Available online at www.sciencedirect.

com

ScienceDirect
Procedia - Social and Behavioral Sciences 104 (2013) 360 – 369

2nd Conference of Transportation Research Group of India (2nd CTRG)

Development of Simulation Model for Heterogeneous Traffic with

No Lane Discipline
Mahendrakumar Metkari a, Anuj Budhkarrb, Akhilesh Kumar Mauryac,1
a,b
Post Graduate student, c Assistant Professor
Department of Civil Engineeing,
Indian Institute of Technology Guwahati, ASSAM - 781 039

Abstract
Indian traffic can be considered as chaotic, with various types of vehicles present with no lane discipline. In order to model
Indian traffic condition, it is necessary to incorporate both the v conditions
in a meaningful way. Research has been made on weak lane discipline or vehicular heterogeneity as a single entity, but both
these criteria are not present in a single model. In the present study, an attempt is made to quantify the unaccounted
parameters of heterogeneity for Indian traffic into the existing car-following models to form a modified car-following model.
A simulation model has been developed as a software program to study the performance of the modified car-following model
in replicating Indian conditions. This model is used to simulate the traffic stream and some preliminary results are obtained.
They are validated with field data collected from a major road in Delhi. The model is able to satisfactorily simulate the real-
time traffic conditions. Analysis is carried out for roadway traffic characteristics, distribution of vehicles along roadway width
and speed distribution of vehicles. The model, after extensive validation at a later stage, can be useful for future traffic experts
for application.

© 2013 The Authors. Published by Elsevier Ltd.

© 2013 The Authors. Published by Elsevier Ltd. Open access under CC BY-NC-ND license.
Selection and peer-review under responsibility of International Scientific Committee.
Selection and peer-review under responsibility of International Scientific Committee.

Keywords: Heterogeneous traffic, traffic simulation model, no lane discipline, car following.

1. Introduction

Indian traffic generally consists of two salient features which distinguish it from the traffic pattern of
developed countries: (i) weak or no lane discipline and (ii) heterogeneous traffic. In such traffic streams, vehicle
not only interacts to vehicle present in front of it but also interact laterally with the vehicles present in its
neighbourhood.

1
* Corresponding author. Tel.: +91-361-2582426; fax: +91-361-2582440.
E-mail address: [email protected]

1877-0428 © 2013 The Authors. Published by Elsevier Ltd. Open access under CC BY-NC-ND license.
Selection and peer-review under responsibility of International Scientific Committee.
doi:10.1016/j.sbspro.2013.11.129
 Mahendrakumar Metkari et al. / Procedia - Social and Behavioral Sciences 104 (2013) 360 – 369 361

Lateral interaction is necessary to be inculcated since vehicles also interact laterally due to absence of lane-
discipline. This is also observed from the field data. Therefore, any study on Indian traffic scenario will be
-
Gunay (2008) had made an attempt to developed a model for weak lane discipline traffic but that did not take into
account the. Later, a modified model incorporating the vehicular heterogeneity is proposed by
Ravishankar and Mathews (2011), however, this did not take into account the lateral interaction. This complexity
and clubbing can be represented as a full-fledged traffic model as attempted in this paper.
The objective of this paper is to develop a traffic simulation model which can mimic the Indian traffic road
conditions of heterogeneity and no-lane discipline. The scope of this paper is limited to unidirectional, mid-block
traffic with ideal conditions on straight level road.

2. Literature Review

Literature review of simulation models is divided in two categories (i) car-following models based on
primarily longitudinal interaction of vehicles, and (ii) car-following models based on longitudinal and lateral
interaction of vehicles. Both categories of model are presented briefly in following two subsections.

