Proceedings of 13th IOE Graduate Conference

Peer Reviewed
ISSN: 2350-8914 (Online), 2350-8906 (Print)
Year: 2023 Month: April Volume: 13

Analysis of Traffic Speed on Urban Mid-Block Section Under the

Influence of Roadside Frictions
Ramesh Dhimal a , Anil Marsani b
a,b Department of Civil Engineering, Pulchowk Campus, IOE, Tribhuvan University, Nepal
a rmsh2028@[Link], b anilmarsani@[Link]

Abstract
The present study focuses on the effect of roadside frictions on the selected undivided urban midblock section. The objective of
this study is to identify the significant variables that influence traffic speed in the mid-block section and using those significant
variables creating a model for predicting the traffic speed. The analysis was based on the video graphic data collection on midblock
sections of Kalimati, Gaushala, Kuleshwor and Maitidevi. The passenger car unit, traffic flow, traffic speed and the frequency of side
friction variables was determined. Correlation matrix was generated for the correlation of different independent variables. Multiple
Linear Regression (MLR) was used to predict the traffic speed at the mid-block section from the traffic flow and roadside friction
variables. The regression model has adjusted R square value of 65.6%. The predicted speed showed good correlation with the
observed speed with MAPE of 11.36%. Hence, the model proposed in this study will be useful for finding the traffic stream speed
on undivided urban mid-block section under the condition of side frictions.
Keywords
Side Frictions, Passenger Car Unit(PCU), Correlation Matrix, Multiple Linear Regression(MLR)

1. Introduction stopping on the roadway, and vehicles entering and exiting the
roadway.
1.1 Background Roadside frictions are the result of the unmanaged road
Speed is the primary indicator of traffic performance of a infrastructure, increase in urban population, increase in vehicle
roadway system for the transport planning, management and ownership, limitation of urban road space, etc.. These road
operation. It denotes the quality of comfort received by the frictions, which are common in developing nations along with
traffic flow. It is one of the elements of the fundamental heterogeneous traffic condition, have a negative impact on the
relationships in traffic flow theory, along with density and capacity of urban roadways and seriously jeopardize the safety
volume. It depends on a number of factors, including how of those who use the roads.
drivers behave, the nature of the traffic, the physical The growth of traffic in the road network of the cities of Nepal is
characteristics, and the surroundings in urban corridors. Traffic the major concern for the traffic engineer. Urban roads in Nepal
characteristics are the traffic flow and density under has highly mixed traffic conditions. In addition to heterogeneous
heterogeneous conditions, whereas physical characteristics are condition, the flow of the vehicles are conflicted by various
concerned with road width, geometry, access, road conditions, interruption on the road. Therefore, the focus of this study is
intersection controls and environmental factors are visibility on how roadside frictions affect the speed of the selected urban
conditions, weather conditions, etc.. Speed has been identified midblock segment in mixed traffic.
as one of the indicator that designers can use to examine the
consistency of road and expectancy of drivers on roadways.
Speed models help designers figure out the operating speed of 1.2 Research Objective
anroadway. The general objective of this research is:
Roadside friction are the activities on the side of the carriageway • To develop the traffic stream speed model considering road side
or on the carriageway which would disrupt the normal flow of frictions.
traffic. There is a significant impact of side frictions on traffic
characteristics on urban roads. The Nepal Urban Road The specific objective of the research is:
Standard-2076 (NURS -2076) has suggested the capacity of the • To determine road side friction factors that affect the traffic
road under various traffic conditions. According to NURS-2076, stream speed.
the capacity of the 4-lane 2-way road ranges 4000 PCU/h to
2000 PCU/h based on different traffic conditions like frontage
access, standing vehicles, parked vehicle and cross traffic. The 2. Literature Review
Indonesian Highway Capacity Manual (IHCM) has considered
side friction as a roadside environment and activities for the 2.1 Introduction
determination of the capacity of the road. It defined side friction There are many studies which have focused on the impact of
as the impact on traffic performance from roadside activities on roadside friction on speed [1][2][3][4]. In some studies, the
the road segment as pedestrians walking, along it or crossing, passenger car unit (PCU) has been calculated for the

