Introduction

Measuring Travel Time

Analysing Travel Time Data

Traffic Congestion Analysis and Modelling

Theoretical Basis for Congestion VOC Modelling in HDM-4

Using ROMDAS to Calibrate the HDM-4 Congestion Model

INTRODUCTION

Travel time is the total time for a vehicle travel over a section of road. Travel time surveys are conducted to:

• identify problem locations on facilities;
• determine the level of service;
• establish the impact of road improvements;
• as input into economic feasibility studies;
• to evaluate the predictions of, or as input to, transportation planning studies.

It is often coupled with delay - usually defined as the time period when a vehicle is stopped - in order to obtain an overall measure of the performance of a facility.

Congestion surveys are similar to travel time surveys except the objective is to establish the level of traffic interactions. These data are particularly useful in economic feasibility studies, such as those using the World Bank HDM-4 Model.

This paper describes conducting travel time and congestion surveys along with the analysis and presentation of the results

MEASURING TRAVEL TIME

The two common methods for conducting travel time surveys are:

• test runs in a vehicle equipped with an instrument such as ROMDAS;
• licence plate observations.

This paper focuses on the use of the ROMDAS.
ROMDAS surveys may be done using different techniques.

Floating Car
The objective of a 'floating car' survey is for the vehicle to travel along the road at the average speed of traffic. This is achieved by ensuring the vehicle passes as many vehicles as pass it.

Average Car
This is a less restrictive survey than the floating car in that the driver does not endeavour to pass as many vehicles as pass it. Instead, the driver travels according to their best judgment as to the traffic stream's speed.

Maximum Car
With this technique the driver travels at the posted speed limit unless impeded by traffic.

Chase Car
The 'chase car' technique sees individual vehicles randomly selected from the traffic stream and followed. This gives a sample of the performance of actual vehicles in the traffic stream.

The travel time survey gives the time and distance travelled. It is advisable to establish check points along the survey route. This enables the data to be analysed for individual sections. These check points are best established using the ROMDAS Location Reference Point survey. This sees the chainage of the check points determined relative to the start point. During the survey the operator resets the chainage at each check point thereby providing a record of the location.

The travel time data can be represented in several ways:

Cumulative Travel Time
For a journey the cumulative travel time from the start to each check point is established. It is often plotted against chainage.

Average Section Speed
The average section speed is defined as the distance between check points divided by the time between check points.

Average Journey Speed
The average journey speed is the cumulative distance divided by the cumulative time.

Stopped Time
The time that the vehicle is stopped is an indicator of the level of service in the 1985 Highway Capacity Manual, although it tends to be established based on intersection surveys as opposed to test vehicle surveys.

It is also useful to record the reason that the vehicle was stopped - queuing, traffic signals, etc.

Traffic congestion leads to a decrease in vehicle speeds and an increase in interactions. These interactions manifest themselves as higher levels of acceleration and deceleration. The ROMDAS system will collect data that will allow for an accurate assessment of traffic congestion.

The discussion which follows is oriented towards the use of congestion in economic appraisals, although the principles apply to other traffic engineering applications as well. It is focused around the forthcoming World Bank 'Highway Development and Management Model' - HDM-4 - which the ROMDAS was specifically designed to collect the input data for. Details of the basic research are found in NDLI (1995) and Greenwood and Bennett (1995).

The speed-flow model adopted for motorised transport in HDM-4 is the "three-zone" model proposed by Hoban, et. al. (1994). This model is illustrated in Figure 1.

Figure 1: HDM-4 Speed-Flow Model

The following notation applies to Figure 1:

• Qo is the flow level below which traffic interactions are negligible in PCSE/h

• Qnom is the nominal capacity of the road in PCSE/h
• Qult is the ultimate capacity of the road for stable flow in PCSE/h
• Snom is the speed at the nominal capacity in km/h
• Sult is the speed at the ultimate capacity in km/h
• S1 to S3 are the free flow speeds of different vehicle types in km/h
• PCSE are passenger car space equivalents (see below)

The model predicts that below a certain volume there are no traffic interactions and all vehicles travel at their free speeds. Once traffic interactions commence, the speeds of the individual vehicles decrease until the nominal capacity where all vehicles will be travelling at the speed of the slowest vehicle class. The speeds can then further decrease towards the ultimate capacity beyond which unstable flow will arise.

