Article citation information:
Wiedemann, R. Method
of occupancy-based traffic light priority for public transport. Scientific Journal of Silesian University of
Technology. Series Transport. 2022, 115,
227-248. ISSN: 0209-3324. DOI: https://doi.org/10.20858/sjsutst.2022.115.16.
Remigiusz WIEDEMANN[1]
METHOD OF OCCUPANCY-BASED TRAFFIC LIGHT PRIORITY FOR PUBLIC TRANSPORT
Summary. This paper
deals with the problem of urban traffic control, considering public transport
as a priority. According to the authors, the occupancy of a means of transport
is one of the key decision variables in the process of prioritisation within a
traffic signal program. This study aims to construct a method of
occupancy-based traffic light priority for public transport and investigate the
possibility of using this information to increase the efficiency of signal
control for time loss. The mathematical model of the priority level
conditioning procedure proposed in the method was tested using a
microsimulation model of the intersection. The simulation results were collated
and compared with an approach that does not consider vehicle occupancy. Under
given traffic conditions, the use of the proposed method allows for reducing
the average time losses per person in the modelled road network.
Keywords: traffic
light control, occupancy of the vehicle, public transport, simulation approach
1. INTRODUCTION
1.1.
Transport systems in
cities
The development in the field of motorisation and the
increase in the wealth of the society significantly influence the transport
behaviour of city residents. On the one hand, due to the process of urban
sprawl, the inhabitants become more mobile. The daily commute to the workplace
or university involves the necessity to travel often up to several dozen
kilometres a day within the city [3, 10]. On the other hand, owning a vehicle, once considered a privilege, is now the norm for most
people. Moreover, almost every adult member of a household possesses a car
these days.
Consequently, a constant increase in traffic intensity in
the road network is noted, causing a reduction in smooth traffic flows in
cities. A manifestation of this dependence is the globalisation of the phenomenon
of transport congestion, the symptoms of which are usually observed in the road
network, especially in the area of intersections. Currently, the
increase in traffic intensity and the adverse effects of transport congestion
are often reduced by favouring journeys made by public transport (PuT). In
general terms, travellers are encouraged to make their daily journeys by bus,
tram or other public transport instead of private transport (PrT).
Although
traffic light systems are mainly used to increase safety at intersections, when
well designed, they can also improve the traffic efficiency of road users [8].
The systems offer the possibility of implementing public transport
prioritisation methods in the traffic light control algorithm. This aspect
constitutes the background for the considerations put forward in this article [9,
27].
1.2.
State-of-the-art
public transport prioritisation
The issues related to the methods of prioritising public
transport are widely discussed in the literature. Presently,
there are two groups of methods for prioritising these types of vehicles in the
transport network (apart from general legal conditions concerning the right of
way in the area of intersection), that is, facility design-based and traffic
light control-based methods [5, 6, 21] (Figure 1).
Fig. 1. Public transport priority methods
classification
Source: author’s work [1, 2, 6]
The
first group of methods includes, among others, special roads for buses or trams
separated from roads reserved for private transport or exclusive lanes for them
along the streets. Also, the appearance/structure of bus stops may offer
preference to this type of (public) transport. One example is a bus bulb, which
ensures safety and speeds up the transfer of passengers. [5, 21].
Within
the road network in contemporary cities, traffic lights play an increasingly
important role regarding public transport priority. The number of traffic light-based
methods developed over the years confirms this fact. Diakaki and Papamichail et
al. [6] introduced an overview of 85 scientific papers and practical solutions
dealing with issues of public transport priority. The time scope of the review
covers the years 1969-2013. Interestingly, numerous new approaches have been
presented since then.
According
to the actual state of the art, traffic light-based priority for public
transport can be implemented using passive or active strategies [1, 2, 22].
Passive strategies feature no need for a detection system and, consequently, an
adaptive system. The traffic light system planned in this way operates on fixed
time signal programs adjusted to common traffic flows and known public
transport lines schedules. They are most effective when PuT vehicles run at
high frequencies and when their dwell time is relatively short [21, 26]. The
most important disadvantage of this approach is related to the lack of
flexibility. Hence, this strategy is rarely used anymore [6, 19].
The
active strategies have become an alternative to the passive ones and are now
gradually replacing them. Their operation is closely related to the use of
information from the system of detectors (located within the area of
intersection) to adapt the signalling program to the situation at the
intersection [2, 21, 26]. They prioritise specific public transport vehicles to
cross the intersection by triggering special procedures in traffic light
control logic. These procedures often include the following actions [2, 5, 6, 19,
20, 22]:
- generating a special
priority stage after detecting a notification from a privileged vehicle (bus,
tram, trolleybus). When there is no notification, the priority stage is omitted
in the signal program;
- green signal
extension that allows extending the duration time of the active signal stage;
- changing the sequence
of signal stages realised as the shortening of the displayed active phase and
the earlier display of the phase for public transport vehicles, which may
include time compensation for the shortened signal stages.
Moreover,
these actions can be implemented individually or in combination. The decision
of which action and when has to be activated and may be made by the signal
controller using a rule-based method [7, 20, 21] or optimisation [5, 11, 23].
The concept of priority levelling and the methods for assessing its
effectiveness are crucial, as it is frequently argued that unconditional
prioritisation may result in the deterioration of overall traffic conditions [5,
12, 20, 26]. In several studies, these decisions are made based on deviations
from the schedule or order of requests of the arriving PuT vehicle [16]. It
means that frequently only late vehicles are considered for priority at the
signal-controlled intersections. Furthermore, in a situation where two equally
delayed vehicles are waiting at the intersection, the vehicle that sent the
request to the signal controller is served first.
