Evaluation method for pedestrian level of service on sidewalks based on fuzzy neural network model

Abstract

For the purpose of creating excellent walking environment, increasing the proportion of pedestrians and providing a planning and designing basis for the newly-built and rebuilt sidewalks, this paper proposed a comprehensive multi-factor evaluation method for pedestrian level of service on sidewalks based on the quantification of environmental factors. Firstly, pedestrians’ satisfaction questionnaires survey was conducted with intercept survey method on 87 typical sidewalks covering different regions, road grades, road facility and environmental conditions. The rating scale form of the questionnaires was 10 grades and 4300 valid questionnaires were obtained. Then, the factors of traffic conditions, road facility conditions and environmental conditions which affected pedestrians’ satisfaction were analyzed in detail. Image recognition and edge detection methods were used to quantify the environmental factors. Combined with Spearman rank correlation method, the 10 significant influencing factors obtained were verified. The more comprehensive and quantified multi-factors evaluation index system for pedestrian level of service on sidewalks could be proposed. Finally, aiming at the characteristics that pedestrian level of service on sidewalks and its influencing factors were multi-type variables, the fuzzy neural network method was used to establish the comprehensive evaluation model for pedestrian level of service on sidewalks. The error result showed that the accuracy of the model in this research was 0.94 which had a significant improvement compared with the existing linear regression models.

Keywords

Edge detection fuzzy neural network green looking ratio pedestrian level of service on sidewalks quantification on landscape

1 Introduction

As a low-carbon traveling mode, pedestrian traffic has been acquiring more and more attention. In order to create excellent walking environment and increase the proportion of pedestrians, just providing a good planning for pedestrian system is not enough. A scientific evaluation method for pedestrian level of service on sidewalks also played an irreplaceable role in guiding planning and designing for urban sidewalks.

In Western countries, the early valuation method for pedestrian level of service on sidewalks classified pedestrian level of service through the relation model among pedestrian flow rate, pedestrian speed and pedestrian density [6]. Then, Jaskiewicz [8] and Kroll (2006) evaluated pedestrian level of service on sidewalks adding trip quality and its related factors on that basis. And they used the Delphi method to obtain the grades of pedestrian level of service on sidewalks and its influencing factors. In recent years, the evaluation method for pedestrian level of service on sidewalks based on pedestrians’ satisfaction provided by FDOT was used extensively (e.g. [11 , 17]). These researches obtained pedestrians’ satisfaction through questionnaire investigations and established the relation models for pedestrians’ satisfaction and the conditions of road geography and vehicle traffic. The conductions of these researches were abstracted in HCM [18].

In China, the evaluation for pedestrian level of service started from the research proposed by Traffic Engineering Manual [19] from the view of pedestrian traffic flow. Based on the summary and reference of the related achievements, Bian et al. [1, 3] studied on traffic facilities level of service on sidewalks in China, taking non-motorized vehicles and pedestrian traffic conditions into account. Recently, the research on pedestrian level of service on sidewalks was at an initial stage. The existing researches mostly provided the evaluation method for pedestrian level of service from the perspective of the conflict between pedestrians and vehicles at intersections [14, 16].

Pedestrians’ psychological feeling was comprehensively affected by traffic conditions, road facility conditions and environmental conditions. The researches above mostly analyzed from the perspective of the traffic characteristics and the influences of road facility conditions on sidewalks. Environmental conditions such as greening and the visual effect of the landscape on sidewalks were rarely mentioned. Moreover, the research on the quantification of the highway landscape was conducted from the view of drivers [9, 12]. Due to the significant differences of speeds and interests between pedestrians and drivers, the evaluation and quantification methods might not be applicable for the sidewalk environment. There were not many achievements on the quantification of sidewalk environment. Thus, it was necessary to develop the research on extracting and quantifying environmental condition factors which affected pedestrians’ satisfaction.

In addition, the existing methods mostly used linear regression model to evaluate pedestrian level of service on sidewalks [2, 15]. However, the linear regression model may not be applicable for pedestrian level of service on sidewalks and some of the influencing factors (e.g. segregations between pedestrians and vehicles, the existence of on-street parking) which were ordinal variables or categorical variables. Therefore, the modeling method should be improved for evaluating pedestrian level of service on sidewalks influenced by multi-factors.

