Integrated Modeling of Road Environments for Driving Simulation

Carlos Campos

1,2

, João Miguel Leitão

and António Fernando Coelho

Department of Electrical Engineering, School of Engineering of Porto Polytechnic, Porto, Portugal

Department of Informatics Engineering, Engineering Faculty of Porto University, Porto, Portugal

Keywords: Procedural Modelling, Road Network, Road Layout, Road Design, Driving Simulation.

Abstract: Virtual environments for driving simulations aimed to scientific purposes require three-dimensional road

models that must obey to detailed standards of specification and realism. The creation of road models with

this level of quality requires previous definition of the road networks and the road paths. Each road path is

usually obtained through the dedicated work of roadway design specialists, resulting in a long time

consuming process. The driving simulation for scientific purposes also requires a semantic description of all

elements within the environment in order to provide the parameterization of actors during the simulation

and the production of simulation reports. This paper presents a methodology to automatically generate road

environments suitable to the implementation of driving simulation experiences. This methodology integrates

every required step for modelling road environments, from the determination of interchanges nodes to the

generation of the geometric and the semantic models. The human supervisor can interact with the model

generation process at any stage, in order to meet every specific requirement of the experimental work. The

proposed methodology reduces workload involved in the initial specification of the road network and

significantly reduces the use of specialists for preparing the road paths of all roadways. The generated

semantic description allows procedural placing of actors in the simulated environment. The models are

suitable for conducting scientific work in a driving simulator.

1 INTRODUCTION

Procedural modelling of realistic road environments

is a research area of great interest, which is

dedicated to generating 3D models, not only for

entertainment but also for scientific applications.

This paper presents the current implementation state

of a methodology for the automatic creation of road

environments for driving simulation for scientific

purposes.

Driving simulators are an important research tool

in several scientific areas, such as psychology,

ergonomics and roadway engineering.

In psychology, they are used to develop

researches related with the driver behaviour. For

example, evaluate the interference of secondary

tasks in principal driving task, like mobile phones,

navigation systems or traffic information systems.

The driving simulators are also used in

ergonomics area to study "In Vehicle Information

System", of interaction with the driver, like

navigation systems (GPS) or mobile phones.

In roadways engineering, they are used to

analyses road paths in design stage, but also real

roads, allowing the test with real drivers, for

example the study of factors that conducts to

dangerous driving.

In traffic engineering also allow the study of

dangerous overtaking maneuver with frontal

collision probability in roads with two lanes, one in

each direction, without compromise the driver

safety.

Driving simulation experiments require the

creation of extensive road environments with

specific technical features, such as the placement of

actors and the production of detailed reports. In

order to place the actors in the simulated

environment and produce the reports, a semantic

description of the entire generated environment is

also required.

Road environments for driving simulation

usually consist of road networks. The strategic

design of a road network involves the previous

definition of nodes (cities, intersections,

roundabouts, etc.) that will be interconnected and the

type of road to implement.

This definition often involves very different

Campos C., Leitão J. and Coelho A..

Integrated Modeling of Road Environments for Driving Simulation.

DOI: 10.5220/0005308600700080

In Proceedings of the 10th International Conference on Computer Graphics Theory and Applications (GRAPP-2015), pages 70-80

ISBN: 978-989-758-087-1

 2015 SCITEPRESS (Science and Technology Publications, Lda.)

disciplines, in addition to the roadways design, and

requires an analysis of the network as a whole.

This process produces the topological definition

of the roads network, that specifies the connection

between nodes and the related typologies of roadway

(highway, secondary roads and rural). The highways

connect cities, secondary roads connect cities with

small towns, and rural roads connect small towns

with villages.

The creation of this road network definition can

become a too hard-working and resource consuming

task, if done by specialists applying traditional

roadway design methods. These difficulties cannot

be completely solved by the use of any of the

available dedicated tools, like the Autodesk Civil3D.

The road path definition is very important to the

realism of the road environment. The road path

should have geometric characteristics like those

found in the real world.

After obtaining the road path, it is still necessary

to generate the visual model. One way to optimize

this process is to use automatic modelling tools to

facilitate this task, like the ones presented in

(Campos, 2007).

Additionally, besides generating the model of the

network and the related road paths, it is still

necessary to edit the terrain model surrounding each

road path that makes up the roads network, to

produce models that meet the required visual

quality.