Nowadays,
many research papers indicate that more parameters need to be considered in the
process of PuT vehicle's prioritisation. Hence, the following factors are
commonly used [6, 7, 18, 25]:
- actual location of
PuT vehicles relative to the signalised intersection [2, 7, 12, 21, 23];
- actual speed of
detected PuT vehicles [2, 7, 15];
- intensity of traffic
on the approaches to the signalised intersection [5, 7, 14, 26];
- number of passengers
in the cabin of PuT vehicle [2, 5, 7];
- the deviations from
the PuT vehicle schedule [7, 11, 12, 20, 21, 26];
- values of PuT vehicle
energy consumption [4, 7, 24];
- values of PuT vehicle
emissions [7];
- PuT vehicle waiting
on the stop [21].
The
occupancy-dependent PuT vehicle prioritisation methods are relatively new.
Making the control method dependent on this parameter may determine the level
of the assigned priority. Hence, it may constitute the basis for the conclusion
that sometimes it is more effective for all road users not to give priority to
a bus or a tram (for example, when the route ends and it is empty) or give it
to other bus/tram (which is more occupied). So far, only a few studies have considered this aspect in
detail.
In
2011, Christofa and Skabardonis [2] presented a traffic light control procedure
that minimises the total delay per person in the network while assigning
priority to transit vehicles based on their passenger occupancy. The presented
approach was tested using a simulation at a signalised intersection
located in Athens, Greece. The results showed that the proposed system might
lead to significant reductions in PuT users' delay and the total person delay
at the intersection under given conditions.
Efimenko
et al. 2018 [7] presented
research dealing with the issue of priority of PuT vehicles on a simple
signalised 4-inlet intersection. The used method assumes delays of PuT vehicles
as the main factor of prioritisation. The actual number of passengers in the
vehicle constitutes an auxiliary criterion in the process of ensuring the
priority of PuT vehicles. In this way, using a simulation approach, those
authors formulated decision rules of the investigated traffic light controller
logic. Consequently, they achieved a notable reduction in the delay of
passengers travelling by public transport.
A
slightly different technique assuming information about occupancy but for its
average value, both for private and public transport vehicles, is presented by
De Keyser et al. [5]. They
compared the results obtained through the microsimulation of four different
strategies of traffic light control. Based on the obtained delay data, they
noted that the deterioration of general traffic conditions in the road network
is caused by a higher priority for PuT vehicles. Thus, a trade-off based on the
calculation of the minimal number of passengers in PuT vehicles, necessary to
justify a higher level of priority, is proposed. Thus, the authors selected the
best strategy depending on the actual traffic conditions in total passenger
travel time (both PuT and PrT passengers).
The authors of the publications mentioned above
emphasised the importance of vehicle occupancy in the traffic control process.
In their opinion, strategies based on this factor are in line with the trend of
increasing people's mobility.
The
use of the occupancy factor in the prioritisation process also raises some
criticism. In 2016, Molecki [18] noted that the direct use of the occupancy of
the vehicle as one of the decision-making factors might not reflect the actual
needs of passengers. Thus, there may be a situation of giving higher
priority to a full vehicle ending the route rather than to an empty vehicle
starting the route for which plenty of people wait at the next stop.
1.3.
Objective of research
According
to the author, in the current state of knowledge, there is little research
focused on a thorough analysis of the impact of the vehicle occupancy
aspect on the transport efficiency in the road network, especially with simultaneous
consideration of parameters such as emissions, energy consumption, delays of
PrT vehicles, deviation from PuT schedule and others. Previous attempts to use
this parameter are limited given the adopted assumptions regarding the
occupancy levels of public transport vehicles and the consequence of their
usage to the traffic condition. Most importantly, virtually none of them
examines the range of usability of the vehicle's occupancy factor depending on
the adopted control parameters. Apart
from some imperfections, the mentioned articles also contain opinions that deny
the practical application of this parameter.
For
this reason, these potential research gaps are the basis for conducting studies
aimed at developing a comprehensive method for active prioritisation of public
transport in the area of intersection. This paper is a description of pilot
studies focused on this aim. The presented analyses and results are included in
a thorough, multidimensional study of the impact of PuT vehicle occupancy on the
control and, consequently, traffic conditions of all road users in the
intersection area.
Thus, at this stage of the work, the author proposes to
verify the following research conjectures:
Conjecture 1: the occupancy of the means of transport is one of
the key variables in urban traffic control strategies accounting for the
prioritisation of public transport.
Conjecture 2: under certain conditions, the occupancy of the means
of transport has a strong influence on traffic conditions for overall
delays for all transport participants.
In the
following section of this paper, the subsequent steps of the proposed method
are described in detail - Section 2. Section 3 presents a verification and
evaluation of the method according to the microsimulation model. Finally,
Section 4 contains the discussion of the obtained research results and provides
conclusions for further work.
2.
PROPOSED METHOD
To
investigate the previously described issues and verify the specific argument
mentioned before, it was decided to prepare a concise sequence of research
tasks. The method presented in this section is based on the simulation approach
technique, which is used to examine a set of different traffic management
strategies. The method consists of five related main steps (Figure 2), and
each of them is a collection of minor sub-steps.