Aiming at the limitations in the existing achievements, this paper extracted significant influencing factors of pedestrian level of service on sidewalks with the consideration of traffic conditions, road facility conditions and environmental conditions comprehensively. Image processing method was used to quantify qualitative variables of environmental conditions then a complete evaluation index system could be built for pedestrian level of service on sidewalks. Finally, fuzzy neural network was used to establish the relation model for pedestrians’ satisfaction and the significant influencing factors. And the multi-factors evaluation method, which was applicable to all types of variables, for pedestrian level of service on sidewalks was proposed.

2 Experiment design

This research used intercept survey method to obtain pedestrians’ sense of safety and comfort through questionnaires. The question was that “From the perspective of the sense of safety and comfort, which grade will you give for the sidewalk?” The form of the questionnaire scale was 1–10 ratings to suit to pedestrians’ rating scheme in China.

In order to make the conclusions more universally applicable, a large-scale survey was conducted on the sidewalks in Beijing. Regions in which the surveyed sidewalks belonged to contained residential areas, commercial areas, transportation hub areas and general areas. Road grades included expressway, arterial, minor arterial and branch road. The actual scene data of sidewalks, including traffic conditions, road facility conditions and environmental conditions, were collected during the survey. The traffic conditions and road facility conditions were captured directly. The results were shown in Table 1.

The influence of the environment of sidewalks affected pedestrians’ satisfaction was complicated, so the environmental conditions were captured by photographing. The photos should be shot along pedestrians’ walking direction and at the height parallel with pedestrians’ eyes for trying to factually represent pedestrians’ field of view in walking. Typical environmental scenes of sidewalks were presented in Fig. 1.

87 typical urban sidewalks were surveyed and 4300 valid questionnaires were obtained. The proportion of male respondents was 54% while that of the female was 46%. The ages of respondents contained all stages from below 20 to above 60. Nearly 3/5 of them were familiar with the surveyed sidewalks. The survey time covered 7:00am to 7:00pm on both working days and weekends.

3 Methods

This research tried to comprehensively consider the factors of traffic conditions, road facility conditions and environmental conditions. After preventing the cross-impact of each factor through analyzing the interaction of them, the significant influencing factors were extracted. Then, image characteristics extraction method and image edge detection method to quantify environmental factors. Finally, the multi-factors evaluation model for pedestrian level of service on sidewalks was established through fuzzy neural network method.

3.1 Influencing factors and their quantification methods

Traffic condition factors influencing pedestrians’ satisfaction contained pedestrian flow rate, non-motorized vehicle flow and motor vehicle flow. Road facility factors contained the forms of road cross-section, effective width of sidewalks, segregated facility forms between pedestrians and vehicles, segregated facility forms between non-motorized vehicles and motor vehicles, frequency of barriers on the sidewalks and the existence of on-street parking. Environmental factors contained greening of sidewalks, building environment and orderliness degree of shops at the side of sidewalks away from vehicles etc.

3.1.1 Index of traffic conditions and road facility conditions

In the circumstance of different road cross-sections, the spatial layout of pedestrians, non-motorized vehicles and motor vehicles differed and the influencing degree of non-motorized vehicles and motor vehicles acted on pedestrians also varied. When pedestrians were adjacent to non-motorized vehicles (Fig. 2a), motor vehicles had negligible effect on pedestrians due to the long distance. Likewise, when pedestrians were adjacent to motor vehicles (Fig. 2b), non-motorized vehicles had little effect on pedestrians. When pedestrians were adjacent to mixed traffic (Fig. 2c), motor vehicles and non-motorized vehicles both affected pedestrians. Therefore, the quantity of vehicles on the lane adjacent to pedestrians and the proportion of motor vehicles were selected for further analysis.

The categorical variables, segregation forms between pedestrians and the adjacent vehicles and between non-motorized vehicles and motor vehicles, should be quantified before modeling. This research used the Delphi method to obtain pedestrians’ satisfaction rates for the segregation between pedestrians and the adjacent vehicles. Combined with the distance between pedestrian and vehicles, the hierarchy of the segregation forms could be obtained. Then the categorical variables were transformed to ordinal variables for modeling. The corresponding relations were shown in Table 2. Level 1 meant the worst and level 5 meant the best.

3.1.2 Index of environmental conditions

People observed architectural space and natural environment mainly through vision [20], which meant that pedestrians’ satisfaction on perceiving environment of sidewalks was obtained mainly through vision. Gestalt theory proposed that people always saw the entire object then concerned the separate parts. And their feeling about the entire object was not consistent with the summation of the feeling about the parts. Thus, the integrality of the landscape on sidewalks was the core factor influencing pedestrians’ visual satisfaction. This research used the image edge detective method [7, 13] to quantify the integrality of the environmental landscape on sidewalks.

