffic delays at level crossings (NCHRP, 1987; U.S. DOT, 2019). Grade approach the level crossing. Gitelman et al. (2006) discussed that an
separation could be a solution to avoid delays and accidents at level increase in the traffic volume and train volume could lead to signifi-
crossings. However, such a solution is generally viewed as radical, as cant traffic delays. The study employed a model, according to which,
it involves substantial capital investments (NCHRP, 2019; Kavoosi the annual cost of roadway traffic delays at a level crossing depended
et al., 2020; Chandra et al., 2021; Pu et al., 2021). Therefore, a sustain- on the traffic volume, number of stopped roadway vehicles at the level
able rail transport should involve the implementation of countermea- crossing per day, average vehicular stoppage cost, and average
2
RESEARCH PAPERS
ABSTRACT
Ever increasing vehicle traffic has over loaded on the performance of toll collecting plazas. Manual toll collection and
Electronic toll collection methods are the most commonly adopted methods of toll collection in India along various
sections of National Highways and Expressways, which were constructed and maintained on Public Private Partnership
(PPP) mode. Manually operated toll collection at such plazas can be very difficult, especially during peak travel times,
particularly during festivals. It is difficult to manage the queue and assign the required number of toll collecting gates
operating in order to allow vehicles quickly and safely. Due to this reason, electronic toll collection servers are the best,
but they are being less used by many drivers and still use manual toll collection. This results the drivers compelled to stop in
front of the server for payment and create long queues and demand more number of toll collecting servers. To
overcome this problem, this research has focused on to determine the performance evaluation of a few electronic toll
collection and manual toll collection gates on National Highway (NH) 65 and Outer Ring Road. The arrival rate, service
rate, delays, and the queue length data were collected. The arrival rates were input into a passion distribution based
queuing model to determine the performance parameters of toll plazas and also the required number of servers at the
study plazas.
Keywords: Arrival Rate, Service Rate, Toll Plaza, Servers, Delay, Queue Length, Queuing Theory.
INTRODUCTION makes the drivers compel to stop in front of the server and
Throughout India, National Highways (NH) and expressways pay their toll. This stoppage of vehicles can create long
are the major road connectivity with the entire capital cities queues if the service time is longer and the required
and commercially important places in the country. These number of toll collecting servers is inadequate.
roads were constructed and maintained based on Public While there are many efforts necessary to create the toll
Private Partnership (PPP) models by different concessioners collection system completely free of delays and also need
and the invested amount will be collected from toll plazas, to be able to run smoothly despite ever increasing vehicles
which are located along the highways. Huge and ever on our highways each year. An analysis of toll collection
increasing vehicle traffic over the past years has characteristics is necessary for evaluating any toll plaza to
overloaded on the performance of toll collecting plazas. be able to run in smooth and efficient way. The
Manual toll collection and Electronic toll collection performance of any toll plaza can be evaluated by several
methods are the most commonly adopted methods of toll factors, particularly safety, queue length, arrival rate,
collection along the various sections of national highways departure rate, and delays. The objective of this study is to
and expressways in India. However, electronic toll evaluate current methods of toll collection system in terms
collection servers are less used by many drivers and they still of arrival rate, departure rate, and queue length and
use manual toll collection as their default practice. This waiting delays at two selected Toll plazas. Determination of
ABSTRACT The existing neural networks suffer from partial observation while modeling and controlling
dynamic systems. In this paper, a new linearized recurrent neural network, the Taylor expanded echo state
network (TESN), is proposed for predictive control of partially observed dynamic systems. Two schemes
of regularization, ridge regression and sparse regression, are imposed on TESNs to tackle the issue of
ill-conditioned estimation. Furthermore, two estimators, lasso and elastic net, are investigated for sparse
regression. Regularized learning is found to improve the estimation consistency of readout coefficients and,
at the same time, suppress the accumulation of linearization residues in a prediction horizon. A series of
experiments was carried out, and the results verified that regularized learning is contributive to TESNs in
predictive control of partially observed dynamic systems.
INDEX TERMS Neural networks, echo state networks, sparse regularization, predictive control.
2169-3536 2016 IEEE. Translations and content mining are permitted for academic research only.