The
first step is to collect and organise all initial data necessary for further
steps. These steps, among others, include the following components:
- properly scaled side
plan of traffic network, which contains geometric intersection, road signs and
horizontal markings. It is used as a background for both traffic modelling and
traffic light design tools as well;
- traffic volumes and
public transport schedules obtained by traffic surveys or received from the
macroscale model;
- vehicle occupancy,
acquired from own measurements or received from traffic management body (for
public transport vehicles).
The
step based on collecting initial data is crucial since it determines the
successful validation of both the microsimulation model and the traffic light
controller.
Fig. 2. Key steps of the proposed method
Based
on these input data, in step 2, the process of designing a traffic light
controller is undertaken. At first (Sub-step 2.1), the controller elements such
as phases of signals are planned, and then intergreen times are calculated.
Afterwards, non-conflicting phases are attributed to the same signal stages.
Minimal tmin and maximal tmax durations of
each signal stage are planned by traffic volumes and vehicle flow structure. In
consequence of this step, a fixed signal program is prepared.
The
signal program intersection has to be provided with an efficient detection
system (Sub-step 2.2). Therefore, it is essential to plan the structure of the
detectors and configure their operation. The detectors are the basis for the
development of the effective structure of the control algorithm and public
transport priority strategy selection.
The
following sub-step 2.3 is dedicated to setting the public transport priority
strategy. The proposed method assumes the occupancy of public transport
vehicles as one of the key factors in the process of traffic condition
improvement. Thus, every PuT vehicle is described by the following parameters:
- number of passengers – pi(t), which is the actual
number of passengers in the i-vehicle,
- vehicle occupancy level – oi(pi(t)), which is a
presentation of the actual number of passengers in the i-vehicle, as a part of
a specific occupancy range; it is classified according to Formula (1):
|
|
|
(1) |
||
|
Where: |
|
|
||
|
N |
– |
number of detected public
transport vehicles (requesting the same priority signal stage), |
||
|
i |
– |
public transport vehicle index, |
||
|
t |
– |
time parameter, |
||
|
pi(t) |
– |
the actual number of passengers
in the i-vehicle, |
||
|
K |
– |
number of occupancy ranges, |
||
|
v0, v1, v(K-2), v(K-1),
vK |
– |
threshold values of ranges. |
||
- reduction factor – ri(oi(pi(t))), the parameter which
determines how much the current signal stage (reserved for PrT flows) can be
reduced (when the demand for its extension is detected) if the priority stage
call of public transport i-vehicle is
also detected. Its value depends on the i-vehicle occupancy level oi(pi) – a
higher level of occupancy causes a lower value of reduction factor. The
received factors are described as elements of a set of reduction factors Xa, Formula (2).
|
|
(2) |
||
|
Where: |
|
|
|
|
a |
– |
set index, a =
1, 2, …, A, |
|
|
A |
– |
number of analysed reduction factor sets, |
|
|
|
– |
value of reduction factor of a-set and j-occupancy range, j = 1, 2, …, K, |
|
|
other markings as
above |
|
||
The
value of the reduction factor is expressed by Formula (3):
|
|
|
(3) |
While
a detection system detects more than one PuT vehicle moving in the same
priority signal stage on non-collision routes, a reduction factor is enhanced.
In the proposed method, the final reduction factor -R(t)
- stands for a product of the
value of reduction factors ri(oi(pi(t))) for each vehicle (Formula 3).
|
|
(4) |
Eventually,
in sub-step 2.4, taking the previously mentioned data and formulas into
consideration, an adaptive traffic algorithm can be designed. The method
presented in this paper assumes rule-based control logic. It means that in each
signalling interval, several decisions (based on information received from the
detection system) are made to determine the further course of the signal
program.
The
next part of the method assumes a modelling approach. At the beginning of
sub-step 3.1, a microscopic simulation model is designed. According to the site
plan, the road infrastructure elements are identified and modelled in the
simulation tool. Next, the traffic generators are introduced, and priority
rules at collision points and velocity limits on arcs are determined. Then sub-step
3.2 starts. It is focused on traffic control rules (designed during step 2),
which are reconstructed and implemented into the simulation model.
The
process of preparing and conducting the experiments is the penultimate stage of
the presented method. It involves the four following sub-steps. First, the
assumptions of individual experiments (as the duration of a single simulation)
are introduced (sub-step 4.1). Then in 4.2, various strategies for traffic
light control are arranged. The total number of scenarios L is equal to
the product of the total number of sets of reduction factors A and the
number of combinations of occupancy threshold ranges C, where C
is the K-value combination of the M-value set, and M is not greater than maximal vehicle
occupancy (MaxVehOcc), Formula (5).
|
|
(5) |
Due
to the numerous combinations of occupancy ranges (For M=MaxVehOcc=200 and K=4, the number of combinations equals
1293699), it is recommended to account for additional assumptions limiting the
size of the set. Each range's upper and lower values can be defined using
Algorithm 1 below. An appropriate selection of variables makes it possible to
select a representative set of scenarios dedicated to simulation experiments.