2.1 Car-following models

Car-following models describe the processes by which drivers follow each other in the traffic stream. These
models have been studied for more than half century (for e.g. Pipes, 1953). It is clear then that a detailed
understanding of this key process is now becoming increasingly important as opportunities for using new
techniques and technologies become available. Literature review yields a long list of car-following models
developed over the years by different. The research team of the General Motors (GM) Company produced five
generations of their car following models which were all based on the analogy that the response of the following
driver (acceleration or deceleration) is a function of the sensitivity of this driver and the stimulus. The Gazis-
Herman-Rothery model (GHR model) (1958), the safety distance or collision avoidance (CA) model (1959),
Linear (Helly) model (1959) are some of the popular car-following models amongst the list. A detailed
description of all such car-following model can be obtained in Brackstone (1999) review paper. Cellular
automata (CA) based models in which the cell is the fundamental units were initially developed by Nagel K. et al.
(1992) for freeway traffic.
Gipps (1981) proposed a car-following model for the response of the following vehicle (FV) based on the
assumption that each driver sets limits to his desired braking and acceleration rates. The model consists of a set of
two equations that the driver of FV selects his speed to ensure that he can bring his vehicle to a safe stop if the
leading vehicle (LV) suddenly stops. The speed of the vehicle is governed by two conditions that the vehicle (i)
speed should not exceed its desired speed; and (2) accelerates to its desired speed with an acceleration initially
increasing with speed and then decreasing to zero as the vehicle approaches its desired speed.
Later Gunay (2003) has modified the Gipps basic car following equation to incorporate the non-lane-based car
following where due to off-centred positions of vehicles, the following driver does not assign leadership fully to
the vehicle most in front. In the case of stopped LV, according to the basic car following case, FV will also stop,
while in the non-lane-based (staggered) car following case, the following driver may have the opportunity to shift
and pass the (stopped) LV. Allowed speed can be the most restrictive of two factors: (a) a speed at which the
vehicle can decelerate (at time of passing) to the maximum speed allowed by the width of available space
referred as escape corridor; and (b) a speed which should allow the following vehicle adequate time to veer
laterally so as to safely avoid a rear-end collision (Gunay, 2007). The former is called maximum escape speed
tveer he latter factor and be used in the second part. However, this
model was developed for only homogeneous traffic but included the weak lane discipline behaviour.
362 Mahendrakumar Metkari et al. / Procedia - Social and Behavioral Sciences 104 (2013) 360 – 369

In 2011, Ravishankar and Mathews have developed a model that incorporates vehicle-type dependent
-type combinations. Three
vehicle classes- namely, car, auto-rickshaw and bus were considered and the parameters b, b* and T
model were evaluated for each vehicle-type pair. These parameters are included in the current study. The
equation of the model for the car-following case is as follows-

(1)

Where, superscripts l and f stand for LV and FV respectively, and subscripts t and denote the quantity at
end of time t (current time) and , where Tj is the reaction time. bij and are the assumed deceleration of
LV and FV respectively. is the sensitivity to the following distance. All the above mentioned parameters are
for a lead-vehicle type i and following vehicle-type j. Si is the effective-size of the LV. stands for longitudinal
position. Similar values of constants (provided in Ravishankar and Mathews, 2011) are also adopted in the
proposed model.

2.2 Lateral and longitudinal control models

Several models have been proposed which have given the relationship between speed and lateral (or
transverse) clearance required for a given vehicle. Arasan and Koshy (2005) developed a simulation model
suitable for replicating heterogeneous traffic flow with weak lane discipline. In their analysis they found that
transverse clearance between LV and FV depends upon their respective speeds and respective types. Transverse
clearance for two different vehicle type pair can be expressed as a sum of each vehicle type. This inculcates the
heterogeneity into practice. Maurya (2007) developed the CUTSiM model for heterogeneous traffic. Here, the
author assumed that Safe Distance Headway (SDH- the headway at which driver feels safe) as well as Lateral
Clearance (LC) linearly increasing with speed. Gupta et al. (1998) and Chakroborty et al. (2004) developed
comprehensive models which describe both lateral and longitudinal control of vehicles based on a force field
(potential field) analogy of the driving environment. But, the computational inefficiency of these models prevents
their use in simulating large traffic streams.

3. Developed model

The section presents the proposed models of this study. The developed model of unidirectional traffic is
motivated by two recent models- car-following model with lateral discomfort by Gunay (2007) and vehicle-type
dependent model by Ravishankar and Mathew (

account the non-lane based car following, or lateral interaction between vehicles. The attempted study clubs these
two models and develops a set of rules for various cases which govern vehicle speed during free flow (Case 0),
basic car following (Case 1), MES-based car following (Case 2), and veer-induced car-following (Case 3),
depending upon the case which would hit at the given time and condition. Following subsection presents a brief
description of these governing rules of the proposed model.

