Download as pdf or txt
Download as pdf or txt
You are on page 1of 6

Capacity Analysis of Three-legged Signalized Intersection

using Cellular Automata


Anupama Krishnan; Leema Peter1

Abstract: Three legged intersections are very common in urban street network for the constraints of space or other factors. They join the main
road and the minor road together. Signalized intersection controls the traffic flow through different directions in terms of time. Under
heterogeneous traffic condition the vehicles vary in its dynamic and static characteristics and share the same carriageway. The cellular automaton
concept is used to simulate the traffic flow at a three legged signalized intersection. A Cellular Automata model is developed for the movement of
the vehicles under heterogeneous traffic conditions. Simulation results show the ability of the model to capture most important features of traffic
phenomenon. The lane group capacity of the intersection was determined on each of the intersection approaches. The validation of the simulation
program was done based on the observed and the simulated discharge at the intersection.

DOI:

CE Database keywords:

Introduction
Rickert et al (1995) introduced a two lane model consisting of two
Intersections are crucial to road networks, because the efficiency of parallel single-lane models with periodic boundary conditions and
intersections would strongly influence the performance of the whole four additional rules defining the exchange of vehicles between the
traffic system. At the controlled intersections different directions of lanes. Li et al (2005) considered the aggressive lane changing
flow share the same road space and flow is segregated in terms of behaviour of fast vehicle in a realistic two lane cellular traffic model.
time. Three legged intersections are very common in urban street Fast vehicle usually has aggressive lane-changing behaviour when its
network for the constraints of space or other factors. They join the preceding vehicle is a slow one. A new set of lane-changing rules
main road and the minor road together. In developing countries like into the basic symmetric two-lane CA model has been introduced
India, the traffic on roads is highly heterogeneous in nature, with considering the different lane-changing manoeuvres of different types
vehicles of widely varying static and dynamic characteristics. In this of vehicles. Sun et al (2007) considered the spatial and temporal
type of traffic, vehicles do not follow lane discipline and they move resolution of the road network on simulation. Emmerich et al (1996)
freely over the entire width of roadway based on availability of introduced an implicit propagation routine in contrast to the
space. To study this type of complex traffic flow and associated propagation routine of the in NaSch model where routine can be done
vehicular interactions, simulation is considered as an effective tool. parallel for all vehicles in the system.
In this paper the traffic movements at a three legged signalized
intersection is simulated under heterogeneous traffic condition.
This paper presents a CA model developed for heterogeneous
Microscopic traffic flow models called cellular automata (CA)
traffic along a two lane road, and for the traffic flow at the three
developed by Nagel and Schreckenberg is used for simulation.
Cellular automata models are discrete models in which the roadway legged signalized intersection. The developed model is then
is divided into blocks which may contain at most one vehicle and the validated.
vehicles move along blocks. CA is an excellent tool for simulating
real traffic flow, because of its efficient and fast performance when Methodology
used in computer simulation. The philosophy of the CA is to bring
about the maximum possible performance velocity of the simulation Simulation of CA Model
for acceptable calculation accuracy. In a CA, space and time are The traffic flow at a signalised T-intersection is simulated
divided in to discrete cells and steps. A cell exchanges transported using Cellular Automata model. The basic CA model proposed by
units only with the neighbouring cells directly within one time step. Nagel and Schreckenberg is modified to simulate the heterogeneous
A CA model is specified by the rules which control these exchanges. traffic conditions. In the simulation model, the road space is divided
in to imaginary cells of equal size. The movement of vehicles is
Larraga et al (2004) introduced a cellular automata model to simulate implemented using the CA rules. The major components of the model
microscopic traffic flow. Rules of interaction in Nagel and are inputs required, model initialisation, vehicle generation,
Schreckenberg (1992) were modified to better capture driver application of CA rules and the desired output.
reactions to traffic that are intended to preserve safety on the The inputs required for CA are cell size, lane width and length in
highway. As a result, a safety distance parameter α was included in terms of the number of cells for defining the road stretch, vehicle size
the model which allows to simulate different driving strategies. Wu et in terms of the number of cells (length and width), arrival pattern,
al (2005) developed a cellular automata model for the traffic flow at a maximum speed (cells per time step), acceleration and deceleration of
three legged unsignalised intersection. The interactions between each type of vehicle, noise probability, lateral movement
vehicles on different lanes and effects of traffic flow states of probabilities, for signal cycle time and green time along with the
different roads on capacity of T-shaped intersection system were position of the signal.
analyzed. 1) Cell Size: Cell size is an important factor as it affects