For driving simulation experiments, the required

high fidelity visual models must also be completed

with the corresponding semantic description of the

entire virtual environment, as also pointed out in

(Thomas, 2000).

The semantic definition is a high-level

description and characterization of the road

environment, which allows parameterizing the

dynamic model of the virtual environment. This

description is fundamental in simulation systems for

placing actors (e.g. pedestrians, other vehicles),

monitoring the entire simulation and report

generation. The availability of this reports and the

details included are very important in scientific

driving simulation experiments. These reports are

essential for further analysis and data processing

relating to experimental test. Traditional modelling

tools for virtual environments do not allow the joint

creation of these two representations, semantic and

visual.

To allow the generation of fully controllable

environments, the preparer of the experiment must

be able to interact with the generation process, by

imposing specific requirements for each element.

This feature is critical in the generation of virtual

environments for driving simulation for scientific

purposes.

In this paper, section 2 presents the state of art in

the generation of road environments and the relevant

related work. The proposed methodology is

presented in section 3, and in section 4 we present

the results obtained with the implemented prototype.

Section 5 presents the conclusions and proposes

some guidelines for future work.

2 STATE OF ART

2.1 Roads Design

As described in (Campos, 2012), a road network can

be defined by a set of nodes and links that connect

them. Each road is planned to connect two or more

points of interest, allowing the movement of

vehicles between them. Planning a road network

aims to satisfy the needs of people, seeking to meet

the standards of service levels. A road network can

be defined by a hierarchical structure of connections

in 3 levels: the highway network, the secondary road

network and rural road network (Teoh, 2008)

(Weber, 2009).

The design of a road starts with a strategic

decision to connect two points on the existing roads

network. Essentially, this decision is made by taking

into account parameters such as population growth,

level of service, traffic studies and estimated

construction cost.

In roadway engineering, a project of a road path

is developed according to the following phases:

preliminary program, prior study and project of

implementation (França, 2011). The preliminary

program is basically a specification consisting of

general rules, with little detail, which include the

general characteristics desired and that may

condition the choice of road path.

The preliminary study is not intended to provide

definitive elements, nor the detail required for the

road works. The main objective is to point out one or

several alternative solutions for tracing the road

path. This stage involves traffic studies, geological

and geotechnical studies, landscape and

environmental impact studies. The results of these

studies will restrict the road path of the roadway, but

leaving some degrees of freedom to a further

decision. When the previous study is concluded and

approved, the project of road path begins. The route

is now defined with necessary detail to its

implementation in the real world. The definition of

IntegratedModelingofRoadEnvironmentsforDrivingSimulation

the horizontal alignments of the road is the first task

to be performed in this step, followed by the

remaining specialities.

In order to produce the horizontal alignments,

one begins by specifying the centreline, with the

introduction of horizontal curves. It follows the

vertical alignment (altimetry), where vertical curves

are defined. Lastly, the road cross-section specified.

The road path definition must conform to the rules

established by the applicable standards design. In

Portugal, these standards are defined by Estradas de

Portugal, which is responsible for maintaining the

road network (EP, 1994).

2.2 Related Work

Several works have been developed for creating road

city environments, focusing mainly on urban layout

(Parish, 2001)(Chen, 2008)(Teoh, 2008)(Coelho,

2007)(Vanegas, 2010).

Typically, the produced networks of urban roads

follow a certain predefined pattern: orthogonal,

radial or branching (Sun, 2002). The results obtained

by these methods are usable for most visual

simulation applications. The problem with these

approaches is the need for an excessive control by

the user to obtain realistic models. Also, the use of a

pre-defined pattern for the road network is not

adequate when specific road layouts are required.

Besides, these methods don’t produce extra-urban

roads network.

Procedural modelling languages, e.g. L-systems

and CGA shapes, are not adequate to produce

extensive road paths for driving simulation (Paris,

2001)(Muller, 2006)(Coelho, 2007). Is known that

the driver easily learn the road layout in an

experimental work, therefore repetition of road

segments is not advisable. The expectation

component of the road path is a very important

variable in a driver behaviour research.

Driving simulation experiments typically require

mixed road environments, with roads models of

good quality, with realistic road paths, combining

roads from urban and rural environments.