3.1 Governing rules

Any vehicle in an Indian traffic condition is considered to undergo any one of following cases governing its
speed-
Case 0: v0 ) It is based on the desired speed of the driver (which should not be exceeded). This is
governing case when vehicle is in free-flow state. The movement is formulated as per conventional
model (1981).
 Mahendrakumar Metkari et al. / Procedia - Social and Behavioral Sciences 104 (2013) 360 – 369 363

Case 1: v1 (= corresponds to basic car-following if the leader and follower are in line (Eq. 2).
Parameters of vehicle-type dependent model (Mathew, 2011) can be used in conventional car-following
model (Gipps, 1981). All these cases are presented in Figure 1 for better description of different scenarios.

(2)

Case 2: v2 (= that also corresponds to car-following condition in which the speed of the leading vehicle
and the MES (maximum escape speed) value are the main constraints (refer Eq. 3).
are used with modifying the deceleration and other parameters for different vehicle types as calculated by
Ravishankar and Mathews. If FV has greater speed than LV, and if available lateral clearance is greater than
1.5 metres, vehicle can pass easily at its free-flow speed; or MES= . If the gap available is zero, vehicle
cannot pass, or in other words, MES=0.

(3)

Case 3: v3 (= when the amount of lateral shift is significant, more than LV, the
time needed to shift the vehicle laterally (tveer) in order to avoid collision will be the governing factor, even if
effective road width is high enough. That is, MES is treated as secondary in this scenario. To modify this
equation to suit the requirements of Indian traffic, vehicle-type dependent parameters needs to be introduced.

equation,
= (4)
Since is also another parameter which depends on the vehicle-pairs, a sensitivity factor (as per
Ravishankar and Mathew, 2011) can be incorporated with this. Hence the modified equation will take the
following form.

(5)

The governing case in that instance would determine the vehicle velocity.

3.2 Simulation model:

A simulation model which can mimic the road conditions in actual Indian traffic scenario is developed for
unidirectional traffic. The entire system is developed in C-language as an automation model. The system is
updated after every time-steps (Tj) which is assumed to be the reaction time. The simulation process consists of
(1) vehicle generation; (2) vehicle position updation and (3) analysis. The mechanisms involved in the vehicle
generation and the analysis part are adopted from CUTSiM model developed by Maurya (2007). Initially, the
road geometry (length and width) is decided. The vehicles are generated from one end of the road and leave from
the other end similar to an open loop system. A parallel system of updating is adopted. Vehicles are generated
according to defined vehicle types and their composition similar to filed condition. Generated vehicles are placed
randomly within entire width of road with proper lateral and longitudinal gaps between generated vehicles. Then
they are assigned various vehicle/driver related parameters like vehicle type (which governs length and width),
current speed, , risk-
speed for a particular vehicle type (values of 15, 20 and 30 m/s for auto-rickshaw, truck and car respectively are
used in this model), interaction range (perceiving capacity of driver) and lateral clearance of vehicles. The lateral
364 Mahendrakumar Metkari et al. / Procedia - Social and Behavioral Sciences 104 (2013) 360 – 369

clearance values obtained from Nagaraj et al. (1990) are used and a linear trend is assumed for intermediate
values.
Following assumptions are made while governing the vehicle movement-
Movement will be governed with that case which gives maximum velocity for FV at t+T interval.
Velocity should not exceed the desired speed of vehicle, Vj.
Driver will maneuver his vehicle towards that gap which will give him minimum veering time .
Depending upon the governing situation described in Section 3.1, speed of FV at the end of the reaction time,
or is calculated. The first case v0 represent the free flowing behaviour while other three 3 speeds v1,
v2, v3 represent car-following cases. Vehicle lateral position ( ) and longitudinal positions ( ) are updated.
= + and = +( T) (6)
Where and represent the lateral position of subject vehicle at times t+T and t. is the lateral
manoeuvring speed of subject vehicle type. A static value of was used as 0.5 m/s for car, 0.75 m/s for auto
and 0.25 m/sec for bus.