computational efficiency and performance of the model. The road
1
Assistant Professor, Civil Engineering Dept, College of Engineering, space is divided in to imaginary cells of equal size. The cell size
Trivandrum, Kerala, University of kerala, India. E-mail: used for the CA model is 1m x 1m to represent the road stretch. The
leems_cet@yahoo.co.in lane width and length of the road stretch is also defined in terms of
number of cells. The intersection considered has a lane width of 7m
and the length of the approach lane considered is 500m.
2) Road layout: The road is having a planar topology.
Square cells of equal size were arranged on a two dimensional lattice
to form a T-intersection. The road layout is illustrated in figure. The
whole layout was divided into four lanes A, B, C and D. Lanes A and
B belong to the major road and C and D are lanes of the minor road.
Each lane was made up of a row or column of cells. There were 500
cells in the major lanes A and B and in the minor lanes C and D.
Number of cells was fixed in such a way that the road stretch
measures about half a kilometer in each leg of the T-intersection,
which is sufficient length to study the flow. The beginning cell of a
lane is indicated by number 1, and the end cell is indicated by Source: Sivanandan et.al, (2008)
number 500. The cell length was considered as 1m. This size was
chosen as it is best suitable to represent heterogeneous traffic in Application of CA Rules
cellular automata. The vehicles are physically represented in the grid as per their
arrival after deciding vehicle size. Location of the vehicle on the road
stretch is considered with respect to x and y coordinates of the left
most front corner of the vehicle. The parallel updation is considered
for updating the vehicle in each time step. All the vehicles are
updated for their speeds as per the rules given for the updation. CA
rules incorporate different maximum speed, acceleration and
deceleration for each vehicle type. The lateral movement is possible
on both sides, if sufficient front gap for improving the speed is not
available. The right and left lateral movement rules are made
according to the right and left front gap availability. Each vehicle has
a whole number speed with values between zero and the maximum
Fig.1 Road Layout Used in the Model speed of the vehicle.
At each time step t, the arrangement of N vehicles on matrix of
cells on road stretch is updated according to the following CA rules:
Step 1: Acceleration
3) Vehicles size : In this study five types of vehicles If vn < vmax then vn = min (vn + ak, vkmax), the speed of the nth vehicle
are considered two wheeler, three wheeler, four wheeler, low is increases by ak, but remains unaltered if vn = ak ≥ vkmax. Here vn is
commercial vehicle and heavy commercial vehicle. After the velocity of the vehicle, vkmax is the maximum velocity of vehicle
type k and ak is the acceleration rate of the vehicle type k in integer
deciding the cell size, length and width of the vehicle in value.
number of cells can be obtained by adding clearances to the Step 2: Decelaration due to other vehicles (vehicles ahead)
vehicles actual length and width. After deciding the size of the If gapfp ≤ vn then vn = max (vn, gapfp-1). Speed of the nth vehicle is
vehicle in terms of number of cells in lateral and longitudinal reduced to gapfp – 1, here gap is as one even a front vehicle is in the
directions vehicle can be physically represented on occupied next cell. Here, gapfp is the front vehicle gap.
number of cells. Step 3: Randomisation
4) Arrival Pattern : The vehicles are generated based If vn > 0 then vn = max(vn – dk, 0). Speed of the nth vehicle is
on Poisson distribution. The mean arrival rate (number of decreased randomly by dk unit with the probability p. But it does not
change, if vn = 0. Here, dk is deceleration of vehicle in cells per time
vehicle) per interval is given as the input. The vehicles are
step of vehicle type k. This rule introduces a random element into the
generated at a given interval of time. This procedure will be model. This randomness models the uncertainties of driver behaviour,
continued until the end of the given simulation time. such as acceleration noise, inability to hold a fixed distance to the
5) Vehicle Speed : Initial speed is generated as per vehicle ahead, fluctuations in maximal speed, and assigns different
normal distribution with a given mean rate and standard acceleration values to different vehicles.
deviation. The maximum speeds assigned for each type of Step 4: Updation
vehicle are shown in the table. Each vehicle is moved forward according to its new speed
Table 1. Maximum Speed Assigned to the Vehicles determined in Steps 1-3.
jn = jn + vn, where jn is the initial position of vehicle in longitudinal
direction and indicates as row number for time step t. The vehicle
movements are similar to the CA traffic flow models.
Step 5: Lateral movement
The lateral movement rules are developed based on the
following criteria:
1. The distance headway in the current frontage (gapfp)
is not larger than the specified speed.
2. The distance headway (gapfn) in the neighbour cells is
larger than that of maximum speed of vehicle type k
Source: Sivanandan et.al, (2008) (vkmax).
6) Vehicle Acceleration and Deceleration: Each type of 3. An empty cell for the vehicle occupations must be empty in
vehicle is assigned an acceleration and deceleration in cells neighbours.
4. The distance headway (gapbn) of the following vehicle in
per time step for the application of the CA rules. the neighbour cells is larger than its specified speed.
Criterion 1 and 2 are known as the trigger criteria or incentive
criteria which imply that the vehicle can be driven faster in the target
neighbour place. The criterion 3 and 4 are known as the safety
criteria.