Pages: 156 – 161

 Proceedings of 13th IOE Graduate Conference

heterogeneous traffic [3][5][6],whereas in other studies, each Where, Vcomb is speed (km/hr) for combined sections, nman is
vehicle type has been considered individually [1][2][4] rather number of parking manoeuvres per minute, td is dwell time (in
than into a common unit (PCU). Most of the studies were seconds), n ped is frequency of pedestrians per minute, nc , n2w ,
focused on the divided four lane two way road [5] [2][1] but the and n3w are the frequency of cars, two wheeler and three wheeler
undivided two lane roads has been only considered in the few respectively[1].
studies[6][7]. Although, undivided roads were considered but
The average speed of vehicles at sections with bus stops,
research was focused on the modelling of capacity and LOS in
pedestrian and on-street parking was found to be reduced by
presence of roadside frictions rather than speed modelling.
21%, 27% and 15% respectively. In the sections with a
For the quantification of roadside frictions, the friction factors combination of all the friction factors, it was found that the
present in the road section has been combined into one common reduction was higher (34%). Among all friction factors, the
unit by the weight of respective friction variable [7][5][3]. The presence of pedestrians had the maximum reduction in speed
impact of roadside frictions were analysed in terms of one than other frictions due to bus stops and on-street parking[2].
common unit of friction that is either Roadside Friction Index The developed model for the stream speed was
(RSFI), Side Friction (SF) or Friction (FRIC)[7][5][3].
Roadside Friction Index (RSFI) was used for the quantification Vcomb = 40.38 − 0.21nman − 0.67td − 0.48n ped
of roadside friction as − 0.26nc − 0.15n2w − 0.29n3w − 0.22nhv
RSFI = ∑ wi Ni
Where Vcomb is speed (km/hr) for combined sections, nman is
Where Ni is frequency of ith type friction elements in a 100 m number of parking manoeuvres per minute, td is dwell time (in
road length and wi is weight factor for the ith type side friction seconds), n ped is frequency of pedestrians per minute, nc , n2w ,
variable [7]. Other study quantified road side friction as Side and n3w are the frequency of cars, two wheeler and three wheeler,
Friction ‘SF’ (events/hr) was calculated as nhv is frequency of heavy vehicles[2].
SF = RW1 ∗ NPSPU + RW2 ∗ NEE + RW3 ∗ NW M In one of the study, it was found that the side friction reduced the
Where, NPSPU is the number of pedestrians and parked vehicles average stream speed in all the study section. Stream speed was
equivalent pedestrian units, NEE is the number of entry- exit reduced by 49-57% in sections with bus stops and bus bays and
vehicles, NW M is the number of wrong movement vehicles and on-street parking reduced traffic stream speed by 45- 67% [8].
RW is the relative weights of friction factor [5]. In another study A model was recommended in other study for calculating the
all side frictions were calculated into single common unit ‘FRIC’ speed for urban roads (undivided four lane two way) with high
as number/200m/h. side friction as
FRIC = 1 ∗ PED + 0.45 ∗ BIC + 0.08 ∗ NMV + 0.37 ∗ PSV
Y = −0.132x2 − 0.126x3 − 0.280x4 − 0.126x5
where, PED is pedestrian cross flow, BIC is bicycle volume, − 0.153x8 + 39.458
NMV is non -motorized traffic volume and PSV is the parked
vehicles [3]. Where, Y is the speed (Km/h), x2 is number of stopping city
buses (veh/200 m/hour), x3 is pedestrian movement
In short, there are only few studies which has focused on the (pedestrian/200 m/hour), x4 is number of parking/stopping
undivided four lane two way road for the modelling of the speed passenger car (veh/200 m/hour), x5 is number of entry vehicles
in presence of roadside friction. Therefore, the present study into the street (veh/200m/hour), x8 is number of heavy vehicles
focuses on modelling of speed in mid-block section of undivided per hour (veh/hour) [9].
road in the presence of roadside frictions under mixed traffic
condition. From the literature review, it was found that the roadside
frictions were selected based on the observation of the study site
for its impact on the speed. Many studies were based on finding
2.2 Findings from the studies
the relationship of speed of traffic stream by considering
The impact of individual friction factors on the traffic speed was roadside frictions individually. In general, the side frictions are
studied on some literature. It was found that the most significant present in combination on the urban roads rather than
parameter was the dwell time of buses in comparison with individually. The speed of the traffic is impacted by roadside
frictions like number of parking manoeuvres and pedestrian frictions in combination. Although few studies has considered
flow[1]. The model was developed to predict the speed on road the multiple side friction elements for modelling the traffic
stretch from the predicted speed of individual friction speed but are limited to the divided road section. Therefore, the
parameters as present study attempts to model speed of traffic on the urban
road considering all possible side friction elements on the
Vcomb = 0.55 ∗Vbs + 0.27 ∗Vpark + 0.18 ∗VPed
undivided four lane two way road.
Where, Vcomb is the speed (km/hr) for combined sections and
Vbs , Vpark , and Vped are the speed (km/hr) for the road with side
friction as bus stops, of pedestrian dense roads, on street parking. 3. Methodology
The speed for the traffic stream was modelled to be
3.1 Research Methodolody
Vcomb = 43.53 − 0.39nman − 0.59td − 0.08n ped
The objective of the study is to develop a model to predict traffic
− 0.18nc − 0.05n2w − 0.27n3w stream speed in presence of roadside frictions. For this, the