Fig. 1 Three governing cases of car-following used in the model. Note: ERW= effective roadway width

that offers a higher speed at the next time-

FV will move in a speed . Else, car-following mode will be switched. All the feasible gaps through which
vehicle can pass are calculated. If the leader and follower are in line, the amount of veer/shift required to overtake
and time to veer is calculated for each possibility. In that case, v3, calculated according to the veer value will be
taken as . However, if there are no feasible gaps, then the subject vehicle will continue following as per basic
car-following case 1 with velocity v1.
In case, there is no leading vehicle in line with following vehicle, available gap between two neighbouring
leading vehicles is considered and maximum escape speed (MES) is calculated based upon the minimum lateral
gap. Here, case 2 (MES-governed car following) will be dominating. If, at some other available gap, v3 calculated
is found dominating over v2 (means vehicle will tend to veer for a better gap) then the case 3 will fire and value
of v3 is taken as . The magnitude of v2 and v3 will depend on the clearance gap to be maintained which is a
function of the escape corridor offered by the leading vehicles. Whenever v3 is found dominating (lateral shift
case), vehicle is shifted laterally according to the Eq. 9.

3.3 Algorithm

Algorithm of the model includes input conditions and initialization, various cases that can be followed and
desired output. Variables such as driver, roadway, vehicle characteristics and traffic composition can be input by
 Mahendrakumar Metkari et al. / Procedia - Social and Behavioral Sciences 104 (2013) 360 – 369 365

the user. Vehicles will be generated according to given flow and traffic composition. At each iteration upto total
simulation defined by the user, position of each vehicle will be updated based on governing conditions. Updated
vehicle lateral and longitudinal positions of the subject vehicle at the end of every time step is given as output.
Fig. 2 shows the flowchart of this simulation model. In the simulated model, total roadway length of 5 km is
divided in 1 km of generation zone, 2 km of warming-up zone to allow vehicles to interact with each other and
remaining stretch is used for data collection

Fig. 2. Flow chart of the developed traffic simulation model.

3.4 Recording of field data

In order to validate the model, it is necessary to compare the simulation results with real world traffic data.
For this purpose, the video recording of traffic stream on Outer Ring Road, Delhi at Pamposh Enclave (towards
Hauz Khas) was carried out on 28th January 2012 during 9 am to 1 pm by Sreekumar (2012). This road was a six
lane divided road with good road surface condition. Weather conditions were sunny. Top-frontal view of vehicles
(recording from camera located on a Foot-Over bridge) was taken. For the field observations, flow and spot-
speed of vehicles was calculated by noting the in-time and out-time of every vehicle within the section length (of
30 m). Similar scenario was simulated using developed model and speed-flow data points from simulated and
observed traffic are compared. Further, simulated and observed speed distributions of different vehicle types are
also compared. Results of this comparative study are presented in next section.
366 Mahendrakumar Metkari et al. / Procedia - Social and Behavioral Sciences 104 (2013) 360 – 369

4. Analysis and results

For the validation and evaluation of the simulated model various analysis has been carried out and their results
are presented in following subsections.

4.1 Analysis of lateral placement of vehicles

Field data collected is analysed manually to check the lateral placement of vehicles. This consists of creation
of virtual grid on screen by using screen marker software. This analysis grid consists of lateral lines spaced at
0.2m gap as shown in Fig. 3 (a). Similar scenario is simulated and lateral placement of vehicles are calculated
from simulated stream. Observed a simulated results of vehicles lateral placement is plotted in Fig. 3(b). It can be
seen that, observed Indian traffic stream as well as simulated stream do not follow any lane discipline as
centrelines of vehicles are placed all along the width of a roadway. It is observed that simulated results show little
higher vehicle frequency than observed ones. In real traffic stream, driver does not prefer to drive in shoulder
lanes due to frequent interruption by unauthorized parked vehicles, stopped auto or buses, etc. This biasness is
not incorporated in the simulated model which resulted in higher number of vehicle playing shoulder lanes in
simulated stream than the real traffic stream. Both results evaluated statistically for 5% significance level. The
results are tabulated in Table 1.
Table 1: Statistical tests for lane distribution data and speed-flow relationship data from field and model.