Trigger criteria:
1. gapfp ≤ vkmax
2. gapfn > vkmax
3. plk ≤ rn
Safety criteria:
1. dwk x dLk = (Empty)(Right side)
gapbn > vkmax
where, rn is the random number, plk is the lateral movement
probability, dwk is the width of the vehicle in number of cells, dLk is
the length of the vehicle in number of cells
If the above trigger and safety criteria are satisfied then only
lateral movement becomes possible at every time step for each Fig2. Site Selected
vehicle.
Step 6: Forced lane changing: 1) Signal Phasing: The signal phasing at the intersection is
In case of forced lane changing where the driver knows the a two phase signal. During the first signal phase the right
direction of movement before reaching the intersection additional turning from lane C is restricted. During the second signal
information is required for driver’s priority of left or right side. phase the straight movements along the major roads is
Trigger criterion is taken as the distance from the intersection. The restricted. Figure below shows the signal phasing at the
distance is taken as 50m. If the position of the vehicle from the intersection.
intersection is less than the prescribed distance from the intersection,
driver changes the lane. Safety criteria are taken as the empty cells
that must be available to move left or right as the case may be. The
vehicle grid is then updated accordingly. Hence, in this case the
vehicle may interrupt the following vehicles and force it to reduce its
speed suddenly.
Trigger criterion:
in > Given distance from the intersection
Safety criterion:
dwk x dLk = (Empty)(Right side or Left side)
The forced lane changing rule is applied to the vehicles when it
reaches the position 50m from the intersection. The direction of
movement of the vehicle is decided by assigning a turning probability
to the vehicles randomly using a random number seed ranging from 0 Fig3. Signal Phase 1
to 1. Based on the assigned turning probability the vehicles moves
left or right. At the three legged intersection a turning probability (p)
of 0.5 is assigned to the vehicles. Considering the movements on the
three lanes A, B and C the turning vehicles are identified as follows:
For lane A there is only through movement.
For lane B:
If p<0.5 the vehicle go straight, and
If p>0.5 the vehicle turns left
For lane C:
If p<0.5 the vehicle turns left, and
If p>0.5 the vehicle turns right.
Step 7: Vehicle movement at intersection
The vehicles approaching the intersection moves forward
according to the signal phase. During the red time the velocity of the Fig4. Signal Phase 2
vehicles entering the exit cell is reduced to zero and the following 2) Signal Timing: The signal timing at the intersection is
vehicles queue behind the leading vehicle. The following vehicles shown in the figure below
queue up based on the CA rules, acceleration, deceleration and
updation. During the green time the vehicles are accelerated based on
the assigned acceleration values and they exits the approach based on
the CA rules. Vehicles will be discharged from the exit cell
based on the turning probability.