The environmental scene photos of the surveyed sidewalks should be converted into gray images first. Then the gray images would be processed with noise debasing and the edges of the patterns in the images would be enhanced at the same time. Finally, the scenes in the photos could be partitioning based on the edges and the integrality of the environmental landscape on sidewalks could be evaluated according to the quantity of connecting images. The fewer the quantity of patterns meant the higher level of integrality of the landscape on sidewalks. The procedure was detailed as follows.

Step 1. Pretreated color image into gray. The reasonable gray image could be calculated through weighted average to the three components of RGB as follows: $f (i, j) = 0.030 R (i, j) + 0.59 G (i, j) + 0.11 B (i, j)$ (1)

The gray image corresponding to Fig. 1 could be calculated as Fig. 3.

Step 2. Smooth gray images. According to the gray values of pixels to filter and their neighbor points, the gray images were processed by the Gaussian filter through two-dimensional Gaussian convolution. The core vector of two-dimensional Gaussian function could be calculated as follows.

$K = \frac{1}{2 π σ^{2}} e^{- \frac{x^{2} + y^{2}}{2 σ^{2}}}$ (2)

Step 3. Image enhancement. Canny detector was used to calculate the first order partial derivative matrix, amplitude and direction of the gradient of the image in the x direction and y direction. The calculations were as follows.

$\begin{matrix} P [i, j] & = (f [i, j + 1] - f [i, j] + f [i + 1, j + 1] \\ - f [i + 1, j]) / 2 \end{matrix}$ (3)

$\begin{matrix} Q [i, j] & = (f [i, j] - f [i + 1, j] + f [i, j + 1] \\ - f [i + 1, j + 1]) / 2 \end{matrix}$ (4) $M [i, j] = \sqrt{P [i, j]^{2} + Q [i, j]^{2}}$ (5) $θ [i, j] = arctan (Q [i, j] / P [i, j])$ (6)

Step 4. Edge detecting for patterns. Firstly, non-maximum suppression was processed on the amplitude of the gradient. It meant to solve the two gray values of each pixel in the gradient direction within its neighborhood, to search for the partial maximum of the pixel and to define the gray value corresponding to the non-maximum point as 0. The probable edge of the image could be obtained. Then, double thresholds detection was used to reduce false edges and the edge of each pattern could be obtained according to the two value gray matrix. The edge detection results of the environmental images in Fig. 1 were presented in Fig. 4.

Step 5. Integrality index. The integrality of images could be evaluated according to the quantity of patterns in the segmented images.

Furthermore, the green environment on sidewalks was another important environmental factor affecting pedestrians’ satisfaction. The indexes to evaluate green environment include green ratio, green space ratio and green looking ratio. Green ratio and green space ratio were the indexes for measuring regional afforestation quality from the perspective of land use and ecological environment. Green looking ratio was a dynamic evaluating factor from the view of pedestrians’ sensitivity on environment (Deng and Wang 2000). Therefore, green looking ratio was selected as a quantitative index to evaluate the green environment on sidewalks. The calculation method of green looking ratio was as follows. $G = \frac{S_{g}}{S} \times 100 {%$ (7)

Where, G was green looking ratio, %. S_g was the area of the green plants in a photo, m². S was the total area of the photo, m².

3.2 Theory of fuzzy neural network model

The evaluation criterion of pedestrian level of service on sidewalks based on pedestrians’ walking demand was pedestrians’ satisfaction, which had the characteristic of randomness and fuzziness. So the traditional mathematics methods were difficult to stimulate the relationships between pedestrians’ satisfaction and influencing factors. Thus, fuzzy neural network system was used to evaluate pedestrian level of service through training the input and output samples with the hybrid learning algorithm and identifying the parameters of membership functions and the accurate fuzzy coefficients.

Pedestrian level of service was iteratively affected by multi-factors, so the evaluation model should be a fuzzy rule with multi inputs and single output (MISO). The fuzzy neural network system based on Takagi-Sugeno (T-S) model was used to establish the evaluation model. It consisted of the premise network and the latter network which were used to match the fuzzy rules [19]. The system structure was illustrated in Fig. 5.

The first floor of the premise network was the input floor. The connections were linked with each significant influencing factor of pedestrian level of service and passing the values of influencing factors x = [x₁, x₂, ⋯ x_n] ^T to the next floor. The total node of this floor n was the number of the influencing factors.