Quantity Decision
Lane distribution 4.9 16.919 Failed to reject Null hypothesis
Speed-flow relationship 22.97 44.97 Failed to reject Null hypothesis
Speed distribution of Cars 16.50 16.919 Failed to reject Null hypothesis
Speed distribution of Bikes 7.87 16.919 Failed to reject Null hypothesis
Speed distribution of Trucks 6.00 11.07 Failed to reject Null hypothesis
Speed distribution of Autos 10.04 14.067 Failed to reject Null hypothesis
Null Hypothesis: There is no significance difference between observed and simulation distribution

Fig. 3. (a) Analysis grid used for lateral placement data collection; (b) distribution of lateral placement of vehicles

4.2 Speed-flow and speed-density relationships

Speed- flow values obtained from simulated and observed traffic stream is presented in Fig. 4(a). It can be
observed that nature of observed and simulated data are similar. Statistical evaluation results are presented in
Table 1. The results are found to be satisfying at 5% significance levels. Derived speed-density plot is also
presented in Fig. 4(b).
 Mahendrakumar Metkari et al. / Procedia - Social and Behavioral Sciences 104 (2013) 360 – 369 367

In order to obtain flow characterization result for homogeneous traffic (i.e. cars-only traffic) a traffic stream is
simulated using developed simulation model. Simulation model is checked for basic traffic flow characteristics
(speed-flow-density relationships) obtained from simulation mode. Model is simulated to obtain speed-flow-
density relationship for varying road width conditions, that is, for 2-lane road and 3-lane road. Obtained speed-
flow and speed-density relationships for two lane and three lane road are shown in Fig. 5(a) to 5(d) respectively.

Fig. 4. (a). Speed-flow relationship; and (b) speed-density relationship for simulated and observed data

4.3 Speed distribution of different vehicle types

Speed distribution of different vehicle types (like car, bike, truck and auto) in observed and simulated streams
are also plotted and compared in Fig. 5. Observed and simulated speeds of different vehicle types are also
matches well. Comparison of speeds from observed and simulated also evaluated statistically at 5% significance
level and found satisfactory.
Speed Distribution for Cars Speed Distribution for Bikes
40 40
Simulation Simulation
Freuency (%)

Frquency (%)

Field
Field
20 20

0 0
0 20 40 60 80 0 20 40 60 80
Speed (kmph) Speed (kmph)
Speed Distribution for Truck Speed Distribution for Auto
40 40
Frequency (%)

Simulation Simulation
Frequency (%)

Field Field
20 20

0 0
0 20 40 60 80 0 20 Speed40
Speed (kmph) (kmph) 60 80
Fig. 5. Speed distribution of different vehicle types from simulated and observed traffic stream

4.4 Speed-flow and speed-density relationships of homogeneous traffic

In order to obtain flow characterization result for homogeneous traffic (i.e. cars-only traffic) a traffic stream is
simulated using developed simulation model. Simulation model is checked for speed-flow-density relationships
obtained from simulation model which are basic traffic flow characteristics defines characteristics of traffic
stream. Model is simulated to obtain speed-flow-density relationship for varying road width conditions, that is,
for 2-lane road and 3-lane road. Obtained speed-flow, speed-density and flow-density relationships for two lane
368 Mahendrakumar Metkari et al. / Procedia - Social and Behavioral Sciences 104 (2013) 360 – 369

and three lane road are shown in Fig. 6(a) to 6(d) respectively. All of these plots show the expected nature of
relationship between basic traffic parameters.

Fig. 6: For two lane roads, (a) speed-flow relationship; (b) speed-density relationship; For three-lane roads, (c) speed-flow relationship and
(d) speed-density relationship.

4.5 Other simulation experiments

In order to analyse the simulation model microscopically, special runs of simulation model are performed
under controlled traffic condition. In the experiment, behaviour of every leading and following vehicle pair for
varying effective road widths and varying relative speed between each pair is recorded. Three different cases are
considered in the experiment (a) Effective road width (ERW) < width of vehicle and relative speed is zero (Case
1a) and 10 kmph (case 1c), (b) ERW is just sufficient to pass (Case 2), and (c) ERW > minimum required road
width to pass (Case3). From all the results, it is observed that the subject/following vehicle makes lateral shift to
look for a better opportunity and after completing sufficient veer, it either follow or pass the leading vehicles
according to the ERW, available. It every case, every case following vehicle behaviour was similar to what is