A proposal for the procedural modelling of

virtual environments suitable for real-time

simulation is presented by Smelik et al. in 2008

(Smelik, 2008). The authors describe a process in

which the virtual environment is created

hierarchically from an initial sketch, including

treatment of the road network. A benefit of this

hierarchy is that each layer can be treated

individually and the final model generation process

takes into account the appropriateness of all settings,

including the terrain model, as discussed in (Latham,

2006). In the proposal presented by Smelik et al., it

isn’t clear what methodology was used to generate

the road path. In a further publication in 2001,

Smelik et al. present a modelling approach of a

virtual environment, using an interactive modeller

(Smelik, 2011). The generation of the virtual

environment requires an excessive control by the

user so that the process ceases to be fully automatic.

For the generation of geometric models of roads,

some procedural modelling techniques of urban

networks presented by Kelly et al. in 2008 are

explored (Kelly, 2008).

Galin et al. present a methodology for procedural

generation of roads paths, where the generation of

the road path is done using a complex algorithm, on

a terrain defined as a digital elevation map (Galin,

2010). At each step of the process the direction to

follow is determined by the local evaluation of the

cost of each alternative. This use of a local

assessment prevents a proper evaluation and global

optimization of the final solution. Depending on the

terrain model and the typologies of the road, the

option for a maximum number of alternatives to be

study may be too complex. The presented

description does not allow to concluding if the

proposed method produces road layouts similar to

those found in real road paths. The realism

evaluation of the obtained models is also not

presented. During the generation of the visual

model, all 3D models, such as bridges, tunnels and

road crossings are instantiated.

A methodology for generating a road network

that integrates different typologies of roads

(highways, main roads and secondary roads) is

presented by Galin et al. in (Galin, 2011). The

generation of the global transport network is

initiated by the creation of the motorway network,

followed by the main network and the secondary

network last. Each type of roads is created

independently. At the end, all the roads are

overlapped in a single road network. The way this

problem is handled, requires that the design of roads

has to be recalculated.

Thomas et al. showed a model implementation of

a virtual city where several actors (like pedestrians,

vehicles and public transport vehicles driven by

other drivers circulating in different directions) are

included and controlled over the complex urban road

environment (Thomas, 2000). The urban

environment was created using the interactive

modeller VUEMS and based on a terrain definition

in a digital elevation map (Donikian, 1997). The use

of an interactive modeller is not suitable for the

GRAPP2015-InternationalConferenceonComputerGraphicsTheoryandApplications

generation of extensive road environments

automatically and efficiently.

In the work presented by Bayarri et al.,

methodologies for the generation of road

environments appropriate for driving simulation

from design data are described (Bayarri,

1996)(Pareja, 1999). Each time you want to get a

model of a roadway, it requires prior interaction with

the area of engineering of roadways, or other

sources, for the delineation of the road paths. Some

of these methodologies have been explored in

previous works (Campos, 2007).

In order to implement experiments in driving

simulation, beyond the generation of road layout and

modelling of virtual environments, the

corresponding semantic definition should also be

created. As mentioned in (Thomas, 2000), the

semantic definition is a description of the

characterization of high-level road environment,

which supports the dynamic model of the virtual

environment. The creation of a semantic description

in accordance with the road map is also addressed in

other works (Campos, 2007)(Bayarri, 1996)(Pareja,

1999).

The OpenDrive Project has developed a standard

for the complete description of a road network to be

compatible with various simulation systems

(OpenDrive, 2010). It is stated that the proposed

standard allows the parameterization of the main

features of real roads. An interactive modeller that

uses this standard is the Road Designer (ROD,

2011).

Creating models of realistic roads for driving

simulation is usually performed by manual,

laborious and slow processes, even when the

available interactive tools are applied. Whenever a

realistic road environment to driving simulation

experiments is needed, a high level of interaction

with the preparer is usually necessary, which

requires experts in road design. An alternative to this

problem is to develop a methodology for automatic

generation of road environments, from the definition

of nodes, topological definition of the network,

generation of road paths and semantic description

until the generation of the visual model.

3 INTEGRATED MODELING OF

ROAD ENVIRONMENTS

The proposed methodology creates, in an integrated

way, a complete road environment from the

generation of inter-changes nodes to the construction

of the three-dimensional model. The process of

generating the three dimensional model is organized

hierarchically in layers, which also gives a semantic

description of the road environment, as illustrated in

figure 1.

Figure 1: Overview of process workflow.