Algorithm 1.
|
1: |
c = 1
// Initialise c-iterator's
first index |
||
|
2: |
For l1 =
0 to N Step = step1 Do |
||
|
3: |
… |
||
|
4: |
For l(K
– 1) = 0 to N Step
= step(K - 1) Do |
||
|
5: |
For lK=
0 to N Step = stepK Do |
||
|
6: |
If (l1
+ … + l(K - 1) + lK = N) and (l1,
…, l(K - 1), lK ≥ MinRangeSize) Then |
||
|
7: |
v0 = MinVehOcc |
||
|
8: |
v1 = l1 * MaxVehOcc /N |
||
|
9: |
… |
||
|
10: |
v(K – 1) = (l1 + … + l(K - 2)
+ l(K - 1))* MaxVehOcc /N |
||
|
11: |
vk = (l1 + … + l(K - 2) + l(K
- 1) + lK)* MaxVehOcc /N |
||
|
12: |
c = c+1
// Incrementing
c-iterator by the value of 1 |
||
|
13: |
End For |
||
|
14: |
End For |
||
|
15: |
… |
||
|
16: |
End For |
||
|
17: |
C = c
//Execute after last "for" loop. Assign
the last value of "c" to "C". |
||
|
Where: |
|
|
||
|
l1,(K-1),K |
– |
auxiliary
variable acting as iteration counters, |
|
|
step1,(K-1),K |
– |
the value of
iteration step, |
|
|
MinRangeSize |
– |
minimal size
of range (0 < MinRangeSize ≤ 100) [%], |
|
|
MaxVehOcc |
– |
maximal
vehicle occupancy [pers.], |
|
|
MinVehOcc |
– |
minimal
vehicle occupancy [pers.], |
|
|
c |
– |
the
combination of occupancy ranges counter, |
|
|
C |
– |
the number
of combinations of occupancy ranges, |
|
|
N |
– |
the
auxiliary variable. |
|
After
that, a complete collection of scenarios is subjected to the main simulation
experiments (sub-step 4.3). Afterwards, all evaluation data of each simulation
are analysed (sub-step 4.4). These data include, among others:
- queue length on
intersection inlets,
- delays (per vehicle
or/and per person),
- emissions of harmful
compounds to the atmosphere,
- energy (electricity
and fuel) consumption (per vehicle and per network),
- the velocity of
vehicles,
- deviations from the
public transport schedule.
Eventually,
in step 5, the traffic light control system based on the analysed data
proceeds. After that, if necessary, any corrections can be introduced, and the
simulation process has to be run again.
3.
VERIFICATION OF
METHOD
3.1.
General assumptions
The
verification of the method for the presented specific objective was carried out
on a microsimulation model of a fictitious isolated tramway crossing
(Figure 3). In a work related to traffic modelling, the author used the PTV
VISSIM tool. In the part devoted to the issues of designing the operation of
signalling systems, the LISA+ tool was used.
In
step 1, the author adopted specific input data values to obtain relevant
traffic conditions. The basic assumptions include:
- the public transport
system is represented by the tram system, which is characterised by isolated
track routes;
- low intersection
complexity is established, at this stage of the study, to reduce the number of
disturbances that needlessly complicate the public transport prioritisation
method. Thus, public transport stops and pedestrian crossings were not included
in the model;
- constant high
intensity of individual transport (2700 veh./h per direction) –
introduced to ensure a high impact of the type (level) of public transport
priority to individual transport traffic conditions;
- random distribution
of vehicle occupancy, assuming an even distribution of the modal split between
public transport and individual transport (Figure 4);
- a fixed tram
timetable to analyse single as well as multiple tram requests (multiple
requests from queuing trams) (Figure 5).
Fig. 3. View of the microsimulation model
Fig. 4. Private transport vehicle occupancy
distribution
Fig. 5. Public transport vehicle occupancy
– schedule
Fig. 6. Signal phases and stages
In
the following step 2, the tasks focused on the traffic light program are
realised. First, signal phases (signal groups) are created (02, 08, 62, 68) and
assigned to the existing traffic flows. Second, the signal phases are grouped
into two signal stages (1, 2), as presented in Figure 6. Then, the minimal tmin
and the maximal tmax times for each stage are defined as
follows (Table 1).
Tab. 1
Time characteristics of signal stages
|
Phase |
Mode of
transport |
Stage |
Duration of the stage |
|
|
tmin [s] |
tmax [s] |
|||
|
02 |
Private |
1 |
5 |
60/∞* |
|
08 |
||||
|
62 |
Public |
2 |
5 |
12 |
|
68 |
||||
* If no call to stage 2 is
detected, then stage 1 remains indefinitely
Subsequently,
localisations of the detectors are planned, and their operations are
programmed. For detecting the requests from private transport vehicles,
standard induction loops are introduced. In the case of trams, the author
choose a method based on radio transmitters. Because of their application in
combination with the localisation data, the information about occupancy volume
is also transmitted to the traffic light controller (sub-step 2.2).
According
to sub-step 2.3, the number of occupancy ranges K and the number of analysed reduction factor sets A are planned. In consequence, the
values of each factor are proposed. To investigate the presented case study,
the values from Table 2 are assumed.
Eventually,
in step 2.4, a traffic-actuated, acyclic program for traffic lights is designed
for the presented road intersection (Figure 7). Stage 1 is a default stage;
thus, it is activated if the detection system does not receive requests for
other induced stages. Stage 2 is induced.