The second floor was the fuzzification floor. This floor was used to calculate the membership function $u_{i}^{j}$ which influenced pedestrian level of service. It could be expressed with Gauss membership function as follows. $u_{i}^{j} = u_{A_{i}^{j}} (x_{i}) = exp (- (x_{i} - c_{i}^{j})^{2} / b_{i}^{j})$ (8)

Where, i = 1, 2, ⋯ n; j = 1, 2, ⋯ m_i; m_i was the number of fuzzy separation of x_i; $c_{i}^{j}$ , $b_{i}^{j}$ were the center and width of the membership function respectively. The total node of this floor was $\sum_{i = 1}^{n} m_{i}$ .

The third floor was the fuzzy rules floor. The multiple multiplication operators were used to calculate the fitness value of each premise rule. The calculation was: $α_{j} = u_{1}^{i_{1}} u_{2}^{i_{2}} \dots u_{n}^{i_{n}}$ (9)

Where, i₁∈ { 1, 2, ⋯ , m₁ }, i₂∈ { 1, 2, ⋯ , m₂ }, ⋯, i_n∈ { 1, 2, ⋯ , m_n }, j = 1, 2, ⋯ , m, $m \in \sum_{i = 1}^{n} m_{i}$ . The total node of this floor was m.

The fourth floor was the normalization calculation. It was used to avoid the oscillating of the model caused by the difference of the magnitudes among each influencing factor. The total node of this floor was the same as the third floor. The calculation was: $β_{j} = \frac{α_{j}}{\sum_{i = 1}^{m} α_{i}}$ (10)

The first floor of the latter network was the input floor. The input value of node 0 was x₀ = 1. It provided the constant of the pedestrian level of service result.

The second floor was the fuzzy rules floor. It was used to calculate the fitness value of each latter rule. The calculation was: $y_{j} = p_{j_{0}} + p_{j_{1}} x_{1} + \dots + p_{j_{n}} x_{n} = \sum_{k = 0}^{n} p_{j_{k}} x_{k}$ (11)

Where, k = 1, 2, ⋯ n, j = 1, 2, ⋯ m. The total node of this floor was m.

The third floor was the output floor. It was used to output the defuzzification result of pedestrian level of service. $y = \sum_{j = 1}^{m} β_{j} y_{j}$ (12)

The structure of the fuzzy neural network established here was a multilayer forward feedback network of local approximation. The network learning process can be achieved with the error back propagation (BP) algorithm. The error can be determined as follows. $e = \frac{1}{2} (y_{d} - y_{c})$ (13)

Where, y_d was the grade of the actual surveyed pedestrian level of service; y_c was the grade of calculating pedestrian level of service from the model.

The connective weights of the latter network can be calculated as: $\begin{matrix} p_{i}^{j} (k) & = p_{i}^{j} (k - 1) - a \frac{\partial e}{\partial p_{i}^{j}} = p_{i}^{j} (k - 1) \\ - a (y_{d} - y_{c}) (a^{j} / \sum_{j = 1}^{n} a^{j} x_{i}) \end{matrix}$ (14)

Where, a was the learning-ratio of the network; x_i was the input influencing factor of pedestrian level of service; a^j was the multiple multiplication operator of the membership of the influencing factors.

The center and the width of the membership function were as follows respectively: $c_{i}^{j} (k) = c_{i}^{j} (k - 1) - γ \frac{\partial e}{\partial c_{i}^{j}}$ (15) $b_{i}^{j} (k) = b_{i}^{j} (k - 1) - γ \frac{\partial e}{\partial b_{i}^{j}}$ (16)

4 Results

4.1 Examination for the effectiveness of pedestrians’ satisfaction questionnaires

Pedestrians’ satisfaction was received through questionnaires and many respondents were asked to access the same sidewalk. Meanwhile, the rating results were rank variables. Thus, Kandall harmony coefficient of the reliability examination method in psychology was used to test the effectiveness of the data. The coefficient could be determined as follows. $W = \frac{(R_{i} - \bar{R_{i}})^{2}}{\frac{1}{12} K^{2}}$ (17) where, K was the number of respondents; R_i was the sum of grades of the evaluated sidewalks; $\bar{R_{i}}$ was the mean of R_i.

The calculated Kandall harmony coefficients of pedestrians’ satisfaction for 87 surveyed sidewalks were in the range of [0.80, 0.88], which complied with the requirement that the test result should be above 0.7. It implied that the data of pedestrians’ satisfaction questionnaires was reliable.

