• Tidak ada hasil yang ditemukan

For the fulfilment of the degree of

N/A
N/A
Protected

Academic year: 2023

Membagikan "For the fulfilment of the degree of "

Copied!
180
0
0

Teks penuh

Therefore, the simulation model developed for such a traffic flow must model the behavior of different vehicles in maintaining longitudinal and transverse spacing. These developed models can then be incorporated into a developed simulation model, alongside a set of algorithms that govern driver behavior in mixed traffic flow.

METHODOLOGY FOR DATA COLLECTION AND EXTRACTION 31 3.1 Representation of interactions between vehicles under heterogeneous traffic with weak lane

81 4.11 Coefficients and residual parameters for plot of LG with CS and speed 86 4.12 Road width-wise classified variation of LG with CS and speed 87 4.13 T-test between model and validation data for difference CS and mean speed.

Units

INTRODUCTION

Introduction

Overview

Traffic stream in developing countries

  • Heterogeneous or mixed traffic
  • Weak or no lane discipline

Due to a high variability of vehicle sizes and operating characteristics, vehicles interact two-dimensionally with their neighboring vehicles and show no (or poor) lane discipline (i.e., vehicles do not move in lanes). Thus, lateral and longitudinal gap maintenance between vehicles become important parameters in such traffic flow.

Staggering

  • Problem statement
  • Objectives of the study
  • Scope of the research work
  • Organization of the thesis
  • LITERATURE REVIEW

Research objective 1: Analysis and modeling of the longitudinal and lateral distances between vehicles at speeds for heterogeneous traffic with weak lane discipline. Research Objective 2: Develop a comprehensive model of driving behavior that includes characteristics and relationships between vehicle differences.

Literature review

Microscopic modeling of traffic stream

  • Conventional car-following models
  • Cellular automaton models
  • Comprehensive traffic simulation models for heterogeneous weak lane disciplined traffic stream
  • Development of traffic simulators
  • Remarks on simulation modeling approaches in mixed traffic stream

Gipps (1981) proposed a car-following model of following vehicle response, based on the assumption that each driver sets limits on his/her desired braking and acceleration speeds. In the developed model, the variation of the longitudinal opening with an increase in dispersion was not studied.

Table 2.1 A brief summary of traffic simulation models developed
Table 2.1 A brief summary of traffic simulation models developed

Modeling parameters of traffic in developing countries

  • Lateral clearance studies
  • Longitudinal headway studies considering lateral staggering
  • Overtaking decision modeling
  • Remarks on literature review of parameters of traffic in developing countries

Thus, few important studies have been conducted on the effect of drivers' lateral behavior in traffic with weak lane discipline (Gunay, 2007; Gurumurthy et al., 2016, Pal and Mallikarjuna, 2013). In heterogeneous traffic with weak lane discipline, overtaking decision-making does not take the form of lane changes, as is the case in the developed world.

Fig.  2.2  Trend  of  the  relationship  between  horizontal  separations  of  the  LV  and  FV,  and  the  following distance
Fig. 2.2 Trend of the relationship between horizontal separations of the LV and FV, and the following distance

Research related to data collection of gap-maintaining parameters

  • Dynamic data collection
  • Remarks on data collection techniques

Measurement of vehicle spacing can be classified based on sensor or image-based techniques. Instrument vehicles can be developed to measure distances between vehicles, as in Wong and Qidwai (2004), Knoetze et al.

Fig. 2.3 Classification of traffic data collection methods  2.3.1 Data collection by static devices
Fig. 2.3 Classification of traffic data collection methods 2.3.1 Data collection by static devices

Outcome of the literature review

Furthermore, a driver's overtaking decisions in mixed traffic conditions based on the distances between vehicles have not been taken into account. There is a need for a combination of these data collection and extraction methods to extract data on vehicle spacing maintenance in mixed traffic conditions.