157
Analysis of Traffic Speed on Urban Mid-Block Section Under the Influence of Roadside Frictions

Table 1: Summary of Data

Parameters Kalimati Gaushala Kuleshwor Maitidevi
Speed (km/h) 23.9 - 31.7 29.1 - 41.5 27.9 - 35.6 27.9 - 35.6
Flow (PCU/h) 861 - 1823 433 - 1024 697 - 1592 460 - 1592
Pedestrian cross (No/100m/h) 28 - 171 24 - 313 152 - 348 115 - 348
Vehicle Entry (No/100m/h) 85 - 370 24 - 217 44 - 218 23 - 218
Vehicle Exit(No/100m/h) 28 - 399 24 - 241 44 - 239 23 - 239
Vehicle Cross(No/100/h) 85 - 370 24 - 120 44 - 327 23 - 327
Heavy Vehicle(%) 0.67% - 7.69% 0.77% - 2.97% 0.67% -3.01% 0% -3.01%
Bicyle (%) 0% -2.33% 0% - 6.36% 0%- 4.85% 0% -6.31%

Table 2: Mode Share of Vehicle

Vehicle Type Kalimati Gaushala Kuleshwor Maitidevi
Motorcycle 73.50% 75.62% 73.32% 78.97%
Car 10.58% 9.21% 8.71% 8.16%
4 wheeler 6.89% 4.34% 6.45% 6.14%
LCV 4.72% 2.98% 9.59% 2.96%
3 wheeler 0.00% 6.87% 0.00% 3.11%
Bus 4.13% 0.42% 1.79% 0.65%
Truck 0.18% 0.56% 0.14% 0.00%

Figure 1: Research Design

• The weather condition, and road visibility should be
normal sunny days.
video graphic technique was used to collect the data for the
study. The classified traffic volume count, traffic speed and side Based on the criteria above, four undivided road sections i.e.
friction elements were measured from the video [Link] Kalimati, Gaushala, Kuleshwor and Maitidevi were selected. All
classified traffic volume count was multiplied with PCU factors the road sections were four-lane two-way, and had similar cross
to get flow (PCU/h) of the traffic stream. The heterogeneous section features like carriageway width, and sidewalk on both
traffic flow was converted to single unit from PCU factors. The side of road.
independent variables were correlated to find the strength of
correlation between the variables. The multiple linear regression
was performed to develop the linear relationship between speed 4. Data Collection and Analysis
(dependent variable) and flow, side friction (independent
variable). The developed model was validated from the observed For the study, data were collected on four road sections. The
speed data from the another road section whose data were not video recordings were taken from the vantage points which
used for regression analysis. Hence, the validated model was would cover the entire section of the roads. The trap length was
used predict the traffic stream speed of the urban undivided set on each road section. The video recording of the sites were
mid-block section of the road. The research design is as shown conducted in between 11 AM to 5PM. The video was taken
in Figure 1. from the mobile phones at 720 pixel resolution at 30 frame per
second. The equipment used for the primary data collection in
the field were IPhone 12 pro, Samsung S9, and tripod stand. The
3.2 Identification of Study Location summary of the data is presented in Table 1.
This study is focused on the predicting the speed on the mid-
block section of undivided road having roadside frictions on it. 4.1 Traffic Volume
Mid-block section is the road section in between the intersection
of the road. The mid-block section should fulfill the following Five minutes of classified traffic volume count data had been
criteria to be selected: collected on each site for the study period. All the vehicles
passing through an imaginary screen line was counted
categorically on the section of the roads. The proportion of
• The road section should be sufficiently away (above 80 vehicle on each category of all the sites are as shown in Table 2.
m i.e. stopping distance for speed 50 km/h ) from the
intersection so that the speed of the each vehicles are not
affected by acceleration and deceleration while moving 4.2 Sampling
towards or away from the intersection. As different mode have different range of speed, random
sampling may not lead to the good result and thus stratified
• The geometric and pavement features of the road should random sampling based on the proportion of different modes
not affect the traffic characteristics i.e. the mid-block was adopted for the research work. The number of samples
section should be straight far from curves, no up or down required to obtain the representative sample was determined by
gradient, pavement surface should be smooth, etc. the formula
• The cross sectional features should be similar i.e. number X 2 ∗ N ∗ (1 − P)
of lanes, carriageway width, median width, footpath width, n=
ME 2 ∗ (n − 1) + X 2 ∗ p ∗ (1 − P)
undivided carriageway (absence of median), etc.
n = sample size required. X 2 = chi-square value for 1 degree of
• Presence of roadside frictions disturbing the traffic flow. freedom at required level of confidence N = the population size,