Tab. 2
Adopted sets of reduction factors, for K=4,
A=8
|
|
|
|
Set of reduction factors Xa |
|
Occupancy range |
|||||||
|
|
|
|
X1 |
X2 |
X3 |
X4 |
X5 |
X6 |
X7 |
X8 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
i-vehicle reduction factor |
xa1 |
0.80 |
0.60 |
0.70 |
1.00 |
0.84 |
0.65 |
0.55 |
0.63 |
1 |
||
|
xa2 |
0.65 |
0.50 |
0.60 |
0.70 |
0.66 |
0.50 |
0.50 |
0.60 |
2 |
|||
|
xa3 |
0.45 |
0.40 |
0.50 |
0.32 |
0.33 |
0.35 |
0.45 |
0.57 |
3 |
|||
|
xa4 |
0.30 |
0.30 |
0.40 |
0.16 |
0.12 |
0.16 |
0.40 |
0.53 |
4 |
|||
The
control algorithm works according to the following principles:
- Stage 1 is always
activated for at least the minimum time – tmin(1); after reaching it, the controller checks
whether there are requests for Stage 2. If not, then Stage 1 time counter
– TCounter1, is stopped at the value of tmin(1);
- The detection of a
request from a public transport vehicle(s) causes the resumption of
Stage 1 time computation and checking of the occupancy level of the
reporting vehicle(s);
- If no extension
reports to Stage 1 are detected, the transition to Stage 2 is immediate.
Otherwise, Stage 1
may be extended up to a maximum time – t'max(1), which is equal to the product of the value tmax(1) and the reduction
factor R(t)
|
|
(6) |
Upon reaching it by TCounter1,
transition to Stage 2 is activated;
- Stage 2 is always
switched on at least the minimum time – tmin(2). After reaching it by TCounter2, the
controller notices the presence of extension requests. Stage 2 may be extended
up to a maximum of tmax(2).
After this time is exceeded, the controller proceeds with the transition to
Stage 1.
Fig. 7. Traffic light control flowchart
3.2.
Microsimulation
experiments
After
completing the actions included in Step 3, the author considers the
microsimulation modelling. A final model (Figure 8), based on common practices
in that traffic modelling approach, is built. The construction of the model in
the Vissim program is completed with the implementation of the traffic light
system constructed in the LISA+ tool. Eventually, the model is used in a series
of simulation experiments (Step 4).
Every
single simulation experiment lasts 11700 seconds. Each of these experiments is
divided into three phases: the warm-up, the main part and the cool-down. During
the warm-up period, in 900 seconds, the network is filled with vehicles to get
stabilised conditions required for further analysis. The successive stage
- the main part of the experiment lasts 3600 seconds. All vehicle movement data
generated in this time interval are evaluated and collected for detailed
analysis. The last stage is the longest (7200 seconds) to ensure that every
vehicle from the previous stage finishes its route and leaves the network. The
vehicles generated in the first and last stages are not considered in the
analysis of the simulation results. According to Formula 3, and the assumptions
from Table 2 (K = 4), a set of examination scenarios is created. To
collect the final set, Algorithm 1 proceeds with some more assumptions:
- step1. step2. step3.
step4 = 1,
- MinRangeSize = 10,
- MinVehOcc = 0 (except driver),
- MaxVehOcc = 200 (assumed maximum of tram occupancy),
- N = 10.
Fig. 8. Structure of the microsimulation
model
Due
to these assumptions, the number of thresholds values combinations of four
occupancy ranges is reduced to
Fig. 9. Obtained combinations of thresholds
values of occupancy ranges
In
consequence, the total number of scenarios L
based on Formula (5) is 8 * 84 = 672.
Furthermore,
the author conducted additional experiments for comparative purposes to the
variants resulting from the presented method. In this case, after detecting a
request to Stage 2, the traffic controller always assumes the same
predetermined value of reduction factor ri(oi(pi(t))) to each PuT vehicle regardless of the actual occupancy.
The
assumed value of the ri(oi(pi(t))) factor ranges from 1 to 0.08 with a gradation equal to 0.02 per
each next experiment, as it is presented in Figure 10. Thus, additional 47
experiments are analysed.
Due
to the high number of experiments in further analysis, the author assumes the
following indication:
-
(a = 1, 2, …, 8; c = 1, 2, …, 84);
-
Fig. 10. The relationship between the
simulation number and the value of the reduction factor in the case of
comparative simulation variants
3.3.
Results of the
simulations
After conducting a series of simulation experiments and
analysing traffic condition data, the evaluation follows. At
this step, vehicle and passenger time losses are considered as only evaluation
data.
The proposed method assumes the occupancy of vehicles as
a decisive factor in the traffic control process. Therefore, the initial
assessment and selection of the results are made based on the average delays
per person in the network. Consequently, the values of delays from experiments
based on the proposed method
In those experiments where traffic control
is independent of vehicle occupancy, the best
score equals 19.05 [s/pass.]. It is obtained for variant
In Table 3, the minimal delay for each of the analysed
set of shortening factor Xa
is underlined. For most of them, that is, for the series (of delay values) where
a = 1, 2, 3, 6, 7, 8, the delays are
higher than the reference variant. The results obtained from simulations where a = 4, 5, tend to be opposite to the
reference variant. Three variants from the first series a=4 and as many as 11 from the second one (a = 5) represent a more efficient solution than variant
An important factor in the interpretation of the results
from Table 3 is the analysis of the coefficient of variation CV defined as the ratio of the standard
deviation σ to the mean value μ. It is used to define how much
the proper selection of the combination of occupancy intervals c impacts the final result of the
experiment depending on the adopted set of reduction factors Xa. Actually, the lowest
value of the coefficient of variation is noted in the series of delay values
where a=8. In this case, it happens due to minimal differences between the
successive values of the reduction factors, concentrated around the value of
0.6, which is similar to the reference variant. The highest value among all
analysed results is obtained for the series of comparative variants
Tab. 3
Obtained results of average
delays per person in a traffic network.