CHAPTER 3: METHODOLOGY FOR DATA COLLECTION AND EXTRACTION

  • Representation of interactions between vehicles under heterogeneous traffic with weak lane discipline
  • Adopted methods for data collection of parameters of vehicle interaction
    • Motivation behind using instrumented vehicle for measuring lateral distances between vehicles
    • Motivation for video-recording for measuring longitudinal distances between vehicles To find a variation of veering with different vehicle types, speed and longitudinal distance, several
  • Methodology adopted to study lateral interactions
    • Instruments used for data collection
    • Data extraction from instruments
    • Estimation of accuracy
    • Data collection locations for lateral interaction study
  • Methodology adopted to study longitudinal vehicle interactions
    • Camera-calibration technique for data extraction
    • Development of data extraction software for extracting longitudinal interactions Data extraction software is developed in MATLAB programming language. It is used for noting
    • Prerequisites for video-recording of traffic stream The prerequisites for video-recording of traffic stream are-
    • Checking accuracy of data extraction
    • Data collection details
    • Data extraction for multiple-vehicle interactions
  • Data collection of overtaking decision-making criteria
  • Summary

These image coordinates are converted to coordinates in the real field (ie the plane of the road) to find out the lateral or longitudinal distance relative to the edges of the road. Four endpoints marked on the ground representing a rectangle in the field are shown in fig. It is used to note the coordinates of the vehicle as it moves in the traffic flow, on a specific time frame.

For this purpose, trap length and sides of the road are determined and marked in the field. Error in the evaluation of the exact position of a point (Table 3.3) on the field leads to an error in the measurement of speed.

Fig. 3.2. Traffic parameters used in analysis of driver behavior
Fig. 3.2. Traffic parameters used in analysis of driver behavior

MODELING OF INTER-VEHICULAR GAPS

Modeling of inter-vehicular gaps

Analysis and modeling of lateral clearance between vehicles

  • Effect of various traffic parameters on lateral clearance
  • Model development using sensor data
  • Variation of developed model across cities
  • Modeling for lateral clearance with camera calibration data
  • Shying away behavior of heterogeneous interacting vehicles
  • Analysis of simultaneous multiple lateral interactions

Table 4.1 shows that the lateral distance is correlated with the average speed and the speeds of individual vehicles. Thus, to model the variation of lateral distance between vehicles, the average speed can be treated as an appropriate independent variable. A scatterplot of the average speed of the test vehicle and the interacting vehicle versus lateral distance for all sites is presented in Fig.

From the obtained contour plot and from positive value of correlation coefficient between lateral clearance and average speed in Table 4.1, it is clear that LC increases linearly with average speed (𝑣̅). There is an increase in side clearance with an increase in average speed for both the restricted and unrestricted cases and for all vehicle pairs.

Table 4.1 Effect of various traffic parameters on lateral clearance (Overall data, sample size: 6,016)  Variable  Statistical test
Table 4.1 Effect of various traffic parameters on lateral clearance (Overall data, sample size: 6,016) Variable Statistical test

Analysis and modeling of longitudinal gaps

  • Preliminary observation of longitudinal gap data and its comparison with earlier studies
  • Variation of longitudinal gap with centerline separation and average speeds
  • Validation of variation of LG with CS and speed
  • Optimal position of FV in steady state flow using developed model

This can be seen from the obtained coefficients a and b for all pairs of vehicles in table 4.11 and also in the figure. In Table 4.13, for each pair of vehicles, there are fewer labeled cells (significantly different data) than unlabeled cells (significantly similar data set) (a total of 44 out of 66 cells show data similarity). The coefficients a and b of Equation 4.6 for validation and modeling are also similar (as can be seen in Table 4.15).

The parameters obtained in Table 15 are compared with 𝐿𝐺 for individual vehicle pairs (assuming no effect of LV2), calculated as mentioned in Section 4.2.2 (Table 4.11). It is observed that 𝐿𝐺̅̅̅̅𝑑𝑜𝑢𝑏𝑙𝑒 obtained for a combined system at lower 𝐶𝑆𝐸𝐹𝐹 values ​​(<2 m) is significantly less than 𝐿𝐺𝑠𝑖𝑛𝑔𝑙𝑒 (obtained from Table 4 .11).

Fig. 4.13 A plot of LG with (a) centerline separation (CS), and (b) with speed (𝑣̅)
Fig. 4.13 A plot of LG with (a) centerline separation (CS), and (b) with speed (𝑣̅)

Overtaking decision-making criteria

  • Modeling of overtaking decisions
  • Interpretation of model estimates for various vehicle pairs
  • Validation

Data for each vehicle pair is separated by decision to overtake or follow into two datasets. This set of rows basically gives sensitivity of predictor variable on the decision to overtake. This is carried out to compare the effect of various variables on the decision to overtake.

It is observed that for all pairs of vehicles, the introduction of an overtaking barrier strongly affects the overtaking decision. Most vehicle pairs are more sensitive to overtaking barrier and centerline separation than other variables.