The preparer of the road environments can interact

at any stage of the model generation process, in

order to customize the final result. This allows a

complete control over each process and the

production of models with the specific required

details.

In this approach, the procedural generation of a

road environment starts at the Nodes module, which

creates a definition of interconnection nodes on a

given area of terrain. This module has, as input

parameters, the definition of the area where the

nodes will be distributed and the number of nodes

that will be generated.

The terrain model can be associated with a

definition of constraints that should be considered in

the generation of nodes. In the produced list of

nodes, each node is associated with its location and

an importance factor. The importance factor can

then be used to organize the nodes by typology.

The Topological Roads module is aimed in

generating a definition of a topological road

network. The topological road network is created

hierarchically by levels, similar to what happens in

other works (Teoh, 2008)(Weber, 2009)(Galin,

2011). The Topological Roads module receives as

input the definition of nodes generated by the Nodes

module and, depending on the importance factor,

organizes the nodes by typologies: cities, towns and

villages. Based on this organization, the topological

definition of each network is created. The secondary

network is originated from town typed nodes, and

the rural network is created from the village nodes.

IntegratedModelingofRoadEnvironmentsforDrivingSimulation

Thus, highways connect cities, secondary roads

connect cities and towns, and rural roads connect

towns and villages. Additional nodes of inter-

connection may be created, with an importance

factor of zero.

The Centerline module is responsible for

generating the definition of road paths for each

roadway in the network. This module receives the

definition of the topological road network generated

by the Topological Roads module. For each

connection, a definition of the road path is created.

The methodology of generating a road path was

inspired by the methods used in roadways

engineering, producing models that meet the design

standards and with geometric characteristics similar

to those found in real world.

The suitability of the terrain model to the

generated road network is performed by Road

Enviro module. This module receives the definition

of the road paths of all the roadways generated by

the Centerline module and fits the terrain model to

the respective geometric definition of each roadway.

The Models Generator module is responsible for

procedurally generating the geometric model of the

roadways, where the previously studied techniques

are explored (Campos, 2007). This module is also

responsible for generating the geometric model of

the terrain.

In the hierarchical process of generating a road

environment, the preparer of the model can interact

at any stage of the process, modifying the result

produced by any module and creating a custom

layout. For example, the preparer can add new

nodes, edit or delete a connection, redefine a road

path or edit the definition of the terrain.

This freedom of interaction with the model at

any stage of the generation process is a special

contribution of this methodology that enables

producing fully customized models, allowing fulfil

of any particular specification required for the

experimental work.

At the end of the generation process, the

semantic description is also made available, which

allows the parameterization and placement of actors

in the simulated environment, e.g. other vehicles,

pedestrians or any other moving element. This

facility is crucial for implementing driving

simulation experiments aimed to scientific purposes.

3.1 Nodes and Topologic Roads

Network

The proposed methodology to create a topological

road network requires the previous definition of the

interchange nodes set. If there is no previous

definition of these nodes, the Nodes module is able

to automatically generate a simplified definition of

nodes on an area of terrain, in which each node is

associated with an importance factor. The

importance factor can match the population and is

used to organize the nodes by typology. The nodes

are organized into three typologies: city, town ad

village. Once the definition of nodes is available,

different topological roads networks are created:

highways, secondary and rural.

The process starts from nodes of city typology,

creating a network of highways. All possible

connections linking a pair of cities are analysed and

the benefit of its existence is evaluated. During the

review process, the decision to remove a connection

depends on the difference between the cost of road

construction and the benefit of keeping that

connection. The benefit is related to the utilization

index (U), which depends on the total length of the

roadway (L) and the population resident in each of

the ends nodes, PA and PB (1).

U = PA * PB / L

(1)

The benefit (B) is determined according to the

utilization index U, the difference between the road

connection length and the shortest alternative route,

and the unitary cost of operation (Cop), which

measures the traveling cost of vehicle per kilometre

(2).

B = U * (L

connection

- L

shortest

) * C

(2)

If an edge connection of the graph is removed, the

cost of this decision makes the movement between

the ends nodes may be longer, what don’t match a

particular problem.

After simplifying the roads network, there can be

many intersections between two road paths. In this

case, additional interconnection nodes are generated.

These nodes have an associated importance factor

(population) of zero. Interconnection nodes with

zero population appear as gray nodes in figure 2.

In order to generate the secondary road network,

the nodes with village typology are added to the

highways network obtained in the previous process.