|
Average delay in the transport network - per person
[s/pers.] |
||||||||||||
|
|
|
|||||||||||
|
|
a |
1 |
2 |
3 |
4 |
5 |
6 |
7 |
8 |
|
b |
- |
|
c |
|
|
||||||||||
|
1 |
32.81 |
37.43 |
23.33 |
31.07 |
34.62 |
38.86 |
32.68 |
20.51 |
1 |
31.28 |
||
|
2 |
27.88 |
35.67 |
21.97 |
32.23 |
35.70 |
39.08 |
32.39 |
20.31 |
2 |
30.11 |
||
|
… |
3 |
29.04 |
||||||||||
|
16 |
20.31 |
24.80 |
21.05 |
19.65 |
19.70 |
29.03 |
24.12 |
19.77 |
4 |
28.16 |
||
|
17 |
20.10 |
23.22 |
21.70 |
19.20 |
18.95 |
25.32 |
24.71 |
19.64 |
5 |
26.87 |
||
|
18 |
19.90 |
24.58 |
19.70 |
18.96 |
19.95 |
28.33 |
24.17 |
19.17 |
6 |
26.12 |
||
|
19 |
20.19 |
24.72 |
20.84 |
19.41 |
19.39 |
28.78 |
24.02 |
20.22 |
7 |
25.28 |
||
|
20 |
20.31 |
24.80 |
21.05 |
19.65 |
19.70 |
29.03 |
24.12 |
19.77 |
8 |
24.73 |
||
|
21 |
20.10 |
23.22 |
21.70 |
19.20 |
18.95 |
25.32 |
24.71 |
19.64 |
9 |
24.01 |
||
|
22 |
19.90 |
24.58 |
19.70 |
18.96 |
19.95 |
28.33 |
24.17 |
19.17 |
10 |
23.12 |
||
|
23 |
20.45 |
24.72 |
20.93 |
20.05 |
19.70 |
29.03 |
24.10 |
19.80 |
11 |
23.03 |
||
|
24 |
19.84 |
23.97 |
21.26 |
18.94 |
18.95 |
25.32 |
24.83 |
19.71 |
12 |
22.84 |
||
|
25 |
19.97 |
24.08 |
19.50 |
18.96 |
19.95 |
28.33 |
24.19 |
19.12 |
13 |
22.45 |
||
|
… |
14 |
21.90 |
||||||||||
|
37 |
20.60 |
25.61 |
20.89 |
20.27 |
19.45 |
29.57 |
23.58 |
19.74 |
15 |
21.69 |
||
|
38 |
20.25 |
24.39 |
21.69 |
19.72 |
18.64 |
25.86 |
23.86 |
19.64 |
16 |
21.40 |
||
|
39 |
20.13 |
21.03 |
20.39 |
19.82 |
21.06 |
22.70 |
21.70 |
19.20 |
17 |
20.97 |
||
|
40 |
20.41 |
25.50 |
20.66 |
20.00 |
19.10 |
29.31 |
23.27 |
20.19 |
18 |
20.85 |
||
|
41 |
20.60 |
25.61 |
20.89 |
20.27 |
19.45 |
29.57 |
23.58 |
19.74 |
19 |
20.16 |
||
|
42 |
20.25 |
24.39 |
21.69 |
19.72 |
18.64 |
25.86 |
23.86 |
19.64 |
20 |
19.96 |
||
|
43 |
20.13 |
21.03 |
20.39 |
19.82 |
21.06 |
22.70 |
21.70 |
19.20 |
21 |
19.05 |
||
|
44 |
20.34 |
25.25 |
21.01 |
20.77 |
19.45 |
29.57 |
24.49 |
19.78 |
22 |
19.40 |
||
|
45 |
20.06 |
23.40 |
21.53 |
19.85 |
18.64 |
25.86 |
24.56 |
19.73 |
23 |
19.77 |
||
|
46 |
20.90 |
20.94 |
20.04 |
19.72 |
21.06 |
22.70 |
21.44 |
19.24 |
24 |
19.19 |
||
|
… |
25 |
20.37 |
||||||||||
|
52 |
20.62 |
24.84 |
20.93 |
21.03 |
19.70 |
28.75 |
23.74 |
19.74 |
26 |
19.48 |
||
|
53 |
20.39 |
23.85 |
21.73 |
20.09 |
18.82 |
23.89 |
23.58 |
19.65 |
27 |
24.24 |
||
|
54 |
20.27 |
20.83 |
20.44 |
20.37 |
21.30 |
21.75 |
21.77 |
19.21 |
28 |
24.89 |
||
|
55 |
20.51 |
24.60 |
20.71 |
20.75 |
19.31 |
28.48 |
23.65 |
20.19 |
29 |
29.36 |
||
|
56 |
20.62 |
24.84 |
20.93 |
21.03 |
19.70 |
28.75 |
23.74 |
19.74 |
30 |
32.82 |
||
|
57 |
20.39 |
23.85 |
21.73 |
20.09 |
18.82 |
23.89 |
23.58 |
19.65 |
31 |
31.03 |
||
|
58 |
20.27 |
20.83 |
20.44 |
20.37 |
21.30 |
21.75 |
21.77 |
19.21 |
32 |
31.73 |
||
|
59 |
20.31 |
24.68 |
21.06 |
21.55 |
19.70 |
28.75 |
23.94 |
19.79 |
33 |
35.38 |
||
|
60 |
20.20 |
23.36 |
21.59 |
20.23 |
18.82 |
23.89 |
25.14 |
19.74 |
34 |
40.38 |
||
|
… |
35 |
39.38 |
||||||||||
|
66 |
20.66 |
23.70 |
20.93 |
21.16 |
19.62 |
22.13 |
23.70 |
19.74 |
36 |
43.44 |
||
|
67 |
20.91 |
22.60 |
21.43 |
20.13 |
18.90 |
21.83 |
23.88 |
19.65 |
37 |
45.23 |
||
|
68 |
20.86 |
20.69 |
20.18 |
20.89 |
21.38 |
21.04 |
21.47 |
19.21 |
38 |
45.63 |
||
|
69 |
20.36 |
23.24 |
21.06 |
21.23 |
19.62 |
22.13 |
24.37 |
19.79 |
39 |
43.29 |
||
|
70 |
20.49 |
22.11 |
21.86 |
20.39 |
18.90 |
21.83 |
23.90 |
19.74 |
40 |
42.55 |
||
|
71 |
20.98 |
20.87 |
20.04 |
20.55 |
21.38 |
21.04 |
22.17 |
19.25 |
41 |
49.61 |
||
|
72 |
21.03 |
22.29 |
22.23 |
20.52 |
19.58 |
22.26 |
24.05 |
19.87 |
42 |
50.51 |
||
|
… |
43 |
52.40 |
||||||||||
|
75 |
20.36 |
23.24 |
21.06 |
21.23 |
19.62 |
22.13 |
24.37 |
19.79 |
44 |
53.01 |
||
|
76 |
20.49 |
22.11 |
21.86 |
20.39 |
18.90 |
21.83 |
23.90 |
19.74 |
45 |
53.88 |
||
|
77 |
20.98 |
20.87 |
20.04 |
20.55 |
21.38 |
21.04 |
22.17 |
19.25 |
46 |
81.35 |