Data Collection
Data collection includes signal phasing, signal timing of
intersection, video recording process of traffic and intersection
geometry. The three legged intersection shown in figure 2 near
LMS in Trivandrum city was selected. Video recording of the
traffic movements have been done at the intersection for one
hour. Fig5. Signal Timing
3) Vehicle Volume: The classified volume of vehicles on each of
the approach is extracted. The obtained volume is converted to
Passenger Car Units (PCU). This traffic volume is used to find the
arrival rate of vehicles on each approach.
Arrival rate, λ=q/3600
Where, q is the volume of vehicles on each approach.

Output
The output obtained from the simulation run includes the
classified volume of vehicles on each of the approach during each
cycle time and the position time data of the vehicles on each lane.
The output is printed as a text file which can be viewed in Notepad or
MS Word. Figures 6 and 7 shows the sample of the output files.

Fig8. Graph showing the vehicle movement along the lane

Fig9. Graph showing the movement of vehicles along the width of


the lane

Capacity of each approach


Fig6. Sample output showing the position of the vehicles
Capacity at signalized intersections is based on the concept
of saturation flow and saturation flow rate. As per HCM 2000 the
capacity of a given lane group is given by the equation
ci = si (gi / C)
where,
ci = capacity of lane group i (veh/h)
si = saturation flow rate for lane group i (veh/h), and
gi / C = effective green ratio for lane group i.

Saturation flow rate is the maximum discharge rate during


the green time. Vehicles are recorded when their front axles cross the
stop line. The measurement starts at the beginning of the green time
or when the front axle of the first vehicle in the queue passes the stop
line. Saturation flow, however, is calculated only from the headways
after the fourth vehicle in queue passes the stop line. For practical
purposes, prevailing saturation flow rates are usually expressed in
vehicles per hour per lane. As a result, their values also depend on
traffic flow composition. The default value is expressed in passenger
cars per hour per lane.
For measuring the saturation flow rate from field an
observation point was selected where the stop line for the surveyed
lane and the corresponding signal heads are clearly visible. The
reference point is normally the stop line. Vehicles should consistently
stop behind this line. When a vehicle crosses it unimpeded, it has
Fig7. Sample output showing the classified volume of vehicle entered the intersection conflict space for the purpose of saturation
discharged during each cycle flow measurement. Left- or right turning vehicles yielding to
opposing through traffic or yielding to pedestrians are not recorded
The position of the vehicles along the lane is represented until they proceed through the opposing traffic.
graphically in figures 8 and 10. Figure 8 shows the movement of the The period of saturation flow begins when the front axle of
vehicles along the length of the lane. Figure 9 shows the movement the fourth vehicle in the queue crosses the stop line or reference point
of the same vehicles along the width of the lane. and ends when the front axle of the last queued vehicle crosses the
stop line. The last queued vehicle may be a vehicle that joined the Table1: Chi-square value for Lane A
queue during the green time. Measurements are taken cycle by cycle.
To reduce the data for each cycle, the time recorded for the
fourth vehicle is subtracted from the time recorded for the last vehicle
in the queue. This value is the sum of all headways for (n – 4)
vehicles, where n is the number of the last vehicle surveyed (this may
not be the last vehicle in the queue). This sum is divided by the
number of headways after the fourth vehicle [i.e., divided by (n – 4)]
to obtain the average headway per vehicle under saturation flow. The
saturation flow rate is 3,600 divided by this value. In order to obtain a
statistically significant value, a minimum of 15 signal cycles with
more than eight vehicles in the initial queue is typically required. An
average of the saturation flow rate values in individual cycles Table2: Chi-square value for straight movement in Lane B
represents the prevailing local saturation flow rate for the surveyed
lane.
The saturation flow rate for the through movement in lane
A, lane B and for the right turn movement in lane C are found out
based on the field data obtained using the above mentioned method.
The green time is continuous for lane B and lane C therefore the
volume of vehicles for one hour is taken as the saturation flow rate.
The simulated saturation flow rate is determined from the position
time output obtained from the simulation run.
The observed capacity of lane group on each approach is
compared with the capacity value obtained from the simulation Table3: Chi-square value for left turn movement in Lane B
output. Table 3.8 shows the percentage error between the capacity
values obtained from the field and from the simulation output.