All possible connections between villages and towns

are then analysed, using the same procedure applied

to the highway network.

The process of generating the topological roads

networks may be parameterized in order to obtain a

representation of highways, adequate and similar to

a real roads network. At the end of the process, we

obtain a definition of the network of connections,

between two nodes, as shown in figure 2.

GRAPP2015-InternationalConferenceonComputerGraphicsTheoryandApplications

Figure 2: Topological roads network.

After obtaining the definition of the topological

roads network, a road path is generated for each

roadway.

3.2 Road Paths

As described in (Campos, 2014), the selection of a

road path (T) connecting two given nodes (a and b),

can be explained by an optimum search function

aimed to minimize a cost function that evaluates the

influence of different indicators, such as the

relationship with the terrain definition and the

constraints.

The proposed procedure for generating a road

path is hierarchically organized in layers. Starting

from the definition of a connection between two

nodes, a set of high level solutions with little detail

is first generated in the first layer. At each layer,

new road layout definitions with growing levels of

detail are generated.

At each layer, several alternatives road paths are

generated. The choice of the best hypothesis at each

layer is performed by a cost function that evaluates

several parameters, such as terrain affection and the

related constraints.

The strategy of generating a road path according

to a specific level of detail mimics the method

usually applied by civil engineers when designing

road path of real roadways. This straight connection

to real design processes eases the selection of every

required parameter and grants that the produced

output will be similar to a real road path. Once the

road path is defined in straight alignments with the

required level of detail, the horizontal alignments

and the altimetry layout are developed. Horizontal

alignments are achieved with the introduction of

horizontal curves and the development of the in with

the introduction of vertical curves. A horizontal

curve consists in an input clothoid, a circular arc and

an output clothoid.

The produced definition of the road path includes

information about every road segment and its

construction class as normal road, bridge or tunnel.

The costs considered in the evaluation of a road path

are: classified area, bridge construction, tunnel

construction, earthwork, roadway construction and

cost of use. A classified zone is defined as a

geographical area in blueprint that represents a

particular constraint, such as water, forest or

buildings.

To determine the impact of a road path on a

classified zone it is necessary to calculate the area of

the zone affected by the road crossing. To determine

whether a portion of the length of the roadway is

built as a bridge or tunnel, an analysis of the

difference between the elevation of the terrain and

elevation of the roadway is made. It’s also

considered the bridge construction where the

roadway crosses water, for example, a river or a

lake.

To evaluate the moved ground volume it is

necessary to analyse the difference between

elevations of the terrain and the road. The reference

values for the costs analysed in the study of

alternative alignments were determined using

specialists in roadways design, and are discussed in

(Campos, 2014).

3.3 Road Environment

The construction of roadways in real world implies

earthworks in the terrain where the road is

implemented. A slope is the terrain surface which is

located along the road and can be originated by

natural or artificial causes. Artificial slopes are

created during construction of roads and are a result

of excavation or embankment on the terrain. They

are aimed to ensure the stability of the natural

terrain. The declivity of cut slopes and embankment

slopes are defined according to geological and

geotechnical studies. Generally, the cut slope and

embankment slopes have declivity of 1 to 1.5 (V/H)

so that the stability of the ground is guaranteed (EP,

1994).

In the proposed methodology, after generating

the trajectory of the road, it is necessary to adjust the

definition of the terrain to generate the visual model

without failures. These changes to the altimetry of

the terrain model typically result in cut and

embankment slopes.

An efficient way to realize the change in the

altimetry of the terrain consists in, for each point of

the terrain, checking if the altimetry value is

IntegratedModelingofRoadEnvironmentsforDrivingSimulation

influenced by the presence of the road. For each

point of the terrain, the minimum distance to the

road is calculated and, according to this distance, the

need to change the z coordinate of the point of

terrain evaluated. This process may result in the cut

or embankment slopes, as shown in figure 3.

The z coordinate of the point of the terrain is not

altered if it is positioned between the definitions of

slopes, as illustrated in figure 3, for the point P3.

We can also see in figure 3 that point P1 is

shifted to P'1 point, and the point P2 is shifted to P'2

point.

Figure 3: Cut and embankment slope.

In order to optimize this process, the terrain points

that are relatively far are not treated. In the edition of

the terrain, the modelling information of the edited

areas is registered so that, it is possible to provide

these areas with a different color (typically brown)

during its visualization.