||
|
… |
47 |
81.35 |
||||||||||
|
81 |
19.61 |
21.57 |
21.59 |
20.57 |
19.69 |
22.85 |
24.54 |
19.80 |
- |
- |
||
|
82 |
21.58 |
20.79 |
20.59 |
20.52 |
21.85 |
21.46 |
22.14 |
19.37 |
- |
- |
||
|
83 |
21.46 |
20.22 |
20.51 |
19.88 |
21.36 |
21.85 |
22.02 |
19.41 |
- |
- |
||
|
84 |
21.34 |
20.47 |
20.25 |
19.83 |
20.56 |
21.91 |
22.17 |
19.60 |
- |
- |
||
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
Minimal
value |
19.61 |
20.22 |
19.50 |
18.94 |
18.64 |
21.04 |
21.44 |
19.12 |
|
- |
19.05 |
|
|
μ |
21.62 |
25.12 |
21.00 |
22.84 |
23.14 |
27.62 |
24.64 |
19.70 |
|
- |
32.61 |
|
|
σ |
2.30 |
4.33 |
0.80 |
5.17 |
5.99 |
5.69 |
2.76 |
0.36 |
|
- |
14.89 |
|
|
CV |
0.11 |
0.17 |
0.04 |
0.23 |
0.26 |
0.21 |
0.11 |
0.02 |
|
- |
0.46 |
|
As the next step of the evaluation, process passenger
delay analysis for the modal split is conducted. Figures 11 and 12 represent
the results received in this approach. The variants mentioned above as better
than
It
is noteworthy as every improvement of the average delay in the network is
strongly related to a high level of priority for public transport. It does not
mean that all scenarios based on the proposed method improve the traffic
conditions, which is shown in Figure 12. Moreover, in all variants marked as
better than the comparative variant, there is a very similar proportion between
the time losses for journeys made by PuT and PrT. For PrT, the delay is between
27 and 30 [s/pers.], and for PrT, is between 7.85 and 10.81 [s/pers.].
The
comparative variant
Ultimately, the delay for public transport (PuT) vehicles
is analysed in Figure 13. In this part of the study, the results obtained from
experiments
Fig. 11. Delay per person for the modal split
Delay values Background
color
Fig. 12. Delay per person for the modal split
In Figure 13, the dependence of the delay splits into seven intervals
of the number of vehicles in a given simulation experiment is shown. The
obtained results show that the three variants based on the proposed method
ensure the possibility of faster travel (delay per vehicle is less than 10
seconds) for more vehicles than in the case of the reference variant.
Additionally, Figure 14 relates these conclusions to the
number of people travelling in the simulated transport network. More than twice
as many people travel with a delay <10 seconds, and approximately five times
fewer people travel with a delay of over 30 seconds for the recommended
variants compared to reference variant
Fig. 13. Delay per PuT vehicle
Fig. 14. Distribution of the number of
travellers for delay intervals
4.
CONCLUSIONS
The traffic light control method presented in this
article assumes the dependence of the priority level for a public
transport vehicle on its current occupancy status. The method was tested at
a simple intersection in the form of a tram line and road crossing.
Under
given traffic conditions, the use of the method allows reducing the average
time losses per person in the modelled road network. This parameter is reduced
by 2% compared to the most efficient variant that does not use the vehicle
occupancy parameter, that is, independent of the occupancy ratio. It was
reached while maintaining the priority of public transport over individual
transport.
The
experiment presented in this article constitutes a pilot study focused on one
of the author's research areas. This study aimed to check whether traffic control
using information about the occupancy of public transport vehicles might contribute, in given initial conditions, to
the improvement of traffic.
The
conducted calculations based on data obtained from the simulation approach
confirm the above method. The use of
information regarding vehicle occupancy and appropriate boundary conditions
significantly affects the efficiency of road traffic and should be considered
in further research processes.