158
 Proceedings of 13th IOE Graduate Conference

Table 3: Estimation of PCU Table 4: Correlation Matrix

Vehicle Type PCU Flow Ped Cross Entry Exit Cross HV Bicycle
Flow 1
Motorcycle 0.19 Ped Cross -.269* 1
Car 1.00 Entry .571** -.327** 1
4 wheeler 1.53 Exit .488** -.443** .377** 1
Cross .563** -0.114 .453** 0.207 1
LCV 1.94 HV .618** -.468** .395** .437** .318* 1
Tempo 1.09 Bicycle -0.191 0.090 -0.242 -0.086 -0.029 -0.207 1
*. Correlation is significant at the 0.05 level (2-tailed).
Bus 3.94 **. Correlation is significant at the 0.01 level (2-tailed).
Truck 3.79

P = the population proportion, ME = margin of error. 4.5 Side Friction Variables

The samples of speed were measured for each category of vehicle The frictions on the road sections were extracted from the video
within confidence of interval of 95% and with the margin of error recordings. The number of the frictions were counted on the trap
of 5%. Out of 12422 vehicles in total, 4935 vehicle samples length for each 5 minute interval of each site and converted to
were taken categorically for speed measurement. their respective unit. The trap length distance was different for
each road section so in order to standardize the trap length for
all road section, the unit was taken per 100 m for the frequency
4.3 Traffic Speed of friction variables. The friction variables observed in the field
were a) The pedestrians crossing the road section within the
The space mean speed for each category of vehicle was calculated
considered trap length and their frequency was converted to
at 5 min interval. All the vehicles moving in given stretch of road
Number /100 m / hour. b) The vehicles entering to traffic stream
in a specified time period is known as space mean speed and is
from on-street parking, roadside premises and access were
calculated using the average travel time and roadway segment
categorized as ’Entry of vehicle’ and their frequency was
length. The space mean speed is calculated as given in equation
converted to Number /100 m / hour. c) The vehicles exiting from
below:
n∗L traffic stream to on-street parking, roadside premises and access
SpaceMeanSpeed(u) = N were categorized as ’Exit of vehicle’ and their frequency was
∑i=1 ti
converted to Number /100 m / hour. d) The vehicles making
Where, u = space mean speed, km/hr, L = Trap length in km, ti = U-turns or any other crossing on road section were categorized
Travel time of individual ith over trap length in hours. as ’Crossing of vehicle’ and their frequency was converted to
Number /100 m / hour. e) The bus and truck, categorized as
For the traffic stream speed, weighted stream speed was ’heavy vehicles’ (HV), and was obtained from the classified
calculated from the classified traffic volume count and space traffic volume count within 5 minute interval and converted to
mean speed of each category of vehicle as percentage (%). f) The proportion of the bicycles, categorized as
’Bicycle’, obtained from the classified traffic volume count
∑Ni ni vi within 5 minute interval and converted to percentage (%).
Vm =
∑Ni ni