The horizontal signalization is made by road

marks, defining the lanes in each road segment.

These marks may have different layouts: continuous,

discontinuous or both. In order to improve the

visibility detection, horizontal signalization models

are hierarchically organized with the road models

using different layers. The vertical signalization is

placed in road environment by instantiating 3D

objects.

In this way the road environment is procedural

modelled, producing 3D models that meet the

detailed standards of specification and required

visual quality.

3.4 Semantic Model

During the creation of a road environment, a

semantic description of the entire model is

generated. This description is fundamental in driving

simulators aimed to scientific purposes, not only for

the production of reports, but also for the

parameterisation of actors and traffic events. The

inclusion of actors and the parameterisation of traffic

events make the driving experience more realistic

and immersive.

In driving simulation, typically road

environments are enriched with the placement of

actors, such as other vehicles and pedestrians. In

driving simulation experiments where the objective

is to study the behaviour of the drivers, the

placement of other vehicles and specification of

precise traffic events is normally required.

In order to place actors and allow the

specification of traffic events, some other levels of

information beyond the definition from the

centerline must also coexist. These include data such

as the roadway width, the width of the road sides,

the number of lanes, the width of each lane and the

road signs.

4 RESULTS

The results presented in section 3.1 were obtained by

the prototype implemented applying the

methodology described in this paper. To

demonstrate the quality of the results obtained by

this methodology and provide a critical analysis, the

following case study was prepared:

The aim of this implementation was to get the road

environment corresponding to two typologies of

roads (rural and highway) for a connection without

nodes, between two cities (Porto and Braga) of the

topological road definition generated, presented in

section 3.1.

The population of Porto is 273 584 inhabitants

and a population of Braga is 93587 inhabitants.

Population was used as an approximation for the

importance factor.

The rural road was designed for a base speed of

60 km/h with two lanes, one in each direction,

corresponding to a total reference width of 12m. The

highway was designed for a base speed of 120 km/h,

with two carriageways, one in each direction, with

two lanes each, corresponding to a total reference

width of 40m.

The model of the real terrain was obtained from

a digital elevation map in GEOTTIF format, from

the United States Geological Survey repository

(USGS, 2014). The constraints due to water and

forest were set to a geographical area affected by the

road layout. In this implementation, the areas that

represent constraints were defined in an image file in

the TIFF format (Libtiff, 2014).

The applied methodology produces several

definitions of the road path, with an increasing level

of detail. The image in figure 4 shows the road path

GRAPP2015-InternationalConferenceonComputerGraphicsTheoryandApplications

generated in straight alignments (black) and the

horizontal alignments, resulting after the

introduction of horizontal curves (blue).

Figure 4: Design of road path.

In figure 5, we present the produced road layout in a

3D view.

Figure 5: Road layout in a 3D view.

It is expected that the road path of a highway is less

winding when compared to a road path designed for

a lower base speed, in order to provide greater

comfort to the driver. Typically, the sinuosity of

horizontal alignments on highways is smaller than

the sinuosity on a rural road. This feature is also

observed in the altimetry layout. One way to analyse

the sinuosity of the horizontal alignment is to use the

discrete Fourier transform and perform an analysis

on the curvature domain. By comparing the results

obtained in the curvature domain of procedurally

generated models to real road paths, allows

concluding that the road paths generated by the

proposed methodology are similar to those found in

real world, like has discussed in (Campos, 2014).

Using the obtained road layout, the terrain model

is modified, in order to generate the corrected visual

model. This terrain correction is performed by a

specially developed module.

In figure 6, we present a rural roadway

environment, designed for a base speed of 60 km/h.

In the figure 7, we present the road environment

of highway, designed for a base velocity of 120

km/h.

Figure 6: Rural road environment.

Figure 7: Highway road environments.

The inclusion of others vehicles can be placed in the

simulated environment, in a procedural way. As we

can see in figure 6, 7 and 8, others vehicle are

procedural placed in the road environment.

IntegratedModelingofRoadEnvironmentsforDrivingSimulation

In figure 8, we show the resulting visual model

with a perfect fit between the generated road and the

surrounding terrain, where we can see cut and

embankment slopes.

Figure 8: Cut and embankment slopes.

Horizontal signs are made by road marks in the

surface of the roads and the vertical signalization are

placed in the road environment instancing 3D

Figure 9: Horizontal and vertical signalization.

models, as we can see in figure 9.

A special characteristic of our approach is the

efficiency and control, to produces tridimensional

models, since the topological road network

definition and the road paths of all roads. In table 1,

we present the time measured to obtain a topological

road definition by our method, with different

number of input nodes.

Table 1: Total number of nodes (city, town, village, and

interconnection nodes), total length of the roads, time (in

seconds) to generate de topological road definition.

Total

Nodes

Cities Towns Villages Total

Length

Time

142 25 16 15 7620 11

342 25 16 15 12166 76

30 18 0 0 3719.2 2

The trajectory of the roads can be controlled by

modifying the parameters of the cost function

(section 3.2). In table 2, we present the time

measured to obtain road paths with different lengths.

Table 2: Length of the road (in kilometres), time (in

seconds).

Length

(km)

Design by

specialists

Time

1000 1 day 3.42

10000 12 days 19

50000 3 months 52

Making an analysis in terms of time, it is concluded

that the proposed method compared to recent

methodologies presented by Galin et al. in 2010

(Galin 2010), significantly reduces the time required

to generate road paths (15%), maintaining the

required standards of realism. This time reduction is

more significant in extensive road paths (> 50 km)

or in a generation of a large road network,

independently of the terrain grid size.

5 CONCLUSIONS

Realistic driving simulation experiments require

prior preparation of models of road environments

correctly designed, with similar characteristics to

those found in real roads. Some experiments also

require models that are recognizable as roads from

the specific simulated country or region. The

presented methodology allows the automatic

generation of any type of road environment suitable

for driving simulation. It integrates all the road

construction stages, from the definition of

GRAPP2015-InternationalConferenceonComputerGraphicsTheoryandApplications

connecting nodes to the visual model.

The proposed generation process of road layouts

was inspired in methods used in roadways

engineering, producing roads according to standards

and similar to those found in the real world. This is a

special contribution of the presented work, that

allow minimize the use of specialists for preparing

the road paths. The introduction of horizontal curves

and vertical curves in altimetry are two prominent

factors that contribute to the realism of the road

paths obtained.

The problem of generating road paths between

two points is globally analysed, at different levels of

detail. The final alignment solution generated at

each level of detail is determined using the

calculation of several costs, by evaluating the

relation with the terrain and related constraints, the

cost of construction and the cost of use. The cost of

use depends with the variation in road layout

altimetry (∆z) and the utilization index (section 3.1).

Throughout the automatic generation process, the

road layout is successively refined, producing

definitions of roadways with the detail required to

generate the visual models.

The designed methodology allows obtaining a

wide range of road layouts, dramatically reducing

work and costs involved in its conception, as it

significantly reduces the use of road design

specialists to obtain the road network definition as

shown by results presented in this paper.

The proposed method eases the inclusion of

actors in the simulated environment. This

improvement is an innovative contribution to the

automatic generation of realistic road environments.

A major area of application of this methodology is

the generation of road environments that are suitable

for driving simulation, allowing the realization of

scientific studies in several science areas.

In the near future it will be possible to present

resulting road environment to road design experts, in

order to best validate the road paths obtained.

ACKNOWLEDGEMENTS

This work had the special contribution of supervisor

of the traffic analysis laboratory, where the driving

Simulator DriS is implemented, Prof. Dr. Carlos

Rodrigues of Civil Engineering Department of

FEUP, and Prof. Ângelo Jacob of Civil Engineering

Department of ISEP.

The Media Arts and Technologies project

(MAT), NORTE-07-0124-FEDER-000061, is

financed by the North Portugal Regional Operational

Programme (ON.2 – O Novo Norte), under the

National Strategic Reference Framework (NSRF),

through the European Regional Development Fund

(ERDF), and by national funds, through the

Portuguese funding agency, Fundação para a Ciência

e a Tecnologia (FCT).

REFERENCES

Bayarri, S; Fernadez, M; Perez, M, 1996. Virtual Reality

for driving simulation, Vol. 39, n.º 5, Communications

of the ACM.

Campos, C.; Leitão, J.; Rodrigues, C.; 2007. Modelação

de Ambientes Rodoviários de Grandes Dimensões,

15.º Encontro Português de Computação Gráfica.