Table 4.18. Overtaking decision model estimates for various vehicle-pairs
Table 4.18. Overtaking decision model estimates for various vehicle-pairs

Other parameters evaluated from field data

  • Lateral speed
  • Overtaking time
  • Vehicle dimensions

The same exercise as mentioned in 4.3.1 is used to predict the responses of the validation dataset, which is not used for modeling purposes. The coefficients obtained from the regression model are used to model the new data used for validation. From Table 4.20, it can be concluded that smaller size vehicles such as two-wheelers and three-wheelers have higher lateral velocities than larger size vehicles.

For all overtaking maneuvers, overtaking time distribution is observed to follow log-logistic distribution (two parameters). It is observed from Table 4.21 that the overtaking time is lower for smaller sized vehicles such as three-wheelers and two-wheelers.

Fig. 4.22 (a) Starting and (b) ending condition for calculation of overtaking time
Fig. 4.22 (a) Starting and (b) ending condition for calculation of overtaking time

Summary

Cars and three-wheelers maintain more space with heavy vehicles and LCVs, while two-wheelers maintain less lateral space. Further, it is observed that vehicles keep less lateral space between vehicles of the same type than vehicles of other type. It is also affected by any obstacles to overtaking in the form of other vehicles or road edges, and the longitudinal gap.

The probability of overtaking can be represented by logistic regression models for individual vehicle pairs. For car-car pairs, there is an 8.9% decrease, a 3.2% increase, a 1.9% increase, and a 23% increase in overtaking probability for a 5% increase in average speed, relative speed, longitudinal spacing and centerline separation.

DEVELOPMENT OF DRIVER BEHAVIOR MODEL

Development of driver behavior model

Framework of driver behavior model

  • Free flowing module
  • Vehicle-following module
  • Overtaking module

These vehicles maintain a stable gap depending on the average speed of LV and FV (𝑣̅), their vehicle types, centerline spacing (CS) or LV-FV spacing (see Equation 4.4 and Table 4.11). Due to the limitation of human perception, the FV driver may not be able to accurately detect the LV speed. FV speed detection error is assumed to be a function of LV speed (𝑣𝐿𝑉), FV deviation from stable longitudinal gap (S), and the aggressive/risky nature of the FV driver.

The sensed LV velocity can be calculated from Equations 5.4 (a) for positive errors and 5.4 (b) for negative errors. Here, 𝑉 is the desired speed of the test vehicle, 𝐶5 is a measure of the ratio of the advantage given the proximity of the gap (k) to its offered speed (𝑣𝑘).

Fig. 5.2 Representation of different parameters under approach to stable vehicle-following
Fig. 5.2 Representation of different parameters under approach to stable vehicle-following

Structure of the simulation model

  • Vehicle generation
  • Updating the vehicle position
  • Printing of results
  • Flowchart of simulation model

Microscopic as well as macroscopic parameters need to be captured from the model to compare them with the field for validation purposes. Macroscopic parameters include speed, flow and density plots with a fixed time interval, whereas microscopic parameters in mixed traffic flow include speed distribution for each vehicle type, time distribution, ratio of lateral clearance to speed for different vehicle types, x and y coordinates of all vehicles on the road section to show the weak lane discipline, the relationship between longitudinal propulsion and centerline separation, and the average speed of vehicle following behavior. Speed ​​for macroscopic calculations is the average speed of all vehicles in 1-minute and 5-minute intervals.

The driving behavior model highlighted in this flowchart is mentioned in detail in the flowchart in Figure. The driving behavior model also lists the different modules that the vehicle must go through to determine its position at the next time step.

Fig. 5.5. Flowchart showing the framework of developed simulation model
Fig. 5.5. Flowchart showing the framework of developed simulation model

Calibration and validation of model

  • Input parameters measured from the field
  • Calibration of unknown model parameters
  • Calibration of VISSIM model parameters
  • Validation of developed model and VISSIM with field data

In Table 5.3, the best f(x) represents the minimum value of the optimization function among the function values ​​for all individuals in the population for the current generation, while average f(x) represents their average. Furthermore, various driving behavior parameters used as inputs include the following: (i) Acceleration-deceleration curves are adopted from Bokare (2013); (ii) Lateral clearance maintaining behavior, taken from Table 4.4 and Equation 4.1, is used as input in the form of clearance maintained at 0 km/h and 50 km/h in the VISSIM model, given in Table 5.4. The sensitive vehicle tracking parameters are optimized between the ranges given in Table 5.5, using genetic algorithms.