3300 Personal use is also permitted, but republication/redistribution requires IEEE permission. VOLUME 4, 2016
See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
K. Xiang et al.: Regularized Taylor Echo State Networks for Predictive Control
and, thereby damage control robustness. A few studies have specific application, the sequence of known states is imported
focused on linearization residues. For example, Yan and into the ESN to transiently warm up from x(0)=0 to the
Wang developed a series of models with an extreme learning initial state. The sequence is then discarded, and more states
machine [17]. Gil et al. [18] modeled linearization residues are collected for readout estimation. Then, teacher forcing is
by using a dual unscented Kalman filter for online updating introduced to enhance the ESN training when there is output
of neural network weights. It is noteworthy that, typically, feedback [23], in which the relationship between the reser-
the aforementioned solutions are all passive, and the mod- voir and the readout is disengaged. The procedure eventually
eling effectiveness is still dependent on observations. becomes similar to that of a feed-forward system. In this
In the present study, an active scheme is proposed to study, by using Wfb , we directly feed the desired outputs,
significantly alleviate the accumulation of linearization instead of the real outputs, into the connections for supervised
residues, which is helpful toward improving control robust- training. Given the state matrix x and the desired output ytchr ,
ness. For a new ESN, it is first linearized by the it is possible to estimate Wout by using an ordinary least
Taylor expansion, and thus converted into a Taylor-expanded square (OLS) estimation:
ESN (TESN). Regularized learning, by which the readout −1
coefficients are constrained, is proposed to suppress residue Ŵout = ytchr xT xxT (2)
accumulation in the prediction horizon. The regularized An ESN can be adjusted and optimized to model any
TESN is effective in modeling typical complex systems. Here temporal dataset. In practice, linearization has been found
regularized learning brings twofold benefits: on one hand, to be indispensable for robust predictive control. Given an
it effectively copes with partially observed dynamic systems; operating point φ (usually set φ = y0 ), the internal state
on the other hand, no additional model is necessary. x of an ESN needs to approach a specified state χ0 = sout
The rest of this paper is organized as follows. Section II (Wout )−1 φ. Eq. (1) can be linearized by applying the first-
introduces the background works on ESNs and TESNs for order Taylor expansion if the ESN is approximately linear in
predictive control are introduced. Section III elaborates the the area around χ 0 :
proposed schemes of regularization, namely, ridge regres-
x (k + 1) = f χ 0 + ∇f χ 0 χ (k)−χ 0 + e χ (k)−χ 0 ,
sion [19] and sparse regression [20], [21]. A series of exper-
iments are reported in Section IV. Section V addresses the (3)
experimental results and the related findings in the preceding
section. Concluding remarks are drawn in the last section. where
χ (k) = ρWx (k) + sin Win u (k + 1) + sfb sout Wfb y (k) (4)
II. TESN FOR NPC
Here, ∇f (χ 0 ) is a Jacobian matrix and e(·) denotes all other
A plain ESN was first reported in 2004 [16], and the
high-order terms. If further defining
authors further proposed a new morphological paradigm with
A = ρ∇f χ 0 W
leaky-integrator in 2007 [22]. These ESNs have a simi-
lar infrastructure that includes a high-dimensional reservoir B = sin ∇f χ Win
0
matrix and a readout matrix. The reservoir matrix is advan- fb sout ∇f χ Wfb
(5)
C = s 0
tageous of accommodating fruitful nonlinear dynamics, and
D = 1/sout Wout
is thus able to approximate highly nonlinear (such as chaos)
systems. Eq. (1) can be converted into a standard state space:
It is common to model an ESN as: (
x (k + 1) = Ax(k) + Bu(k + 1) + Cy(k) + c
( (6)
x(k +1) = f sin Win u(k +1) + ρWx(k) + sfb sout Wfb y(k) y (k + 1) = Dx(k + 1)
y(k + 1) = 1/sout Wout x(k + 1) where the constant term c = f (χ 0 )−∇f (χ 0 )χ 0 . Note that e(·)
(1) in Eq. (3) has been omitted for a fully linear state space. In this
study, Eq. (6) is termed as a TESN model to differentiate it
where f (·) is an unknown activation function, u(k) denotes from naive linear ESNs with linear activation functions.
the input vector, x(k) is the internal state matrix, and y(k) To eliminate the constant term c, it is appropriate to rewrite
is the output vector. Here, y(k) may be optionally fed back Eq. (6) in an incremental vector format:
to x(k + 1). All weighting matrices W∗ , except Wout , are (
randomly generated and normalized before supervised train- 1x (k + 1) = A1x(k) + B1u(k + 1) + C1y(k)
(7)
ing. All the scalar parameters (ρ, sin , sout , and sfb ) in Eq. (1) 1y (k + 1) = D1x(k + 1)
are subject to manual settings; however, ρ should be smaller
Given the prediction horizon Np , the control horizon Nc ,
than 1. The superscripts are added to discriminate the weight-
the dimension of input vector nu , the dimension of output
ing matrices, where ‘‘in’’, ‘‘out’’ and ‘‘fb’’ denote the input,
vector ny and a quadratic coefficient matrix H, which is
output and feedback loop, respectively.