4.2 Evaluation index system of pedestrian level of service on sidewalks

Based on the selection of influencing factors affecting pedestrians’ satisfaction discussed above, the quantized influencing factors could be obtained. The factors were reported as the second column in Table 3. Spearman rank correlation analysis method was used to analyze the significance of correlation between pedestrians’ satisfaction and each influencing factor. The evaluation index system of pedestrian level of service and its correlations with pedestrians’ satisfaction scores were shown in Table 3. It can be seen in Table 3 that the ten indexes above were all significant influencing factors (p < 0.05). The positive correlation showed the positive relation between each factor and pedestrians’ satisfaction, whereas, the negative correlation showed the negative relation between them. It was worth noting that when the proportion of motor vehicles in the lane adjacent to pedestrians was high, the quantity of vehicles adjacent to pedestrians would be low because of the limited road width. Thus, the higher proportion of motor vehicles could cause the higher score of pedestrians’ satisfaction.

4.3 Classification of pedestrian level of service on sidewalks

Since the original data was pedestrians’ satisfaction ratings of 10 grades, it would transform into the pedestrian level of service of 6 grades with fuzzy clustering method before establishing the models. The detailed calculating process could be found in another paper written by the same authors of this paper [21]. The results were shown in Table 4.

The relationship model of pedestrian level of service and its significant influencing factors was established with the fuzzy neural network method. The evaluation result of pedestrian level of service on the surveyed 87 sidewalks under different levels of influencing factors was presented in Fig. 6.

As shown in Fig. 6, pedestrian level of service on the surveyed sidewalks contained level A to level F. Judging from the variation trend of the input and output curves, the training results of fuzzy neural network model and pedestrians’ actual satisfaction scores were nearly the same. The accuracy of the model could be achieved as the error curve presented in Fig. 7.

The expected results of the model were integer values of 1 to 6 transformed from pedestrians’ satisfaction scores. The actual results of the model were the calculating values which might not be integer according to the self-adaption relationship among pedestrians’ satisfaction and significant influencing factors. Therefore, all of the errors in the range of –0.5 to 0.5 could be regarded as accurate. On this basis, the accuracy of the model was 94%.

5 Discussion

5.1 Accuracy comparisons

The pedestrian level of service on sidewalks surveyed in this research was reviewed with the models in the existing research [2]. The comparison results of the accuracy of the models were calculated and shown in Table 5.

It could be seen in Table 5 that the accuracy of the evaluation model established with fuzzy neural network method had greatly improved comparing with the existing model. It presented that the evaluating method in this research could evaluate pedestrian level of service on sidewalks more accurately.

5.2 Limitation in this research

Aiming at the problem that the influence of environment affecting pedestrians’ psychological feeling was neglected in the existing researches, this paper comprehensively considered multiple influencing factors, i.e. traffic conditions, road facility conditions and environmental conditions. Image recognition and processing technology were used to propose a quantification method for environmental factors of sidewalks. The significant influencing factors of pedestrians’ satisfaction were determined through Spearman rank correlation analysis method and the factors were as follows: pedestrian flow rate, quantity of vehicles in the lane adjacent to pedestrians, proportion of motor vehicles in the lane adjacent to pedestrians, effective width of sidewalks, segregated facilities between non-motorized vehicles and motor vehicles, segregated facilities between pedestrians and adjacent vehicles, frequency of barriers on the sidewalks, existence of on-street parking, green looking ratio and the integrality of environmental landscape on sidewalks. On this basis, a quantified multi-factor evaluation index system was proposed for pedestrian level of service on sidewalks.

For the limitation that the existing evaluation model with linear regression method for pedestrian level of service was just adapted to continuous variables, this paper established the model for pedestrian level of service and significant influencing factors through the fuzzy neural network. This model was widely adapted to ordinal variables and categorical variables. Through the self-adaption training of the model, it could be seen that the accuracy of the evaluation model for pedestrian level of service on sidewalks with the fuzzy neural network method was 0.94. It had a significant improvement comparing with the existing linear regression models.

Footnotes

Acknowledgments

This research was supported by Beijing Technology Planning Project (No.Z141100000714008) and National Natural Science Foundation of China (NFSC) (No. 51208008).

References

Bian

, Wang

and Lu

, Pedestrian level of service for sidewalks, Journal of Southeast University (Natural Science Edition)37(4) (2007), 695–699. (In Chinese).