Results of Case 1a are presented in Fig. 7. In this case, ERW is not sufficient for FV to overtake the LV.
Initially FV searches for better opportunity to overtake LV and if there are no interacting vehicles then FV tries to
achieve its desired speed and then continues with its desired speed. However, if there is a LV present in the
interaction zone and FV realises the presence of vehicle and will slow down. As there is no sufficient space for a
FV to pass, it will continue to follow LV with speed equals to speed of LV. This situation where FV follows LV
is termed as car-following phase. At car-following phase, follower will follow the leader by maintain safe
distance headway and is dependent on FV type. Fig. 7 explains behaviour of various leader-follower pairs. It can
be seen from speed-time plot that whenever FV realises that LV is present within the zone of interaction, it slows
down and FV starts following by attaining its speed equal to speed of leading vehicle. From distance headway-
time plot, it can be observed that in car-following phase, FV starts following leader by maintaining sufficient
time headway in order to avoid collision. This minimum distance headway is different for each vehicle type
(although it is not visible in figure due high scale on Y axis).
Although the developed model is calibrated and validated with limited field data set, validation with bigger
field database is recommended. Further, this model can be extended for inclusion of various roadway geometric
conditions such as effect of curves, effect of gradients, etc. Thus, it is hoped that further exploration into this
would open up opportunities to better utilize these modified car-following models in an Indian perspective.
 Mahendrakumar Metkari et al. / Procedia - Social and Behavioral Sciences 104 (2013) 360 – 369 369

5. Concluding Remark

In the present study, an attempt is made to develop a traffic simulation model accounting heterogeneity and
together. Two different models (developed from Gipps model) which deals the
weak lane discipline or heterogeneity issues individually are clubbed with necessary modifications. Scope for
simulation model is limited to uninterrupted unidirectional traffic on straight and level mid-block section.
Developed model is validated at macroscopic and microscopic levels with traffic data collected from Delhi.
Speed-flow-density relationships and speed distribution of different vehicle types obtained from simulated and
real world traffic stream are compared and match is found satisfactory. Certain control experiments are also
carried out to evaluate the performance of different vehicle pairs at microscopic level. Behaviour of vehicle pairs
in simulated stream was found as expected in real world.
300
7 Lateral Position vs Time Distance Headway vs Time

Distance Headway (m)

LV
Lateral position from
shoulder edge(m)

5 200

3 100
LV1

1 0
0 50 Time (s)100 150 0 Time (s)50 100 150

70 Speed vs Time 2000

Longitudinal Position vs Time
1600
Distance (m)
Speed (kmph)

1200
50
800
400
30 0
0 50 Time (s)100 150 0 50 Time (s)100 150
Figure 7 Simulated behaviour for various leader-follower pairs when ERW is lesser than width of follower and relative speed is 0 (Case 1a)

References
Arasan, V. T. and Koshy, R. Z. (2005). Methodology for modeling highly heterogeneous traffic flow. Journal of Transportation Engineering,
131(7):544 551.
Brackstone, M, McDonald, M. (1999). Car-following: a historical review. Transportation Research Part F, 2:181 196.
Gipps, P. (1981). A behavioural car following model for computer simulation. Transportation Research Part B, 15:105 111.
Gunay, B. (2007). Car following theory with lateral discomfort. Transportation Research Part B, 41:722 735.
Gunay, B. (2009). Rationality of a non-lane-based car following-theory. Proc. of the Institution of Civil Engineers, Transport, 162 27 37.
Gunay, B. and Erdemir G. (2011). Lateral Analysis of Longitudinal Headways in Traffic Flow. International Journal of Engineering and
Applied Sciences (IJEAS) Vol.3, Issue 2(2011) 90-100.

International Symposium on Industrial Robotic System - 98, Bangalore, India.

IRC:SP:41- - ublication
Kikuchi, S. and Chakroborty, P. (1992). Car-following model based on fuzzy inference system. Transportation Research Record 1365,
Transportation Research Board, National ResearchCouncil, Washington, DC, pages 82 91.
Maurya, A. K. (2007). Development of a Comprehensive Microscopic Model for Simulation of Large uniterrupted Traffic streams without
Lane Discipline. PhD thesis, Indian Institute of Technology, Kanpur.
Ravishankar, K. V. R. and Mathew, T. V. (2011). Vehicle-type dependent car-following model for heterogeneous traffic conditions. Journal
of Transportation Engineering, ASCE, 137 (11):775 781
Special Report 87:(1965) Highway Capacity Manual HRB, National Research Council, Washington D.C
Sreekumar, M (2012). Development of Car-following Mod . M tech Thesis, Indian
Institute of Technology Guwahati.
- 328.