In Table 5.7, the null hypothesis is that data from each group of speed and centerline separation ranges for field and simulation model are statistically similar at 5% significance levels. The exercise in Table 5.7 does not reject the null hypothesis for the majority of speed and centerline separation ranges (a total of 82 of 105 cells or 79% of cells in Table 5.7).

Table 5.1 Input parameters in simulation model from field data
Table 5.1 Input parameters in simulation model from field data

Application of the developed model

  • Capacity of the road section
  • Effect of road width on capacity

Homogeneous flows (machines only) with different free flow velocities are generated at different flow rates and their capacities are calculated based on the flow velocity diagrams. Variations in desired car speeds are fixed at 10% of the average value to represent free flow speed. As expected, it is observed that the capacity of the section increases as the free flow speed increases, from 4,940 cars/hour for the desired speed of 50 km/hour to 8,140 cars/hour for the desired speed of 120 km/hour for the 3-lane road, comprising 100% cars.

As expected, it is observed that the capacity for both flows increases as the road width increases. As the road width increases, motorists feel more comfortable and try to maintain more safety.

Fig. 5.16. (a) Speed-density; and (b) Speed-flow relationships of simulated section  5.4.2 Effect of free-flow speed on capacity
Fig. 5.16. (a) Speed-density; and (b) Speed-flow relationships of simulated section 5.4.2 Effect of free-flow speed on capacity

Summary

The developed comprehensive simulation model updates the position of each vehicle (using a parallel updating scheme) in a discretized time step of 1 second, which is assumed to be equal to the drivers' average perception-reaction time. Field parameters such as roadway characteristics, traffic composition, desired lateral and longitudinal speeds are extracted from the calibration field section and used as input data in the simulation model. The values ​​of additional unknown parameters in the simulation model are calculated indirectly so that the error between the field traffic characteristics and the simulated flow characteristics is minimal.

It is observed that the developed model performs similarly to VISSIM while representing macroscopic features and microscopic features such as headway distribution, speed distribution, etc. It is also designed to understand the effect of road width, desired speed and traffic flow segregation on the calculate road capacity.

CONCLUSIONS AND FUTURE SCOPE

Conclusions and future scope

  • General
  • Conclusions
    • Analysis and modeling of longitudinal interactions
    • Overtaking decision-making criteria
    • Developed simulation model based on inter-vehicular gaps
  • Contribution of this research
  • Future scope

A simulation model is developed based on lateral and longitudinal gap-keeping behavior and overtaking decision modeling. The simulation model is also used to separate bus traffic from other vehicles in the stream and to model the two separate streams. The simulation model is validated with field data using genetic algorithms and also compared with the commercial VISSIM software.

Development of Driving Simulators in Mixed Traffic: The developed simulation model can be used to generate a traffic flow with desired flow and traffic composition in a driving simulator, and a driver's performance can be evaluated based on his/her maneuverability in this traffic flow. Development of road ecology models in mixed traffic flow: The road ecology models (i.e. the environmental effect of traffic) could be developed by combining this simulation model with added inputs from emission modeling of pollutants from different vehicle types.

Bibliography

Conference/ Journal publications

In: Proceedings of the 13th COTA International Conference of Transport Professionals (CICTP 2013), Jul 2013, Shenzhen, China. In: Proceedings of the 14th International Symposium on Transport and Traffic Theory, July 1999, Jerusalem, Israel. In: Proceedings of the 2nd Conference of the Indian Transport Research Group (2nd CTRG), Dec 2013, Agra, India.

In: Proceedings of the second international symposium on road traffic flow theory, Paris, France, 44–59. In: Proceedings of the 7th International IEEE Conference on Intelligent Transportation Systems, October 2004, Washington D.C., USA.

Theses

Books/ manuals

Websites

List of Publications

International Journals

International Conferences

Gambar

Fig. 2.3 Classification of traffic data collection methods  2.3.1 Data collection by static devices
Fig. 3.1 Flowchart depicting the overall study framework adopted in research work
Fig. 3.3 Equipments used for measuring lateral clearance- (a) Microcontroller board, (b) Ultrasonic  sensor, (c) Connecting wires and (d) Breadboard with power supply
Fig. 3.4: Illustration showing sensor readings corresponding to time of detection.
+7

Referensi

Dokumen terkait

Chapter 1: Introduction Chapter 2: Background Chapter 3: Representing Text Commentary Data Chapter 4: Mining Strength Rules and Weakness Rules of Cricket Players Chapter 6: