LEABHARLANN CHOLAISTE NA TRIONOIDE, BAILE ATHA CLIATH TRINITY COLLEGE LIBRARY DUBLIN
OUscoil Atha Cliath The University of Dublin
Terms and Conditions of Use of Digitised Theses from Trinity College Library Dublin Copyright statement
All material supplied by Trinity College Library is protected by copyright (under the Copyright and Related Rights Act, 2000 as amended) and other relevant Intellectual Property Rights. By accessing and using a Digitised Thesis from Trinity College Library you acknowledge that all Intellectual Property Rights in any Works supplied are the sole and exclusive property of the copyright and/or other I PR holder. Specific copyright holders may not be explicitly identified. Use of materials from other sources within a thesis should not be construed as a claim over them.
A non-exclusive, non-transferable licence is hereby granted to those using or reproducing, in whole or in part, the material for valid purposes, providing the copyright owners are acknowledged using the normal conventions. Where specific permission to use material is required, this is identified and such permission must be sought from the copyright holder or agency cited.
Liability statement
By using a Digitised Thesis, I accept that Trinity College Dublin bears no legal responsibility for the accuracy, legality or comprehensiveness of materials contained within the thesis, and that Trinity College Dublin accepts no liability for indirect, consequential, or incidental, damages or losses arising from use of the thesis for whatever reason. Information located in a thesis may be subject to specific use constraints, details of which may not be explicitly described. It is the responsibility of potential and actual users to be aware of such constraints and to abide by them. By making use of material from a digitised thesis, you accept these copyright and disclaimer provisions. Where it is brought to the attention of Trinity College Library that there may be a breach of copyright or other restraint, it is the policy to withdraw or take down access to a thesis while the issue is being resolved.
Access Agreement
By using a Digitised Thesis from Trinity College Library you are bound by the following Terms & Conditions. Please read them carefully.
A New Approach To Driving Cycle
Development And Its Applications
by
TIMOTHY PATTISON
A dissertation submitted to Trinity College Dublin, in partial fulfilment
of the requirements for the degree of Doctor of Philosophy
Department of Civil, Structural and Environmental Engineering
Declaration
I declare that this thesis has not been submitted as an exercise for a degree at
this or any other university and it is entirely my own work.
I agree to deposit this thesis in the University’s open access institutional repository
or allow the library to do so on my behalf, subject to Irish Copyright Legislation and
Trinity College Library conditions of use and acknowledgement.
_ Timothy Pattison
Abstract
Driving cycles describe the general driving pattern of a vehicle in a road network,
and are usually represented by speed-time profiles of equal spacing. They are used
to approximate and represent driving conditions as an input for laboratory chassis
dynamometers, so that fuel consumption, exhaust emissions and emission coefficients
can be evaluated. Traditional methods of data collection for use in driving cycle
development involve attaching expensive data loggers to vehicles and often hiring
professional drivers to sit in traffic, sometimes for hours at a time. This can be very
costly and often needs to be repeated over the course of several months. The work
in this thesis proposes to use microsimulation as an alternate method which brings
many benehts such as the ability to generate much larger data sets at a significantly
faster than real-time rate. It also allows for the construction of unique and controlled
networks that would not be accessible in the real-world, as well as opportunities to
predict the effects on the driving cycle of a new traffic control strategy that might be
considered for implementation.
Another fundamental part of this research was the development of a unique au
tomated cycle building algorithm, which does not require any decision making input
from the user, and even negates the need for any user knowledge of the subject. Ben
efits to this are, the potential for much faster cycle creation, the ability to be used by
anyone, and better suitability for cycle comparison due to the identical methodology
and consistency of results.
A detailed study of Dublin’s South Quays was used to validate the concepts that
had been proposed. A synthetic driving cycle was developed using data extracted
from a microsimulation model, and was compared against a real-world cycle that was
based on data from GPS loggers. The same automated cycle generation algorithm
was used and the cycle ontputs were found to match up reasonably well.
at certain average speeds, and applying a correction factor based on the acceleration
mode of the vehicle. It places a heavy emphasis on accounting for the dynamics of
a vehicle’s acceleration and deceleration over the total cycle duration, unlike many
other models that just focus on average speed.
Acknowledgements
I would like to express my gratitude to my supervisor Prof. Biswajit Basu for all his
guidance and expertise.
I would like to thank Tigran Tchrakian for his help while we shared an office together.
1 would also like to thank my brother Roger for having the patience to be the dedi
cated professional driver for testing, and Dermot Kinane who provided me with access
to the SCATS database in the Dublin City Council.
Finally, I would like to thank my whole family for their support.
Contents
Declaration
ii
Abstract
iii
Acknowledgements
v
List of Tables
xi
List of Figures
xiii
1 Introduction &: Literature Review
1
1.1 Introduction... 1
1.2 Literature Review... 3
1.2.1
Driving Cycles ... 3
1.2.1.1 Driving Cycle Classification... 3
1.2.1.2 Methodology... 6
1.2.1.3 Data Acquisition... 10
1.2.2
Microsimulation... 12
1.2.2.1 Applications in Transportation... 13
1.2.2.2 Overview of Available Models... 14
1.2.2.3 VISSIM... 17
1.2.3
Urban Traffic Control Systems (UTC) ... 18
1.2.3.2
SCATS ... 19
1.2.3.3
Traffic Detectors... 20
2 Algorithm for Cycle Generation
23
2.1 Introduction... 23
2.2 Average Speed... 28
2.3 Average Distance... 29
2.4 Speed Acceleration Frequency Distribution Matrix... 31
2.5 Microtrip Generation... 40
2.6 Microtrip/Segment Filter... 44
2.7 Microtrip Duration... 46
2.8 Microtrip Average Speed... 47
2.9 Microtrip Composition... 48
2.10 Microtrip Comparison ... 49
2.11 Final Sorting ... 57
2.12 Cycle Generation... 61
2.13 Conclusions... 82
3 A New Approach To Driving Cycle Development
83
3.1 Introduction... 83
3.2 Methodology ... 83
3.2.1
Description of Network... 83
3.2.2
Recording Data... 86
3.2.3
Comparison of test networks... 87
3.2.4
Driving cycle generation... 88
3.3 Results... 90
3.3.1
Evaluation of Final Cycles... 92
4 Validation of the Method: A Case Study of Dublin’s South Quays 95
4.1 Introduction... 95
4.2 Real-World Cycle... 96
4.2.1
Description of Network... 96
4.2.2
Data Collection... 98
4.2.3
Results... 102
4.3 Synthetic Cycle... 104
4.3.1
Description of Network... 104
4.3.1.1
Junctions and Signals ... 107
4.3.2
Recording Data... 115
4.3.3
Results... 115
4.4 Comparison of Synthetic Cycle to Real-World Cycle... 117
4.5 Discussion of Research Significance... 118
4.6 Conclusions... 120
5 Integration of an Emission Model
122
5.1 Introduction... 122
5.2 Literature Review of Emission Models ... 123
5.2.1
Average Speed-Based Models ... 123
5.2.2
Dynamic Emission Models... 125
5.2.2.1
Emission Maps... 126
5.2.2.2
Regression-Based Models ... 127
5.2.2.3
Load-Based Models... 128
5.2.2.4
Traffic Situation Models... 129
5.3 Theory... 132
5.3.1
Introduction... 132
5.3.2
Constant Velocity Mode... 134
5.3.3 Acceleration Mode... 135
5.3.5 Final Form ... 137
5.4 Implementation of model... 138
5.5 Results... 147
5.6 Further Analysis ... 150
5.6.1
Theory... 150
5.6.2
Results... 150
5.7 Calibrated Testing ... 151
5.8 Conclusions... 155
6
Conclusions
157
References
162
Appendices
168
A Pertains To Chapter 2
168
A.l Average Speed Function... 169
A.2 Average Distance... 170
A.3 Speed Acceleration Frequency Distribution Alatrix Function... 171
A.4 Microtrip Generation Function... 177
A.5 Microtrip/Segment Filter Function... 180
A.6 Microtrip Duration Function... 181
A.7 Microtrip Average Speed Function... 182
A.8 Microtrip Composition Function... 183
A.9 Microtrip Comparison Function... 185
A. 10 Final Sorting Function... 190
A. 11 Cycle Creation Function... 192
B.2 Uncongested Network Data ... 212
B.3 Congested Network Data... 217
B. 4 Signal Change Network Data ... 222
C Pertains To Chapter 4
227
C. l Off-Peak Network Data... 228
C.2 Peak Network Data... 233
C. 3 Synthetic Off-Peak Network Data... 238
D Pertains To Chapter 5
243
List of Tables
1.1 Classification of models based on applicable traffic conditions (Algers
et al., 1998)... 15
2.1 Description of function variables 1/2 ... 26
2.2 Description of function variables 2/2 ... 27
3.1 Uncongested vehicle inputs ... 87
3.2 Congested vehicle inputs... 88
3.3 Comparison of uncongested final cycle to base data... 92
3.4 Comparison of congested final cycle to base data... 93
3.5 Comparison of signal change final cycle to base data... 93
4.1 Data collection times for off-peak cycle... 99
4.2 Data collection times for peak cycle... 100
4.3 Evaluation of off-peak final cycle ... 103
4.4 Evaluation of peak final cycle... 104
4.5 Off-peak vehicle inputs... 106
4.6 Evaluation of synthetic off-peak hnal cycle... 116
4.7 Comparison of real cycle to synthetic cycle... 117
5.1 Off-peak synthetic cycle emissions... 147
5.2 Off-peak cycle emissions... 148
5.4 Off-peak cycle emissions... 152
5.5 Calibrated off-peak synthetic cycle emissions... 152
5.6 Synthetic cycle emissions deviation from real cycle... 153
List of Figures
1.1 New European Driving Cycle ... 5
2.1 Flow chart showing program inputs and outputs... 25
2.2 SAFD matrix... 32
2.3 Normalised SAFD matrix ... 32
2.4 Microtrip data sample... 45
3.1 Network diagram showing link numbers... 84
3.2 Default signal timings ... 80
3.3 Final Driving cycles... 90
4.1 Satellite view of Dublin Quays... 97
4.2 Sample from one of the recorded videos... 101
4.3 Final driving cycles... 102
4.4 Map showing route ; Network diagram... 105
4.5 Example of VISSIM 3D view [Junction TCS 37]... 107
4.6 Junction layout for TCS625 ... 108
4.7 TCS 625 signals... 108
4.8 Junction layout for TCS26... 109
4.9 TCS 26 signals ... 109
4.10 Junction layout for TCS193... 110
4.13 TCS 37 signals ... Ill
4.14 Junction layout for TCS152... 112
4.15 TCS 152 signals... 112
4.16 Junction layont for TCS760 ... 113
4.17 TCS 760 signals... 113
4.18 Junction layont for TCS190... 114
4.19 TCS 190 signals... 114
4.20 Synthetic off-peak driving cycle... 115
4.21 Comparison of real cycle to synthetic cycle... 117
5.1 CO emission rates vs. average speed ... 133
5.2 NOx emission rates vs. average speed... 133
5.3 HC emission rates vs. average speed ... 134
5.4 PM emission rates vs. average speed ... 134
5.5 Comparison of calibrated synthetic cycle to original synthetic cycle . 151
B.l
Uncongested Network Data Sample 1/5... 212
B.2
Uncongested Network Data Sample 2/5... 213
B.3
Uncongested Network Data Sample 3/5... 214
B.4
Uncongested Network Data Sample 4/5... 215
B.5
Uncongested Network Data Sample 5/5... 216
B.6 Congested Network Data Sample 1/5... 217
B.7 Congested Network Data Sample 2/5... 218
B.8 Congested Network Data Sample 3/5... 219
B.9 Congested Network Data Sample 4/5... 220
B.IO Congested Network Data Sample 5/5... 221
B.ll Signal Change Network Data Sample 1/5... 222
B.12 Signal Change Network Data Sample 2/5... 223
B.14 Signal Change Network Data Sample 4/5... 225
B. 15 Signal Change Network Data Sample 5/5... 226
C. l
Off-Peak Network Data Sample 1/5... 228
C.2
Off-Peak Network Data Sample 2/5... 229
C.3
Off-Peak Network Data Sample 3/5... 230
C.4
Off-Peak Network Data Sample 4/5... 231
C.5
Off-Peak Network Data Sample 5/5... 232
C.6 Peak Network Data Sample 1/5... 233
C.7 Peak Network Data Sample 2/5... 234
C.8 Peak Network Data Sample 3/5... 235
C.9 Peak Network Data Sample 4/5... 236
C. 10 Peak Network Data Sample 5/5... 237
C.ll Synthetic Off-Peak Network Data Sample 1/5... 238
C.12 Synthetic Off-Peak Network Data Sample 2/5... 239
C.13 Synthetic Off-Peak Network Data Sample 3/5... 240
C.14 Synthetic Off-Peak Network Data Sample 4/5... 241
Chapter 1
Introduction
&i.
Literature Review
1.1 Introduction
Vehicular exhaust emissions are a major source of pollution globally, and are the
predominant source of pollution in modern developed cities. The main air pollutants
being emitted from road traffic are carbon monoxide (CO), nitrogen oxides (NO^),
hydrocarbons (HCs), particulates (PM), sulphur dioxide (SO2) and carbon dioxide
(CO2) (Goyal and Krishna, 1998). In Ireland alone, 112kt of NOx was released by
the transport sector in 2005, which was 40 percent of the national total (Kelly et ah,
2009). In the same year, Irish transport accounted for 19 percent of CO2 emissions.
Vehicle emissions are affected by driving patterns on the road, which in turn are
mainly dependent on traffic conditions (Tzirakis et ah, 2006). Driving cycles are used
to approximate and represent driving conditions as an input for laboratory chassis
dynamometers so that fuel consumption, exhaust emissions and emission coefficients
can be evafuated (Simanaitis, 1977).
situation. Driving cycles have an extensive range of applications in both commercial
and governmental fields. They are required by traffic engineers to aid in the design
of traffic control systems and the simulation of traffic flows and delays (Tamsanya
et ah, 2006). Environmentalists are interested in evaluating how a vehicle performs
regarding pollutant emissions while engaged in specific driving patterns. As fuel
consumption is also directly related to driving patterns, a speed-time profile can be
used to estimate fuel consumption for a particular urban area.
Traditional methods of data collection for use in driving cycle construction involve
attaching expensive data loggers to either a whole fleet of vehicles, or perhaps only one
if the car chasing technique is being employed. Additional costs are also incurred with
fuel and if professional drivers need to be hired. These methods are also extremely
time consuming as the drivers must sit in traffic while data is being recorded, and this
is further exacerbated if data is needed during peak hour flows. In order to collect
sufficient data such that the sampling is representative of real driving conditions,
sometimes months worth of data needs to be collected.
As described by Merz (1991), simulation becomes attractive when conventional
analytic, numeric or physical experimental methods would be too time-consuming, ex
pensive, difficult, hazardous, irreversible or even impossible as real world experiments
intended to solve a problem.
planners want to know how it will affect the driving cycle in that locale before it
is implemented, then microsimulation is really the only option. Generally speaking
simulation typically allows for the collection of signihcantly greater data sets too.
There are also many cost benehts to this approach as data loggers do not need to be
purchased, nor drivers hired. Finally, it is a more environmentally friendly approach
as huge quantities of expensive fuel do not need to be burned while vehicles are
collecting data.
The work presented in this thesis aims to demonstrate the use of microsimulation
in driving cycle development, propose a new automated method for cycle generation,
investigate how altering signal timings affects driving cycles, and also to develop an
emissions model that can estimate emissions based on the input of the driving cycle.
1.2 Literature Review
1.2.1
Driving Cycles
1.2.1.1
Driving Cycle Classification
Driving cycles can be classified as legislative and non-legislative depending on how
they are being used (Tong and Hung, 2010). Non legislative cycles are mainly used in
the field of research and are developed primarily for pollution evaluation and energy
conservation. Examples include the Improved European Cycle (lEC), the Sydney
cycle, the Melbourne peak cycle, the Perth cycle and the Hong Kong driving cycle.
Japanese 10-15 cycle. These particular cycles are also examples of legislative cycles
that are not only used to control vehicle emissions in the geographical location that
they were designed for, but also in numerous developing countries that do not have
their own cycle (Tong and Hung, 2010).
Driving cycles are often categorised into two broad types, the first being transient
driving cycles, which are developed from on-road driving data. These real world cy
cles are more dynamic than modal cycles, which is reflected in more rapid acceleration
and deceleration profiles experienced during on road conditions (Tzirakis et ah, 2006).
The second being the “modal” or “polygonal” driving cycles, which are composed of a
sequence of steady-state phases (Tong and Hung, 2010). After the cycles’ operational
modes are measured, they are smoothed into phases of steady speed, acceleration and
deceleration (Esteves-Booth et ah, 2002). Examples include the ECE 15 European
cycle and NEDC, US FTP cycles, Melbourne peak cycle and Japanese cycles (Tong
and Hung, 2010). Previous research has indicated that “modal” cycles do not accu
rately describe real-world driving behaviour due to the statistical smoothing effect
of the vehicles’ operational modes, which tends towards underestimating emissions
when used for testing (Pelkmans and Debal, 2006; Tzeng and Chen, 1998).
New European Driving Cycle(NEDC)
in g/km for each pollutant. The permissive limit values for various emissions have
become vastly more stringent over time as governments are eager to create a health
ier environment. Under Euro 4 regulations, which came into effect in 2005, limits for
Particulate Matter (PM), Carbon Monoxide (CO), Nitrogen Oxides (NO^) and Hy
drocarbons (HC) are roughly 5-10 times lower than those before the Euro regulations
came into effect in 1990 (Pelkmans and Debal, 2006). The Euro 4 regulations have
been superseded by Euro 5 since the start of 2011, imposing even more stringent lim
its on vehicles. The NEDC is openly critised for not correlating to real-world driving
patterns due to its very smooth acceleration profile, which translates to a low relative
positive acceleration value (Andre and Pronello, 1997). It greatly underestimates the
vehicle load compared to real conditions, and the operating range of the engine is
greatly reduced. Thus engine manufacturers need only to optimise emissions in these
particular zones in order to pass the certification tests. In studies undertaken by
Pelkmans and Debal (2006), vehicles that comply with Euro 4 standards can exhibit
CO and NO^ emissions that are as much as 10 times greater under real traffic con
ditions compared to the NEDC test bed. Also fuel consumption and CO2 emissions
could be underestimated by 10-20%.
New European Drive Cycle (NEDC)
uo 120 100 60 60 40 20
Part On# UThar
A... 4..a i
■t
bh iilMi
a
Part Two E >-tra-Ort»an
0 100 :oo 300 400 500 m *00 sco 000 1000 uoo 1:00 Txm« (%ec4)
1.2.1.2 Methodology
In general, cycle construction methods typically include four steps: collecting real
world driving data, segmenting the driving data, constructing cycles, and evaluat
ing and selecting the final cycle. Depending on the type of driving activity that is
being used to construct the cycle, existing cycle construction methodologies for light-
duty vehicles can be generalised into four types: micro-trip cycle construction, trip
segment-based cycle construction, cycle construction based on pattern classification,
and modal cycle construction.
Micro-trip based cycle construction
One of the most common methods for creating driving cycles is to chain “micro
trips,” which are defined as the driving activity between adjacent stops, including
the leading period of idle. Examples include the Sydney Cycle, the Melbourne Peak
Cycle, the Unified Cycles, the Unified Correction Cycles, and the Hong Kong driving
cycles.
In this method, real world activity data, for example, the modal activity of a
number of vehicles with each being tracked for a period of time while they are driv
ing on a freeway or an arterial, is separated into micro-trips with the aim that the
cycle closely matches that of the observed data. Micro-trips are usually chosen using
one of three methods: random selection; “best incremental,” meaning incrementally
searching for and adding a micro-trip with specific modal characteristics; or a hybrid
of both approaches (Austin et ah, 1993). Candidate cycles are assessed using param
eters that typically include maximum speed, minimum speed, average speed, average
acceleration, average deceleration and the Speed-Acceleration Frequency Distribution
(SAFD).
particular driving condition can be difficult (Andre, 2004). For example, in smooth
traffic conditions, a vehicle rarely stops and a single micro-trip may cover different
road segments or different traffic conditions. Use of micro-trip based methods has
been limited to developing cycles designed to represent a single type of trip or cycles
designed to replicate region-wide driving conditions.
Segment-based cycle construction
A trip “segment” is obtained by segregating vehicle speed-time profiles using
changes in roadway type or LOS, in addition to stops (Carlson and Austin, 1997).
Therefore, overall vehicle driving activity can be arranged into sequences based on
roadway type or LOS, and cycles can be built to represent driving activity for specific
roadway types and traffic conditions. One major application of the segment-based
cycle construction method is US Environmental Protection Agency (EPA) facility-
specihc speed correction cycles.
As with the micro-trip based method, trip segments are selected and linked to
gether in series using a hybrid of random and best-incremental logic. The differen
tiating element compared to micro-trips is that trip segments can start and end at
any speed. Accordingly, the chaining of segments needs certain constraints on speed
and acceleration between two connecting segments, independently of previous and
succeeding micro-trips. For example, in constructing EPA’s facility specihc cycles,
the difference in speed between two connecting segments was required to be within
0.5 mph, and the difference in acceleration within 0.5 mph/sec. The “best” cycle is
then selected using two main parameters: 1) the sum of difference in SAFD between
the test cycle and the target population, and 2) the amount of operation occuring in
high power mode (Carlson and Austin, 1997).
Cycle construction with pattern classification
This kind of method has been vastly applied in generating European driving cycles
(Andre et ah, 1995; Andre, 1996). In this approach, “kinematic sequences” (which
are similar to micro-trips) are categorised into heterogeneous classes using statistical
methods. Also, the approach uses succession probabilities to approximate and con
sider the likelihood that one class of activity precedes or follows another activity class.
Driving cycles are generated by re-connecting kinematic sequences randomly picked
from each of the activity classes in compliance with the probability and chronology
of kinematic sequences (Andre et ah, 1995).
As described by Andre et al. (1995) while constructing the European Urban Cy
cles, kinematic sequences are characterised by 20 variables including duration, idle
time and distance throughout the sequence; the means, maximums, standard devi
ations, 20% and 80% percentiles of driving modes; and other parameters such as
instantaneous speed and acceleration, and the distance between two accelerations.
Following the application of Principal Component Analysis (PCA) to these 20 vari
ables, cluster analysis was performed on the kinematic sequences. Four distinctive
classes were identified, representing congested and free-flow urban traffic, extra-urban
and motorway driving conditions.
Following this, a trip was considered as a series of kinematic sequences. The
ascertained trips were categorised according to the frequency of sequences in each
kinematic sequence class and also the number of transitions between two classes of
kinematic sequences. Rural road, urban, and motorway trips were identified as the
three primary types of trip.
used to classify these sequences into twelve driving conditions. Urban, rural road
and motorway cycles were created according to the observed composition of driving
conditions.
This method also has its limitations, namely: 1) the basic units for cycle devel
opment, including micro-trip and uniform sequences, are not immediately related to
emissions, and are possibly not the best units to be used in defining emission-related
driving activity, and 2) the classification of sequences is based on the chi-square dis
tance of speed-acceleration joint distribution. Even though such classification differ
entiated the kinematic driving activity, it does not necessarily differentiate emissions
related to these activities.
Modal cycle construction
Lin and Niemeier (2002) describes the application of a mode-based cycle construc
tion method, where real world driving is considered as a sequence of acceleration,
deceleration, cruise, or idle modes. It is known that running emissions relate to a
vehicle’s modal operation as well as average speed, hence, when estimating emis
sions, logic suggests that driving activities should be analysed and replicated from
a modal perspective. Assuming that the likelihood of an individual modal event
(eg.acceleration, cruise, or deceleration) occurring depends exclusively on the mode
of the previous modal event, driving activities can be modelled as a Markov Chain.
modal nature of the current snippet and the transition matrix. One snippet is picked
from the predicted bin without replacement. The snippet selection requires that
the selected snippet best improves the match to the observed SAFD, and that the
initial speed of the snippets matches the end speed of the preceeding modal snippets
with an acceptable difference. Sirippet selection is repeated until the desired cycle
length is obtained. The final driviirg cycles are selected using a composite assessment
measure, which integrates parameters such as differences in average speed, differences
in minimum and maximum speeds, and percentage of idling operation.
As with all the methods this too has its limitations, these include: 1) the criterion
used to connect the snippets is arbitrary, and 2) cycles are constructed for a specific
facility and LOS, and the number of cycles required to represent the emission related
driving activity are not well researched.
1.2.1.3 Data Acquisition
Basic data that is acquired for developing a driving cycle is aimed at describing
traffic conditions for a defined vehicle group (light duty vehicle, heavy duty vehicle,
bus, motorbike etc.) and is often limited to a small number of models and vehicles
that is representative of the studied fleet. Constraints are often put on the data
collected such as vehicle use types, trips to work during the morning period, the peak
hour period, urban driving conditions, road types or typical trips for a particular road
network.
When collecting on-road vehicle data for use in developing a driving cycle there
are typically four main methods that are used as described by Andre (1996):
the technical details of the study, their knowledge can heavily influence their
driving behaviour while recording data.
The instrumentation of vehicles which are driven by a sample of private drivers.
This method is better at taking into account the variability of individual driving
patterns.
The car chase technique can be used to measure the speed of a large quantity
of target vehicles, without a high risk of disturbing the drivers’ behaviour. The
method involves randomly selecting a vehicle while driving around in traffic, and
a survey vehicle then follows the car while keeping approximately a constant
distance during different modes of operation to ensure that its speed matches
that of the target vehicle. The distance between the two vehicles can be ap
proximated by eye or a more accurate way is to use a laser distance meter. The
problems with this method are that it does not allow for describing trip start
and end conditions, nor does it allow for recording any other vehicle parameters
of the target vehicle aside from speed (eg. gearbox ratio).
The instrumentation of privately owned vehicles being driven by their usual
drivers for their usual duties. The prime advantage of this method is that the
vehicles are driven under normal use conditions. It also allows for the test
vehicles to be instrumented with data loggers to record other parameters aside
from speed (temperatures etc.), but it should be noted that this method does
not provide any information about traffic densities or road conditions.
derived from the calculated speed and not directly measured) and more infrequently
parameters such as intake air pressure (engine load indicator), road gradient, engine
torque, and traffic conditions (Andre, 1996).
1.2.2 Microsimulation
DefimUon
1.2.2.1 Applications in Transportation
Traffic simulation techniques have been used since the early days of the development
of traffic theory. Traffic microsimulation involves the process of developing a virtual
computer model of a city’s transportation infrastructure so that the interactions of
road traffic (and other modes of transportation) can be simulated in microscopic de
tail. Every vehicle, bus, train, tram, cyclist, pedestrian etc. in the model is considered
as a unique entity with its own goals and behavioural characteristics; each having the
ability to interact with other entities within the model.
Traffic microsimulation computer models replicate the interactions and movement
of real world road traffic using a variety of complex algorithms describing car following,
lane changing, gap acceptance, and spatial collision detection. Furthermore, free form
pedestrian movement is simulated using agent based spatially aware models allowing
road traffic to interact with pedestrians just as they would in the real world.
Tr affic microsimulation models are rapidly becoming the de facto standard for the
evaluation and development of road traffic management and control systems world
wide, especially due to the increasing levels of system complexity in the operation of
urban traffic networks.
Traditional models provide a simplified aggregated description of traffic, usually
expressed in terms of total vehicle flows per hour. In such models, all vehicles of
a certain group obey the same rules of behaviour. This does not allow for some
of today’s most common transport planning/traffic engineering applications to be
modelled with accuracy. Conversely, microsimulation models provide a much better
representation of actual driver behaviour and network performance. They are the
only modelling tools available with the capability to examine certain complex traffic
problems, i.e. intelligent transportation systems, complex junctions, shockwaves,
effects of incidents.
traffic operations, is only possible with microscopic simulation. Microsimulation can
be used to develop new systems and control measures and also optimise their effec
tiveness. They can estimate the effects of a new scheme by producing outputs on a
vast range of measures of effectiveness. These effects can be difficult to quantify in
the field, for example the amount of pollution emissions.
Microsimulation is particularly well matched to the development, testing and eval
uation of intelligent transportation systems (ITS). Many such systems interact with
individual vehicles. Responsive signal control, public transport priority and ramp
metering systems react to vehicles approaching junctions. Dynamic Route Guidance
systems supply specific information to individually equipped vehicles, and Intelligent
Cruise Control systems adjust the speeds of equipped vehicles. With this in mind to
explore the potential benefits of using ITS it is logical to use an assessment tool that
is capable of modelling interactions at the level of individual vehicles.
1.2.2.2 Overview of Available Models
Microsimulation models are primarily used as research tools. Nine of the most em
inent software models used in the transportation sector are commercial products,
listed as follows; AIMSUN2, FLEXSYT II, HUTSIM, INTEGRATION, PARAM-
ICS, THOREAU, VISSIM, ERESIM, NETSIM. Note that CORSIM is a software
package that integrates both NETSIM (for surface street simulation) and ERESIM
(for freeway and highway simulation). There are three other models that are obtain
able through request, and their use is subject to user agreement restrictions (MIXIC,
NEMIS and PHAROS) (Algers et ah, 1998).
Urban
Motorway
Combined
Other
CASIMIR
AUTOBAHN
AIMSUN2
ANATOLL
DRACULA
FREEVU
CORSIM
PHAROS
HUTSIM
FRESIM
FLEXSYT II
SHIVA
MICSTRAN
MIXIC
INTEGRATION
SIMDAC
NEMIS
SISTM
MELROSE
NETSIM
MICROSIM
PADSIM
MITSIM
SIGSIM
PARAMICS
SIMNET
PLANSIM-T
SITRA-B
TRANSIMS
SITRAS
VISSIM
THOREAU
Table 1.1: Classification of models based on applicable traffic conditions (Algers et al.,
1998)
urban, motorway and combined columns are capable of addressing such objectives.
Microsimulation is used for evaluation purposes prior to or in conjunction with on
street operation. This covers a vast array of objectives such as the study of dynamic
traffic control, incident management schemes, real-time route guidance strategies,
adaptive intersection signal controls, ramp and mainline metering, toll plazas and
lane control systems (lane use signs, electronic toll collection, high occupancy vehicle
lane etc.). In addition some models attempt to assess the effects and sensitivity of
alternative design parameters (number of lanes, length of ramps, road curvatures and
grades, and lane change regulations). It is important to note that not all of the afore
mentioned models are designed for all these tasks as they all have different properties
and capabilities.
objectives. For example the modelling of the tactical level of driving and the testing
of intelligent vehicle algorithms (to aid with the writing of Artificial Intelligence pro
grams that could potentially drive vehicles in traffic), to provide a detailed roadway
environment for a simulated robot driving vehicle, to evaluate the safety and comfort
conditions of a line of cars on a single lane, or to predict queues at toll booths.
Interface
The interface used by microsimulation models is comprised of two parts. The
input part is the simulation conhguration, which includes network description, and
the output part is the simulation result.
The majority of the models accept inputs by text hies. These input hies de
scribe the network conhguration with respect to nodes, links, traffic signals, paths,
vehicle arrival rates, link capacities, incidents, signal timing etc. and specify general
parameters of the simulation. Notably hve of the models, AIMSUN2, MELROSE,
PARAMICS, TRANSIMS and VISSIM contain a network Computer Aided Design
Graphical User Interface (GUI) to input road network topology and geometry data
(Algers et ah, 1998). CORSIM and FREEVU only provide tools to graphically create
the input data hies.
1.2.2.3 VISSIM
VISSIM, which is developed by PTV AG, is the microsimulation software package
used for the work described in this thesis. As described in PTV (2010), VISSIM is
a microscopic, time step and behaviour-based simulation model developed to model
urban traffic and public transport operations and flows of pedestrians.
Below is a list of some of VISSIM’s prior applications in transportation:
• Development, evaluation and fine-tuning of signal priority logic. VISSIM can
use numerous types of signal control logic such as fixed-time and vehicle-actuated
controls. There is also an external signal state generator (VAP) whereby user-
defined signal control logic can be designed. Almost any signal control type
can be modelled providing that the controller details are available, or in the
case of adaptive systems such as SCOOT and SCATS if there is direct interface
available with VISSIM.
• Evaluation and optimisation of traffic operations in a combined network of
coordinated and actuated traffic signals.
• Feasibility and traffic impact studies of integrating light rail into urban street
networks.
• Analysis of slow speed weaving and merging areas.
• Capacity and operations analyses of complex station layouts for light rail and
bus systems.
• Comparison of design alternatives including signalised and stop sign controlled
intersections, roundabouts and grade separated interchanges.
• VISSIM has a built-in Dynamic Assignment model and so it is possible to study
route choice dependent scenarios such as the effects of variable message signs.
• Modelling and simulation of pedestrian flows (in streets and buildings). VISSIM
can also simulate the interactions between road traffic and pedestrians.
The accuracy of a traffic simulation model depends largely on the quality of the
vehicle modelling. Unlike some of the less complex models which use constant speeds
and deterministic car following logic, VISSIM uses the psycho-physical driver be
haviour model which was developed by Wiedemann in 1974. Fundamentally in this
model the driver of a quicker vehicle begins to decelerate as an individual perception
threshold to a slower moving vehicle is met. Since it cannot exactly be determined
what speed the vehicle in front is travelling at, the driver’s speed will drop below that
of the vehicle in front until another perception threshold is reached in which case
the driver starts to accelerate again. Thus the whole process results in an iterative
sequence of acceleration and deceleration (PTV, 2010).
VISSIM simulates the flow of traffic by moving “driver-vehicle-units” through
the network. Each driver has specific behaviour characteristics and is assigned to
a specific vehicle, and resulting from this, the driving behaviour corresponds to the
technical capabilities of the vehicle.
1.2.3 Urban Traffic Control Systems (UTC)
1.2.3.1 Adaptive Control Systems
adaptive UTCs and these function by unifying the two philosophies of fixed-time UTC
and vehicle actuation. The principle of fixed-time UTC adheres to the idea that where
traffic signals are in close physical proximity, good vehicle throughput is achievable
by linking the lights together on a fixed-time plan controlled by computer. This time
plan can then be altered relative to different times of day (e.g. there can be separate
“morning peak”, “off-peak”, “evening peak” and “night-time” plans). The principle
of vehicle actuation relies on detectors that are capable of detecting the difference
between traffic volumes on each approach to the intersection. This data is then used
in order to extend or reduce the amount of “green-time” assigned to each approach at
an intersection, depending on the relative traffic volumes at that approach. SCATS,
for example, uses inductive loop detectors in order to measure the flows of traffic
travelling through an intersection. From these detectors SCATS can compute three
fundamental values (original volume, reconstituted volume and degree of saturation)
for each lane that has detector implementation. By constantly changing the signal
timings according to traffic demand, the result is reduced delay, shorter queues and
decreased travel times. Some of the most commonly implemented adaptive control
systems are SCOOT, SCATS, OPAC and RHODES (Martin et ah, 2003).
1.2.3.2
SCATS
SCATS (Sydney Coordinated Area Traffic System) was originally developed in the
early 1970’s by the Roads and Traffic Authority of New South Wales, Australia and
was first implemented in Dublin in 1989 (Dineen, 2000).
SCATS utilises a hierarchical control architecture which is composed of 2 levels,
strategic and tactical. At the strategic level, a “subsystem” or a network of up to 10
intersections, is managed by a regional computer to coordinate signal timings. These
subsystems are capable of linking together to create a larger “system” operating on
a common cycle time. At the tactical level, optimisation occurs at the intersection
level within constraints that are dictated by the regional computer’s strategic con
trol. Tactical control permits premature termination of green phases when the traffic
demand is less than average and for phases to be omitted entirely when there is no
demand. All the extra green time is deposited to the main phase or can be used by
subsequent phases.
SCATS relies on 4 modes of operations. The first or “normal mode” provides
integrated traffic responsive oi)eration. fn the second or “fail-back mode,” time-of-
day plans are implemented following a computer or communication failure. In the
third mode, known as “isolated control mode,” there is just local vehicle actuation
with isolated control, while the fourth mode, the normal signal display shows flashing
yellow or flashing red on all approaches (Martin et ah, 2003).
1.2.3.3 Traffic Detectors
The resurfacing of roadways and utility repair can also result in the necessity to
reinstall these types of sensors.
Non-intrusive sensors are installed aboveground and can be mounted above the
lane of traffic that they are monitoring or on the side of a roadway where they can
view multiple lanes of traffic at angles perpendicular to, or at an oblique to the di
rection of flow. The technologies presently in use are video image processing, laser
radar, microwave radar, ultrasonic, passive infrared, passive acoustic array, and com
binations of sensor technologies such as passive infrared and microwave Doppler or
passive infrared and ultrasonic (Mimbela and Klein, 2000). Similar to the intru
sive sensors, the aboveground sensors measure vehicle count, passage, and presence.
A large number also provide vehicle speed, vehicle classification, and multiple-lane,
multiple-detection zone coverage.
Inductive Loop Detectors(ILD)
ILDs are the most extensively used sensors in traffic surveillance and management
applications. Presently, most incident detection systems and algorithms use traffic
data derived from ILDs (Parkany and Xie, 2005).
The standard ILD is a length of insulated wire bent into a closed shape, typically a
square or a rectangle, and attached to a power source/sensor on both sides of the wire.
The wire loops are embedded in a shallow cutout in the pavement. A lead-in cable
connects from a roadside pull box to the controller cabinet, then to an electronics unit
located in the controller cabinet. When a. vehicle stops on or drives over the loop, the
inductance of the loop reduces, which in turn, increases the oscillation frequency and
causes the electronics unit to transmit a pulse to the controller, indicating the transit
of a vehicle and registering its presence in its detection zone. More recent versions of
ILDs use higher frequencies to identify specific metal components of vehicles, which
can be used to classify vehicles (Parkany and Xie, 2005).
locations and environments (Parkany and Xie, 2005). In use, ILDs tend to go out
of “tune” over time and need readjustment. The pertained presence information can
be used to compute volume and occupancy. Occupancy is calculated by taking the
ratio of time the detector registers the presence of vehicles in its detection zone to
the overall sample time.
ILD Reliability
As with all sensors, reliability of inductive loop detectors is an issue. It has been
claimed by Underwood (1990) that in most cities with mature systems it is reported
that 25 to 30 percent of their detectors are not operating correctly at any particular
time. Detection errors can be attributed to inclement weather, improper connections
in pull boxes, and to the application of sealants over the cutouts (Parkany and Xie,
2005). These problems are exacerbated in cases where ILDs are installed in pavements
of poor condition.
Chapter 2
Algorithm for Cycle Generation
2.1 Introduction
This chapter is a detailed overview of the algorithm used for driving cycle generation
throughout the work presented in this dissertation.
Presently, there is no universally approved or standard method for developing a
driving cycle to represent driving conditions in a particular location. Currently, every
method is open to interpretation by the team of researchers creating the cycle and
can be adjusted to suit the needs of the network being represented. They do have
one commonality however, there is always some form of human input and decision
making during the process of cycle creation.
based on a microtrip methodology. The method should be suitable for a wide vari
ety of driving conditions and road types. The main benefits of having a completely
automated process are: 1) the program can be used by anyone, even if they have a
very limited knowledge on cycle creation practices or even driving cycles in general,
2) having no need for human decision making allows the cycle creation process to be
massively sped up, and limitations rest solely on the computer hardware being used,
3) the fact that there is no decision making bias combined with the consistency of
results, makes this method ideal for the comparison of different driving cycles. The
combination of these benefits results in an ideal research tool.
microtrip microtrip microtrip duration average speed composition
(V, A, B) |a,A, B)
microtrip comparison (mav, mpa, mpd, mpc)
H
group identifiers
microtripgroup^nums, microtrip_group^frequency
cycle creation
(mav, mpa, mpd, mpc, mt. microtnp_group_freqi>ency, microtrip_group_nums avg_trip_length. avg_idle_tim€, bav, bpa, bpd, bpc, bit, A, B, v, t, a)
T
Variable
Description
V
Speed, [km/hr]
a
Acceleration, [m/s^]
t
Time, [seconds]
N
Unique vehicle number.
distX
Total distance travelled at each time step,
[rn]
numVeh
Number of vehicles for which data was recorded.
bav
Average speed of the base data, [km/hr]
bpa
Percentage of the base data in acceleration mode.
bpd
Percentage of the base data in deceleration mode.
bpc
Percentage of the base data in cruise mode.
bit
Percentage of the base data that is idle time.
Avg_dist
Average distance that vehicles travelled, [m]
SAFDnnatrix
Speed acceleration frequency distribution matrix.
A, B
Start and end indexes dehning each microtrip.
mt
Duration of each microtrip. [seconds]
mav
Average speed of each microtrip. [km/hr]
mpa
Percentage of each microtrip in acceleration mode.
inpd
Percentage of each microtrip in deceleration mode.
mpc
Percentage of each microtrip in cruise mode.
Variable
Description
group identifiers
Numbers to categorise the properties of rnav, mpa,
mpd, and mpc in each microtrip.
UI
Unique identifier number.
MN
Microtrip number.
microtrip_group_nums
Microtrips with similar properties are assigned the
same group number.
microtrip.groupTrequency
The frequency that each microtrip group number occurs.
avg_trip_length
Average trip length for vehicles in the base data, [seconds]
avg_idle_time
average idle time for vehicles in the base data, [seconds]
mnJinal
Microtrip numbers chosen for the hnal cycle.
perfJndex
Describes how closely the final cycle matches up to
the base data.
level
Shows which algorithm was used in the cycle creation
function.
finaLcycle_max_v
Max speed in the final cycle.
finaLcycle_max_a
Max acceleration in the final cycle.
\m/s^\
finaLcycle_niax_d
Max deceleration in the final cycle,
[m/s^]
finaLcycleJength
Duration of the final cycle, [seconds]
finaLcycleJdle.time
Total idle time in the final cycle, [seconds]
final_cycle_per_idle_time
Percentage of the final cycle that is idle time.
finaLcycle_per_accel
Percentage of the final cycle in acceleration mode.
finaLcycle_per_decel
Percentage of the final cycle in deceleration mode.
finaLcycle.per .cruise
Percentage of the final cycle in cruise mode.
The following is a list of the main functions contained within the program in order
of occurrence:
• 1) Average Speed
• 2)Average Distance
• 3)Speed Acceleration Frequency Distribution Matrix
• 4) Microtrip Generation
• 5)Microtrip/Segment Filter
• 6) Microtrip Duration
• 7)Microtrip Average Speed
• 8)Microtrip Composition
• 9)Microtrip Comparison
• 10)Final Sorting
• ll)Cycle creation
These functions are described in the succeeding sections, and are presented in a
pseudocode format. Within the pseudocode, the notation is used to represent
assignment, and “//” is used to denote comments. The original MATLAB code for
the functions in this chapter are presented in Api)endix A.
2.2 Average Speed
start function
vtot:=the sum of all v values
bav:=vtot/the number of values for v
end function
2.3 Average Distance
This function is specific to data generated from VISSIM microsimulation software. Its
purpose is to calculate the average distance travelled by a vehicle. This will be used
in conjunction with average speed to work out the average trip length. For data col
lected by alternative means, the average trip length may be calculated using a different
method. This function also outputs the total distance travelled by all vehicles, but is
not needed for subsequent functions and is merely provided as additional information.
start function
add arbitrarily large number to the end of N
//Searches for when N cheinges & records last distX value for i:= 1 to (length of N)-l
if N(i+1) does not equal N(i)
X:=distX(i) Z:=Z+X end
end
total_dist:=Z Avg_dist:=Z/numVeh
output total_dist output Avg_dist
2.4 Speed Acceleration Frequency Distribution Ma
trix
This function takes in speed and acceleration, and creates a speed acceleration fre
quency distribution matrix. Each column represents a new acceleration bracket and
each row represents a new velocity bracket. It sequentially goes through every row
of data and counts the frequency of each speed acceleration occurrence, adding the
value one to the relevant matrix cell. Note that idle time is represented by the zero
acceleration/zero speed cell, and cruise is represented by the zero acceleration/non
zero speed cells. An example of an SAFD matrix is provided in Figure 2.2. This ma
trix is then normalised to 100 percent as is shown in Figure 2.3. From this normalised
matrix, parameters describing the base data are extracted, such as percentage accel
eration (bpa), percentage deceleration (bpd), and percentage cruise (bpc).
SAFD Matrix
acceleration {m/s^2] Total
< 9 -9^7 6 -5 -4 .3 -2 -1 0 1 2 3 4 5 6 78 9 >9 speed rkm/hl
0 0 0 0 0 0 0 0 0 0 0 797 0 0 0 0 0 0 0 0 0 0 797
0-5 0 0 0 0 0 0 0 0 14 144 76 119 7 0 0 0 0 0 0 0 0 360
5-10 0 0 0 0 0 1 0 5 29 44 10 21 32 0 0 0 0 0 0 0 0 142
1015 0 0 0 0 0 0 0 2 27 51 10 30 33 3 0 0 0 0 0 0 0 156
1520 0 0 0 0 0 0 1 1 32 51 14 51 33 2 0 0 0 0 0 0 0 185
20-25 0 0 0 0 0 0 0 1 27 112 59 80 36 1 0 0 0 0 0 0 0 316
25-30 0 0 0 0 0 0 0 2 24 147 91 102 41 1 0 0 0 0 0 0 0 408
30 35 0 0 0 0 0 0 0 0 23 190 101 167 33 2 1 0 0 0 0 0 0 517
3540 0 0 0 0 0 0 2 1 7 273 231 244 29 1 0 0 0 0 0 0 0 788
4045 0 0 0 0 0 0 0 0 S 255 282 252 27 4 0 0 0 0 0 0 0 825
45-50 0 0 0 0 0 0 1 0 3 126 153 136 17 0 0 0 0 0 0 0 0 436
50-55 0 0 0 0 0 0 0 0 0 66 72 76 3 0 0 0 0 0 0 0 0 217
55-60 0 0 0 0 0 0 0 0 0 16 17 23 0 1 0 0 0 0 0 0 0 57
60-65 0 0 0 0 0 0 0 0 0 3 3 5 1 0 0 0 0 0 0 0 0 12
>65 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
Total 0 0 0 0 0 1 4 12 191 1478 1916 1306 292 15 1 0 0 0 0 0 0 5216 average speed [krrVtir] (bav) 26 989
average tnp length [s] 274 53
•die time (%] (brt) 15 28
average idle time fsl 41 947
Figure 2.2: SAFD matrix
Normalised SAFD Matrix