defined as
Three essential steps are involved in establishing a new
ESN: warming, collection and estimation. First, for a H = (STu ITy Ty Iy Su + Tu ) (8)
@ IJTSRD | Unique Paper ID – IJTSRD28068 | Volume – 3 | Issue – 6 | September - October 2019 Page 140
International Journal of Trend in Scientific Research and Development (IJTSRD) @ www.ijtsrd.com eISSN: 2456-6470
toll collection methods, number of toll booths, and types of MEAN PERFORMANCE PARAMETERS
vehicles have been studied [14]. The assessment of 1. Traffic Intensity
uninterrupted traffic flow is traditionally based on empirical The average number of customers being served is the ratio
methods [15]. The issues of traffic management are the of arrival and service rate. For a stable system the service
process of increased disturbance in the movement of traffic. rate μ should always exceed the arrival rate λ and thus ρ
In India the traffic congestion is rapidly increased due to should always be less than one. Therefore, it is also known as
increasing vehicles on the road which produces many utilization factor of the server. This can be represented as
problems [16]. Due to the escalating traffic issues a large ρ=λ/μ
number of highways have been built in the recent past which
are maintained by tax collection at toll plazas by various 2. Average Number of Customer in the System
operating agencies. Due to smooth and hassle-free driving on The average number of customers in the system is equal to
highways, the arrival rate of vehicles at toll plazas increases. the average number of customers in the queue together with
The arrival rate goes beyond control if the traffic on the those being serviced. It is expressed as
highway increases in an uncontrolled manner with the Ls = ρ / (1-ρ)
passage of time [17]. The toll plaza on the Kaduna – Zaria
highway has been modeled as a queue system by adapting 3. Average Number of Customer in Queue
the Makino model [18]. It can be viewed as average queue length that is, the average
number of customers who are waiting in the queue. It is
FEATURES OF THE QUEUING SYSTEM defined as
Generally, any queuing arrangement is composed of units, Lq = Ls – ρ
denoted as customers, requiring some kind of service. When
a customer arrive at a service facility, join a queue if service 4. Average Time Spent in the System
is not instantly available, and ultimately leave after receiving The average time spent in the system is equal to the total
the service. A server refers to mechanism that delivers time that a customer spends in a system i.e. waiting time
service to the customer. Upon arrival, if a “customer” finds plus the service time. It is given by
the server busy, then he may form a queue, join it or leave Ws = Ls / λ
the system without getting any service even after waiting for
some time. This therefore permits a number of different 5. Average Waiting Time in Queue
possible arrangements such as the following, to describe The average waiting time in queue is the average time a
vehicular traffic flow as a queuing system: customer waits in queue forgetting service. It is expressed as
1. The arrival pattern: This is the manner in which arrivals Wq = Lq / λ
occur, indicted by the inter-arrival time between any
two consecutive arrivals. For our stochastic modelling METHODS OF TOLL COLLECTION
framework, the inter-arrival time may vary and may be There are three methods of toll collection available. These
described by a specific probability distribution that best are: (i) Manual, (ii) Automatic, and (iii) Electronic system.
describes the arrival pattern observed.
2. The Service Pattern: This is the manner in which the 1. Manual Toll Collection
service is rendered and is specified by the time taken to Manual toll collection is most broadly used collection
complete a service. Similar to the arrival pattern, technique in India. It requires a toll collector or attendant.
distribution of the service time must be specified under Based on the vehicle classification, cash toll is received by
stochastic modelling considerations. the collector. The collector, who also dispenses change, may
3. Arrival Rate (λ): This is the average number of vehicles accept and sell scrip, tickets, coupons, making an entry of the
arriving per unit time. vehicle in the system and issuing receipt to the patron. Due
4. Service Time (μ): This gives the average number of to manual intervention, the processing time is highest
vehicles served per unit time.
5. Server Utilization (ρ): This gives the average utilization 2. Automatic Toll Collection
of the server. Automatic toll collection is based on the use of Automated
Coin Machine (ACM). These accept both coins and tokens
NEGATIVE IMPACTS OF TRAFFIC CONGESTION issued by the operating agency. Depending on the toll rate,
1. Wasting time of drivers and passengers in blocked the use of automated coin or token collection instead of
traffic affect the economic health of the nations. manual collection reduces transaction and processing time
2. Wasted fuel increases the air pollution and carbon as well as the operating cost
dioxide emission owing to increased idling, acceleration
and braking. 3. Electronic Toll Collection
3. Due to blocked traffic, emergency vehicles may delay in Electronic Toll Collection (ETC) is a system that
reaching to their destination where they are urgently automatically identifies a vehicle equipped with a valid
needed. encoded data tag or transponder as it moves through a toll
4. Spillover effect from congested main routes to lane or checkpoint. The ETC system then posts a debit or
secondary roads and side street as alternative routes are charge to a patron’s account, without the patron having to
attempted which affect colony amenity and real estate stop to pay the toll. ETC increases the lane throughput
prices. because vehicles need not stop to pay the toll.
5. Delays, which may result in late arrival for employment,
meetings and education, resulting in loss of business, REVIEW OF LITERATURE
disciplinary action or other personal losses. YAN Qiu et al. (2013) made an evaluation of the design
requirements when expected traffic flow is in Erlang
@ IJTSRD | Unique Paper ID – IJTSRD28068 | Volume – 3 | Issue – 6 | September - October 2019 Page 141