Campos, C., Leitão, J., Coelho, A.; 2012. Modelação

Procedimental de Ambientes Rodoviários para

Simulação de Condução, 20.º Encontro Português de

Computação Gráfica.

Coelho, António; Bessa, Maximino; Sousa, A. Augusto;

Ferreira F. Nunes; 2007. Expeditious modeling of

virtual urban environments with geo spatial L-

systems; Computer Graphics Forum, Vol. 26, N. 4, pp.

769–782.

Campos, C., Leitão, J., Coelho, A.; 2014. Geração

Procedimental de Traçados Rodoviários para

Simulação de Condução, 21.º Encontro Português de

Computação Gráfica.

Chen,G.; Esch,G.; Wonka,P.; Muller,P.; Zhang,E., 2008.

Interactive Procedural Street Modeling, ACM

SIGGRAPH.

Donikian, S., 1997. VUEMS: A Virtual Urban

Environment Modeling System, Computer Graphics

International, Proceedings. IEEE, p. 84-92.

EP, 1994. Book of Road path Standards, Junta Autónoma

das Estradas, Estradas de Portugal (EP), ISBN-96379-

6-2.

França, A., 2011. Book of roadways, Civil Engineering

Department of Engineering Faculty of Porto

University.

Galin, E.; Peytavie, A.; Maréchal, N.; Guérin, E., 2010.

Procedural Generation of Roads, EUROGRAPHICS,

Volume 29.

Galin, E.; Peytavie, A; Guérin, E.; Benes, B., 2011.

Authoring Hierarchical Road Networks, Pacific

Graphics, Volume 30.

Kelly, G.; McCabe, H., 2008. Citygen: An Interactive

System for Procedural City Generation. GDTW, UK.

Latham, R.; Burns, D., 2006. Dynamic Terrain

Modification Using a Correction Algorithm, IMAGE

2006 Conference Scottsdale, Ari-zona.

Libtiff, 2014. TIFF Library and Utilities, available at

http://www.libtiff.org.

Muller, p.; Wonka, p.; Haegler, S; Ulmer, A.; Gool, L.;

2006. Procedural Modeling of Buildings, ACM

SIGGRAPH.

OpenDRIVE, 2010. OpenDRIVE Format Specification

rev 1.3, available at www.opendrive.org/docs/

IntegratedModelingofRoadEnvironmentsforDrivingSimulation

OpenDRIVEFormatSpecRev1.3D.pdf.

Pareja, I.;Bayarri, S.;Rueda, S., 1999. Modelado de

calidad para la visualizacióninteractiva de carreteras,

IX Congresso de Engenharia Informática.

Parish, Y.; Muller, P., 2001. Procedural Modeling of

Cities; ACM Computer Graphics (Proceedings of

SIGGRAPH’2001), pp. 301-308.

ROD, 2011. Road generation tool, available at

http://www.vires.com/Products_Ted.htm.

Smelik R. M.; Tutenel T.; Kraker K.J.; Bidarra R., 2008.

A Proposal for a Procedural Terrain Modelling

Framework, EGVE Symposium.

Smelik, R., Tutenel, T., Kraker, J, Bidarra, R., 2011. A

declarative approach to procedural modeling of virtual

worlds, Computer& Graphics Journal, Elsevier.

Sun, J.; Baciu, G.; Yu, X.; Green, M., 2002. Template-

Based Generation of Road Networks for Virtual City

Modeling, VRST’02, pp 33-40, NY USA.

Teoh, S., 2008. Algorithms for the Automatic Generation

of Urban Streets and Buildings, Proceedings of the

International Conference on Computer Graphics and

Virtual Reality (CGVR'08).

Thomas, G. and Donikian, S., 2000. Modelling Virtual

Cities Dedicated to Behavioural Animation.

EUROGRAPHICS 2000, vol. 19(3).

USGS, 2014. United States Geological Survey, available

at earthexplorer.usgs.gov.

Vanegas, C. A., Aliaga, D. G., Wonka, P., Müller, P.,

Waddell, P., & Watson, B. 2010. Modelling the

appearance and behaviour of urban spaces. In

Computer Graphics Forum (Vol. 29, No. 1, pp. 25-42).

Weber, B.; Müller, P.; Wonka, P.; Gross, M., 2009.

Interactive Geometric Simulation of 4D Cities,

Eurographics, vol. 28.

GRAPP2015-InternationalConferenceonComputerGraphicsTheoryandApplications