where, Vm is average stream speed (km/h); ni is total frequency

of ith category of vehicle during each 5-min period; vi is the 5. Results and Discussion
space mean speed of ith category of a vehicle during 5-min
period (km/h); and N= total number of categories of vehicles The data from three road sections were used for model
considered. development and one road section data was used for validation
of the model.
4.4 Estimation of Passenger Car Unit
A Passenger Car Unit is a measure of the impact that a mode 5.1 Correlation Matrix
of transport has on traffic variables (such as headway, speed,
density) compared to a single standard passenger car. This is The correlation matrix was obtained to know the strength of
also known as Passenger Car Equivalent. PCU for each category correlation between the independent variables. The independent
of the vehicle was determined as variables were flow, pedestrian crossing, entry of vehicle, exit
of vehicle, crossing of vehicle, proportion of heavy vehicles,
(Vc /Vi ) and proportion of bicycle. The data of the three road sections
PCU =
(Ac /Ai ) were used for generating the correlation matrix. The correlation
matrix is as shown in Table 4.
Where, Vc and Vi are the average speed (km/h) of passenger car
and vehicle category ‘i’ respectively. Ac and Ai are the respective From the correlation matrix in Table 4, no two independent
projected area (m2 ). The PCU was determined for each category variables had strong correlation with one another. So, all the
of vehicle as shown in Table 3. independent variables were taken for the multiple linear
regression analysis.
Flow (PCU/h) of traffic stream was determined by multiplying
each vehicle category with their PCU factors.

159
Analysis of Traffic Speed on Urban Mid-Block Section Under the Influence of Roadside Frictions

Table 5: Summary of Model

Model R R Square Adjusted R Square Std. Error of the Estimate
6 .819f 0.671 0.656 2.06490
f. Predictors: (Constant), Flow, Entry
Dependent Variable: Speed

Table 6: Coefficients of Variables

Model Unstandardized Coefficients Standardized Coefficients t Sig.
B Std. Error Beta
6 (Constant) 40.308 1.114 36.188 0.000
Flow -0.007 0.001 -0.611 -6.102 0.000
Entry -0.015 0.005 -0.318 -3.169 0.003
a. Dependent Variable: Speed

5.2 Multiple Linear Regression

Figure 2: Predicted Vs Observed Speed
Using all the independent variables, the models were generated
from SPSS software. The significance test was conducted to each
of the predictors at 95% confidence interval. Variable having
p-value of less than or equal to 0.05 and t-test value less than traffic stream speed yielded the following results.
1.96 was considered as a significant variable. The six models R square value = 0.5829 (i.e. 58.29 % variance of original field
were generated. In all the models, traffic flow and ’entry of data is explained by the variance of field data obtained from the
vehicles’ were found significant and other independent variables MLR equation)
were found insignificant. The model with significant variables
and with higher value of r-square was chosen as the developed Root Mean Square Error (RMSE) = 3.66 ( i.e average differences
equation. The summary of model is presented in Table 5 and between values predicted by the model and actual values), Mean
coefficients of significant variables is presented in Table 6. The Absolute Percent Error (MAPE) = 11%, Mean Absolute Error
developed model is (MAE) = 3.51, Chi square calculated value = 19.68, and Chi
square critical value at 5% significance = 4.63
Speed = 40.308 − 0.007 ∗ Flow–0.015 ∗ Entry The RMSE value is 3.66 which shows that the predicted traffic
speed is slight deviant from the regression line i.e. there is slight
The R- Square value = 0.656. The 65.6% variance of the
difference between the observed and predicted speed. Similarly,
dependent variable is explained by the independent variable.
the MAE and MAPE is 3.51 and 11.36%. Generally, the MAPE
Significance F= 0.000 . The p-value is less than the significance value from 10% to 20% is considered good for the accuracy
level (0.05), sample data provides sufficient evidence to conclude of the model. The Chi-Square tests shows that the calculated
that regression model fits the data better than the model with no value of Chi-square is greater than the critical value at 5% of
independent variables (i.e. intercept-only model). The predictor significance level. So, null hypothesis is accepted i.e. there is
variable flow and entry of vehicle in the model fits the data better no significant difference between the values of predicted and
than the intercept only [Link], the F-test determines all of observed traffic stream speed.
the predictor variables are jointly significant.
The side friction variable ’entry of the vehicle’ and traffic flow
were found significant at p-value less than 0.05 and t-test value 6. Conclusions and Recommendations
less than 1.96. The speed of the traffic stream is affected by
the entry of the vehicles from the on-street parking, roadside 6.1 Conclusion
premises and access.
The present study proposed speed model for undivided four lane
The traffic stream speed in absence of roadside friction and flow two way urban mid-block section under the influence of roadside
is 40.308 km/hr (free flow speed). The stream speed in absence of frictions.
flow (flow = 0) and maximum observed roadside friction (Entry
1. This study identified the potential side friction parameter
of vehicle = 370 Number/100m/h) is 34.758 km/h i.e. stream
having influence on the traffic stream speed. The speed on urban
speed reduces by 13.768% w.r.t. free flow speed. Also, in case of
mid-block section is affected by the traffic flow and entry of
maximum observed flow (1823 PCU/h) and maximum observed
vehicle from on-street parking, roadside premises or access.
roadside friction (Entry of vehicle = 370 Number/100m/h), the
stream speed is 21.997 km/h i.e. stream speed reduces by 45.42% 2. The developed traffic stream speed model showed R-square
w.r.t. free flow speed. value of 65.6%. The predicted speed showed the good correlation
with the observed speed at MAPE of 11%.
5.3 Validation of Model 3. The passenger car unit was determined for each vehicle
category on the undivided urban mid-block section.
The model was calibrated from three road sections of Gaushala,
Kuleshwor and Kalimati and for the validation, Maitidevi road 4. The traffic stream speed in absence of roadside friction and
section was used. The speed was predicted from the speed model flow is 40.308 km/hr.
for the Maitidevi road section as shown in Figure 2.
5. The stream speed in absence of flow (flow = 0) and maximum
The regression analysis between the predicted and observed observed roadside friction (Entry of vehicle = 370