acceleration (m/s*?] %
<-9 9 -8 -7 -6 -5 4 -3 2 1 0 1 2 3 4 5 6 7 8 9 >9 speed (km/h)
0 0 0 0 0 0 0 0 0 0 0 15 28 0 0 0 0 0 0 0 0 0 0 15 27991
0-5 0 0 0 0 0 0 0 0 0 268 2 761 1 457 2 281 0 134 0 0 0 0 0 0 0 0 6 90184 510 0 0 0 0 0 0 019 0 0 0959 0 556 0 844 0.192 0 403 0613 0 0 0 0 0 0 0 0 2 722393 10-15 0 0 0 0 0 0 0 0 0383 0 518 0 978 0 192 0 575 0 633 0 058 0 0 0 0 0 0 0 2 990798 15-20 0 0 0 0 0 0 0 019 0 0192 0 613 0 978 0 268 0 978 0 633 0 038 0 0 0 0 0 0 0 3 546779 20-25 0 0 0 0 0 0 0 0 0192 0 518 2.147 1 131 1.534 0 69 0 019 0 0 0 0 0 0 0 6 058282 25-30 0 0 0 0 0 0 0 0 0383 0.46 2 818 1 745 1 956 0 786 0 019 0 0 0 0 0 0 0 7 622086 30-35 0 0 0 0 0 0 0 0 0 441 3 643 1 936 3 202 0 633 0 038 0 02 0 0 0 0 0 0 9 91181 3540 0 0 0 0 0 0 0 038 0 0192 0 134 5 234 4 429 4 678 0 556 0 019 0 0 0 0 0 0 0 15 10736 4045 0 0 0 0 0 0 0 0 0 0% 4 889 5406 4 831 0 518 0 077 0 0 0 0 0 0 0 15.81672 45-50 0 0 0 0 0 0 0 019 0 0 058 2416 2 933 2 607 0 326 0 0 0 0 0 0 0 0 8 3588% 50 55 0 0 0 0 0 0 0 0 0 1 266 1 38 1 457 0 058 0 0 0 0 0 0 0 0 4 160276 55-60 0 0 0 0 0 0 0 0 0 0 307 0 326 0441 0 0 019 0 0 0 0 0 0 0 1 092791 60-65 0 0 0 0 0 0 0 0 0 0058 0.058 0 0% 0019 0 0 0 0 0 0 0 0 0230061
>65 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
% 0 0 0 0 0 0 019 0 077 0 2301 3 662 28 34 36 73 25 04 5.598 0.288 0 02 0 0 0 0 0 0 100 base% acci (bpa) 30 943
base% decel (bpd) 32 324 base%cruise (boc) 21 453
start function
create a 15 by 21 matrix full of zeros and call it M
for i:=l to length of v
//Sets initial lower and upper bounds for speed and //acceleration brackets
vel_lower:=5 vel.upper:=10 accel_lower:=-9.1 accel_upper:=-8.1
//Counts idle time if v(i)=0
add 1 to row 2, column 11 of matrix end
if v(i)>0 and v(i)<5
if a(i)<-9.1
for j:=1 to 9
if a(i)>=accel_lower and a(i)<accel_upper
add 1 to row 2, column j+1 of matrix end
increase accel_lower by 1 increase accel_upper by 1 end
if a(i)>=-0.1 and a(i)<=0.1
add 1 to row 2, column 11 of matrix end
if a(i)>0.1 and a(i)<l.l
add 1 to row 2, column 12 of matrix end
accel_lower=l.1 accel_upper=2.1
if a(i)>=accel_lower and a(i)<accel_upper
add 1 to row 2, column j+1 of matrix end
increase accel_lower by 1 increase accel_upper by 1 end
if a(i)>=9.1
add 1 to row 2, column 21 of matrix end
end
for k:=l to 12
if v(i)>=vel_lower and v(i)<vel_upper
if a(i)<-9.1
add 1 to row k+2, column 1 of matrix end
if a(i)>=accel_lower and a(i)<accel_upper
add 1 to row k+2, column j+1 of matrix end
increase accel_lower by 1 increase accel_upper by 1 end
if a(i)>=-0.1 and a(i)<=0.1
add 1 to row k+2, column 11 of matrix end
if a(i)>0.1 and a(i)<l.l
add 1 to row k+2, column 12 of matrix end
accel_lower:=1.1 accel_upper:=2.1
for j:=12 to 19
if a(i)>=accel_lower and a(i)<accel_upper
increase accel_lower by 1 increase accel_upper by 1 end
if a(i)>=9.1
add 1 to row k+2, column 21 of matrix end
end
increase vel_lower by 5 increase vel_upper by 5
end
if v(i)>=65
if a(i)<-9.1
add 1 to row 15, column 1 of matrix end
for j:=1 to 9
add 1 to row 15, column j+1 of matrix end
increase accel_lower by 1 increase accel_upper by 1 end
if a(i)>=-0.1 and a(i)<=0.1
add 1 to row 15, column 11 of matrix end
if a(i)>0.1 and a(i)<l.l
add 1 to row 15, column 12 of matrix end
accel_lower:=1.1 accel_upper:=2.1
for j:=12 to 19
if a(i)>=accel_lower and a(i)<accel_upper
add 1 to row 15, column j+1 of matrix end
end
if a(i)>=9.1
add 1 to row 15, column 21 of matrix end
end
end
output matrix M to a text file called ‘'SAFD_matrix.txt’’
2.5 Microtrip Generation
The inputs for this function are vehicle number and speed data, and it generates mi
crotrips for data sections that start and end with zero speed. In some cases, namely
where a vehicle enters or leaves the network, the speed at one end may not be zero.
The microtrips which correspond to these situations should technically be classified
as segments as they don’t both start and end with zero speed, and will be filtered
out in the next function. An index system is used to identify microtrips, whereby the
first number indicates the row of data for the start point and the second number the
row of data for the end point. Using this method is very efficient as all other data
relating to each microtrip can be stored in a small amount of space. The indexes are
stored in a two-dimensional array where each row represents a new microtrip, and
the column entries represent the start and end points.
The first part of this function runs through a “for loop” and sei)arates the data
by vehicle number (N) using an index system. It stores the starting indexes in vector
“S” and the final indexes in vector “F”
start function
add arbitrarily large number to end of N first value in S:=l
x:=l k:=l kl:=l
for i:=l to (length of N)-l
if N(i+1) does not equal N(i)
F(kl):=i
increase kl by 1
S(kl):=i+l
end
end
This part of the function searches through the data for each individual vehicle and
determines where the starting point of the first microtrip is, and does this for each
new vehicle, “indexl” is always the starting point of a microtrip.
for i:=l to length of S
for j:=S(i) to F(i)
//This finds indexl when the start of the data set consists
//of zero speed values
if j>l and j<F(i) and v(S(i))=0 and v(j)=0
and v(j+l) does not equal 0
indexl:=j
//Sets indexl as the starting index S if the first speed
//value is zero but the second isn’t
else if v(S(i))=0 cind v(S(i) + l) does not equal 0