The
influence of a given type of intersection on the achieved results is worth
noting. In the case of a tram crossing intersection, each trip of a public
transport vehicle causes only losses for individual transport. For junctions
with a more complex geometric layout and control system, the priority phase may
allow vehicles to move in a conflict-free manner unless such conflict is
permitted, given the public transport vehicle routes.
References
1.
Bhouri Neila,
Fernando J. Mayorano, Pablo A. Lotito, Salem Habib Haj,
Jean Patrick Lebacque. 2015. „Public transport priority for
multimodal urban traffic control”. Cybernetics
and Information Technologies 15(5): 17-36. ISSN: 1314-4081. DOI:
https:/doi.org/10.1515/cait-2015-0014.
2.
Christofa Eleni, Alexander Skabardonis. 2011.
„Traffic signal optimization with application of transit signal priority
to an isolated intersection”. Transportation
Research Record 2259(1): 192-201. ISSN: 0361-1981. DOI:
https:/doi.org/10.3141/2259-18.
4.
Czerepicki Andrzej, Anna Górka,
Jarosław Szustek. 2019. „Analysis
of trams' consumption depending on the type of traffic light used”. Science & Technique 18(6): 490-494.
ISSN: 2227-1031. DOI: https:/doi.org/10.21122/2227-1031-2019-18-6-490-494.
9.
Gaca Stanisław, Wojciech Suchorzewski,
Marian Tracz. 2011. Inżynieria ruchu
drogowego. Teoria i praktyka. [In Polish: Road traffic engineering. Theory and
practice]. Warsaw: WKiL. ISBN: 978-83-206-1707-8.
11. Kaczmarek Mariusz,
Jeremi Rychlewski. 2006. „Tram priority traffic control on complex
intersections”. IFAC Proceedings
Volumes 39(12): 416-420. ISSN: 1474-6670.
DOI: https:/doi.org/10.3182/20060829-3-NL-2908.00072.
13. Kuang Xianyan, Xu Lunhui.
2012. „Real-Time Traffic Signal Intelligent Control with
Transit-Priority”. Journal of
software 7(8): 1738-1743. ISSN: 1796-217X.
DOI: https:/doi.org/10.4304/jsw.7.8.1738-1743.
17. Ma Wanjing, Liu Yue,
Yang Xiaoguang. 2010. „Development and evaluation of coordinated and
conditional bus signal priority approach”. Transportation Research Record 2145(1): 49-58. ISSN: 0361-1981.
DOI: https:/doi.org/10.3141/2145-06.
18.
Molecki Adam. 2016. „Powiązanie
systemów nadzoru transport publicznego i sygnalizacji
świetlnej dla podnoszenia jakości obsługi kominikacją
zbiorową”. [In Polish:
“Increase of public transport quality through connection of PT and
traffic lights management applications”]. Autobusy: technika, eksploatacja, systemy transportowe 17(1-2): 30-35.
ISSN: 1509-5878. Available at: http://yadda.icm.edu.pl/baztech/element/bwmeta1.element.baztech-fbdb8b3d-44cc-4ace-a6c5-51ef81c4f54b.
19.
Oskarbski Jacek, Grzegorz Kuprewicz. 2014.
„Priorytety dla transport zbiorowego z wykorzystaniem system
sterowania ruchem”. [In Polish:
“Priorities for public transport system with the use traffic
control”]. Logistyka 3:
4878-4887. ISSN: 1231-5478. Available at:
http://yadda.icm.edu.pl/yadda/element/bwmeta1.element.baztech-53f30153-4fb5-4239-8b32-a8aaa90a6bd7.
22. Smith Harriet R.,
Brendon Hemily, Miomir Ivanovic, Gannett Fleming. 2005. Transit Signal Priority (TSP): A Planning and Implementation Handbook.
Funded by the Unites States Department of Transportation. Corpus ID: 107522795.
Available at: https://nacto.org/references/smith-harriet-r-2/.
24.
Szeląg Adam, Przemysław Gorczyca.
2006. „Energochłonność tramwajów w systemach
dynamicznego priorytetowania”. [In Polish:
“Energy consumption of trams in system with dynamic priority for street
traffic”]. TTS Technika Transportu
Szynowego 12(5):
41-44. ISSN: 1232-3829. Available at:
http://yadda.icm.edu.pl/baztech/element/bwmeta1.element.baztech-article-BGPK-1269-5263.
27.
Wiedemann Remigiusz. 2019. “Ocena zmian
organizacji i systemu sterownia ruchem w wybranym fragmencie sieci
transportowej”. [In Polish:
“Evaluation of the improvement of organisation and traffic control
system on selected part of the transport network”]. In: Współczesne problemy transportu:
opracowanie monograficzne. T. 5, Sterowanie ruchem. [In Polish: Modern problems of transportation:
monographic study. Part 5, Traffic engineering], edited by Banet Krystian.
P. 20-33. Cracow University of Technology, Cracow, Poland. ISBN: 978-83-953673-5-9. Available at: https://repozytorium.biblos.pk.edu.pl/resources/42018.
Received 11.02.2022; accepted in
revised form 30.03.2022
Scientific Journal of Silesian University of Technology. Series
Transport is licensed under a Creative Commons Attribution 4.0
International License
[1] Faculty of Civil and Transport
Engineering, Poznan University of Technology, Piotrowo 3 Street, 60-965 Poznan,
Poland. Email: remigiusz.j.wiedemann@doctorate.put.poznan.pl.
ORCID: https://orcid.org/0000-0002-8360-0997