Bian

, Research on Urban Pedestrian Planning Method [Ph. D. thesis]. Southeast University, Nanjing, Jiangsu, P.R. China, 2007, (In Chinese).

Bian

, Wang

, Lu

, Ma

J.L.

and Tan

D.D.

, Pedestrian Level of Service for sidewalks in China. TRB, National Research Council, Washington, D.C., 2007.

Cheng

F.H.

, A Study on Air Quality Assessment Applied BP Neural Network and Fuzzy Neural Networks [Ph. D. thesis]. Guangdong University of Technology, Guangzhou, Guangdong, P.R. China., 2012, (In Chinese).

Deng

X.J.

and Wang

H.G.

, ratio, space ratio, green looking ratio, New Architecture (4) (2002), 75–76. (In Chinese).

Fruin

J.J.

, Pedestrian Planning and Design, New York, Metropolitan Association of Urban Designers and Environmental Planners, 1971.

Guerrero

J.M.

, Guijarro

, Montalvo

, Romeo

, Emmi

, Ribeiro

and Pajares

, Automatic expert system based on images for accuracy crop row detection in maize fields, Expert Systems with Applications40(2) (2013), 656–664.

Jaskiewicz

, Pedestrian level of service based on trip quality, Transportation Research Board Circular. E-C019: Urban Street Symposium. Washington, D.C.: Transportation Research Board of the National Academies, 2000, p. 114.

Jia

Z.R.

, Guo

Z.Y.

and Liu

X.B.

, Research on quantitative evaluation of visual effect of highway cut-slope landscapes, Journal of Highway and Transportation Research and Development25(11) (2008), 154–158. (In Chinese).

10.

Kroll

, Assessing the Environmental Quality of Walking: Steps Toward a Person-Centered Level of Service. Proceedings of the 4th Asian-Pacific Transportation Development Conference, California, Oakland, 2003.

11.

Landis

B.W.

, Vattikuti

V.R.

, Ottenberg

R.M.

, Mcleod

D.S.

and Guttenplan

, Modeling the roadside walking environment: A pedestrian level of service, Journal of the Transportation Research Board, Transportation Research Record1773(1) (2001), 82–88.

12.

S.W.

, Wang

L.H.

, Sun

W.C.

, Sun

and Jand

F.Y.

, Qiao, Feature quantification method for freeway roadside landscape, Journal of Jilin University (Engineering and Technology Edition)41(4) (2011), 956–962. (In Chinese).

13.

Montalvo

, Guerrero

J.M.

, Romeo

, Emmi

, Guijarro

and Pajares

, Automatic expert system for weeds/crops identification in images from maize fields, Expert Systems with Applications40(1) (2013), 75–82.

14.

Pan

F.Q.

, Lu

, Xiang

Q.J.

and Zhang

G.Q.

, Level of safety service for highway signalized intersections, Journal of Southeast University (Natural Science Edition)38(2) (2008), 298–303. (In Chinese).

15.

Petritsch

T.A.

, Landis

B.W.

, McLeod

P.S.

, Huang

H.F.

, Challa

and Skaggs

C.L.

, Pedestrian level-of-service model for urban arterial facilities with sidewalks, Transportation Research Record: Journal of the Transportation Research Board1982(1) (2006), 84–89.

16.

Qian

D.L.

, Zhang

M.M.

and Zhao

W.T.

, Semi-quantitative evaluation of level of service for pedestrians, Journal of South China University of Technology (Natural Science Edition)40(7) (2012), 33–40. (In Chinese).

17.

Steiner

R.L.

, Landis

B.W.

and Miller

D.W.

, Refinement of the Roadway Segment Level of Service, TRB, National Research Council, Washington, D.C., 2002.

18.

Transportation Research, Board, Highway Capacity Manual, TRB, National Research Council, Washington, D.C., 2010.

19.

“Traffic Engineering Manual” Editorial Board of China Highway and Transportation Society. Traffic Engineering Manual. China Communications Press, Beijing, 1998, (In Chinese).

20.

Zhang

Y.M.

, Zhou

C.L.

, Wang

H.S.

, Liu

Y.C.

and Wang

X.Y.

, Environmental Behavior and Human Engineering. China Electric Power Press, Beijing, 2011, (In Chinese).

21.

Zhao

, Bian

, Rong

and Liu

X.M.

, Evaluation on pedestrian level of service of sidewalks in beijing transportation junction area, Journal of Beijing University of Technology5 (2014), 725–732. (In Chinese).