The HDM-4 speed-flow model departs from traditional speed flow models through its use of passenger car space equivalents (PCSE) instead of passenger car units (PCU) or veh/h. The PCSE reflect the differences in space occupied by each vehicle type on the road, which directly affects road capacity (Hoban, et al., 1994). The PCSE are used because the model already considers the differences in speeds by vehicle type which, along with space, are implicitly considered by PCU factors.

Hoban, et al. (1994) indicate that the key parameters for use in the model vary depending upon the road type and width. Table 1 lists the recommended values for these parameters. The values for Q0 and Qnom are expressed relative to Qult. Hoban, et. al. (1994) suggest that the nominal speed is equal to 85 per cent of the free speed of the slowest vehicle.

Table 1
Speed-Flow Model Parameters by Road Type

THEORETICAL BASIS FOR CONGESTION VOC MODELLING IN HDM-4

The HDM-4 speed-flow model presented in Figure 1 shows that as flows increase, there is an increase in vehicle interactions and a decrease in speeds. These interactions are accompanied by an increase in the frequency and magnitude of the vehicle accelerations and decelerations. The HDM-4 congestion model is built around this theory.

The magnitude of vehicle interactions are to be represented in HDM-4 by the acceleration noise, i.e. the standard deviation of acceleration. The total acceleration noise for a vehicle can be considered to be comprised of the following two components:

a =

where a is the total acceleration noise in m/s²

at is the noise due to fast-moving vehicle interactions in m/s²

an is the natural noise ascribed to the driver and road in m/s²

Figure 2 illustrates the different acceleration distributions which arise with congested and uncongested conditions.

Figure 2: Congested and Uncongested Acceleration Distributions

The natural noise can be described as:
an = f(adr, aal, asf, anmt, airi)
where adr is the noise due to natural variations in the driver's speed in m/s²
aal is the noise due to the road alignment in m/s²
asf is the noise due to side friction in m/s²
anmt is the noise due to non-motorised transport in m/s²
airi is the noise due to roughness in m/s²
In HDM-4 traffic noise is considered to be due to motorised transport. On the basis of previous research and experiments conducted by the ISOHDM HTRS team in Malaysia, the following equation was developed which gives the traffic noise as a function of relative flow - namely the volume to capacity ratio (NDLI, 1995):
at = atmax

RELFLOW =

where RELFLOW is the relative flow (i.e. volume to capacity ratio)

a0 and a1 are regression coefficients quantified as:

a0 = 4.2 + 23.5

a1 = -7.3 - 24.1

Figure 3 illustrates the predictions for various relative flows.

 
Figure 3: Effect of Relative Flow on Traffic Noise

In HDM-4 the driver noise (adr) and the alignment noise (aal) are combined into a single value as it is difficult to differentiate between these two components. The other three components of natural noise - side friction (asf), non-motorised transport (anmt) and roughness (airi) - will be modelled as linear functions such as those shown in Figure 4. The maximum values preliminarily estimated for these components are:
asf = 0.20 m/s²

anmt = 0.40 m/s²

airi = 0.30 m/s²

Figure 4: Proposed HDM-4 Natural Noise Models

These maximums apply at side friction and non-motorised transport ratios of 1.0 and a roughness of 20 IRI m/km.

The total natural noise is given by:

an =

Using the above equations, the total acceleration noise at any relative flow can be characterised by the natural noise (an) and the maximum traffic noise (atmax).

The maximum traffic noise is calculated as:

atmax =

Experiments were conducted in Malaysia with cars, medium trucks and buses led to a value of 0.60 m/s² being recommended for the maximum total acceleration noise (amax) and a minimum value of 0.10 m/s² for the natural noise. These were adopted as the default parameters for all vehicle classes.

ROMDAS has been specifically designed to collect the necessary data to calibrate the HDM-4 congestion model.

The calibration exercise should focus on the two extreme values of the traffic noise: the natural noise and the maximum traffic noise. These values can be established by conducting surveys on high standard roads with little traffic and on the same roads under highly congested conditions. By eliminating stopped time the noise can be directly calculated from the ROMDAS output.

In the same manner the noise can be established by conducting surveys on roads with varying alignments, surface or traffic conditions. These will allow for the other components of the HDM-4 model to be quantified.