Table 3.8 Percentage error of capacity values


Capacity (veh/hr)
Percentage
Observed Simulated error (%)
Lane A Straight 1872 2062 9.21
Straight 1934 2160 10.46 Table4: Chi-square value for right turn movement in Lane C
Lane B
Left turn 711 765 7.06
Right turn 131 139 5.75
Lane C
Left turn 114 124 8.06

The percentage error of the capacity values obtained from


the field and simulation output ranges from 5.75 to 10.46%.

Validation of the Simulation Program


Table5: Chi-square value for left turn movement in Lane C
Validation of the simulation program is carried out by comparing
the simulated value to the observed value. The traffic flow at the
intersection on three different days in a week i.e., Wednesday, Friday
and Saturday are recorded. One hour traffic volume during the peak
and off peak period is extracted. The arrival rate during the peak and
off peak time is given as the input and the obtained simulation output
is compared with the observed volume.
Chi-square test is used to measure the discrepancy between the
observed value and the simulation output. The observed chi-square The obtained chi-square values are less than the critical chi-square
(χ2) value is estimated using the following equation: value 42.56 at 95% confidence interval and 5% level of significance;
χ2 = (O-E)2 / E hence the discrepancy between the observed and the simulated values
where, is less.
O is the observed traffic volume, and
E is the simulation output. Conclusion
The observed chi-square value is compared with the critical chi- A Cellular Automata traffic flow model is an efficient tool in
square value corresponding to 5% level of significance and degree of modelling traffic flow. The CA model by Nagel and Schreckenberg is
freedom (n-1), where n is the number of variables. modified to develop heterogeneous traffic flow model. To
The chi-squared values obtained is shown in the table below incorporate heterogeneity, NaSch model is modified by considering
different types of vehicles depending on their dynamic characteristics
like maximum speed, acceleration and deceleration. The lateral
movement of the vehicles along the lane is also considered. The
position of the vehicles moving along the lane is obtained from the
simulation run which shows the movement of the vehicles based on
the CA rules. The discharge of vehicles at the signalized intersection
from each approach has been obtained as the output. . Capacity of
the intersection at each approach was found out based on the HCM
method. Percentage error between the capacity values obtained from
field data and from simulation has been determined, to quantify the
difference between the values. The degree of error obtained ranges
from 5.75% to 10.46%, which shows that there is not much
difference between the observed and simulated capacity values. The
validation of the simulation program has been done based on the
observed and the simulated discharge at the intersection.

References
Larraga, M., Rio, J., Alvarez-Icaza, L., (2005), “Cellular automata
for one-lane traffic flow modeling”, Transportation Research Journal.

Xi-Gang, Li., Bin, Jia., (2009), “Capacity analysis of un-signalised T-


intersection by using cellular automata model”, Physica A.
Maerivoet, S., (2005), “Cellular Automata models for road traffic”,
Physics Reports.

Rickert, M., Nagel, K., Schreckenberg, M., Latour, A., (1995), “Two-
lane traffic simulation using cellular automata”, Physica.

Emmerich, H., Rank, E., (1997), “An improved cellular automata


model for traffic flow simulation” , Physica A

Xi-Gang, Li., Bin, Jia., (2006), “ A realistic two-lane cellular


automata traffic model considering aggressive lane-changing
behaviour of fast vehicle” , Physica A

Gowri, A., Sivanandan, R., (2008), “Evaluation of left turn


channelization at signalized intersection under heterogeneous traffic
conditions”, Transport.

Gundaliya, P.J., Mathew, T .V., Dhingra, S,L., (2010),


“Heterogeneous traffic flow modelling for an arterial using grid
based approach”, Journal of advanced transportation.

You might also like