160
 Proceedings of 13th IOE Graduate Conference

Number/100m/h) reduces by 13.768%. Also, in case of on transport policy, 8(1):94–100, 2020.

maximum observed flow (1823 PCU/h) and maximum observed [3] Masatu LM Chiguma. Analysis of side friction impacts on urban
roadside friction (Entry of vehicle = 370 Number/100m/h), the roads: Case study Dar-es-Salaam. PhD thesis, KTH, 2007.
stream speed reduces by 45.427 %.
[4] Hareshkumar Dahyabhai Golakiya, Manish Patkar, and Ashish
Dhamaniya. Impact of midblock pedestrian crossing on speed
6.2 Recommendation characteristics and capacity of urban arterials: civil engineering:
transportation engineering. Arabian Journal for Science and
The outcome of this study will certainly justify the Engineering, 44:8675–8689, 2019.
implementation of policy measures to manage the vehicles from [5] Pallavi Gulivindala and Arpan Mehar. Analysis of side friction
on-street parking, access to road and roadside premises. These on urban arterials. Transport and Telecommunication Journal,
all can be managed by providing service roads in urban streets 19(1):21–30, 2018.
in future cities of Nepal. Also, for the mitigation, it is [6] Subhadip Biswas, Satish Chandra, and Indrajit Ghosh. Side friction
recommended to provide the raised median which discourages parameters and their influences on capacity of indian undivided
the vehicles and pedestrians to cross road haphazardly. Decision urban streets. International journal of transportation science and
technology, 10(1):1–19, 2021.
could be made to restrict the on-street parking along the urban
streets so that the roadside frictions could be reduced. [7] Sudipta Pal and Sudip Kr Roy. Impact of roadside friction on
travel speed and los of rural highways in india. Transportation in
Developing Economies, 2:1–12, 2016.
References [8] Amudapuram Mohan Rao, S Velmurugan, and KMVN Lakshmi.
Evaluation of influence of roadside frictions on the capacity of
[1] S Salini, Sherin George, and dan R Ashalatha. Effect of side roads in delhi, india. Transportation research procedı́a, 25:4771–
frictions on traffic characteristics of urban arterials. Transportation 4782, 2017.
research procedia, 17:636–643, 2016. [9] Ahmad Munawar. Speed and capacity for urban roads, indonesian
[2] S Salini and R Ashalatha. Analysis of traffic characteristics of experience. Procedia-Social and Behavioral Sciences, 16:382–387,
urban roads under the influence of roadside frictions. Case studies 2011.

161