A Simulation Tool to Assess the Impact of Deviation Plans on Disruptive

Events of Urban Trafﬁc

Davide Andrea Guastella

1,2 a

, Mois

es Silva-Mu

noz

2,3 b

, Eladio Montero-Porras

2 c

and Gianluca Bontempi

2,3 d

Aix Marseille University, CNRS, LIS, Marseille, France

Machine Learning Group, Universit

e Libre de Bruxelles, Brussels, Belgium

FARI Institute, Universit

e Libre de Bruxelles-Vrije Universiteit Brussel, Brussels, Belgium

Keywords:

Urban Trafﬁc, Simulation Models, Road Deviation Planning, Decision Making, Digital Twin.

Abstract:

Urban trafﬁc management faces growing challenges in evaluating and mitigating the impact of disruptive

events, such as road closures, on vehicular trafﬁc ﬂow. This paper presents the design and development of

an interactive tool to deﬁne and assess the impact of road deviation plans on vehicular trafﬁc. The proposed

tool targets trafﬁc management experts and is expected to support them in deﬁning and comparing alternative

solutions to mitigate disruptive events (e.g. road/tunnel closures for maintenance). The proposed tool, called

TrafﬁcTwin, can be adapted to different areas of the town, make use of different trafﬁc models (either synthetic

or calibrated) and visualize several quantitative statistics to assess and compare alternative deviation plans.

We evaluate the proposed tool using a synthetic trafﬁc model and assess the pertinence of the simulation tool

to support the decision-making process in transportation infrastructure management.

1 INTRODUCTION

A challenge in the trafﬁc domain is to provide man-

agement experts and decision-makers with methods

to assess policies to optimize trafﬁc and avoid con-

gestion, a major responsible of CO

emission (Hus-

sain et al., 2023). On the one hand, the impact of

the control strategies on road infrastructures is not ap-

parent until they are deployed in real-world motor-

way applications (Ku

c et al., 2023a). On the other

hand, testing in real-life settings is costly, risky, and

often unfeasible (Argota S

anchez-Vaquerizo, 2021).

In this context, urban trafﬁc simulation models have

become an indispensable asset: these tools provide a

lens through which it is possible to analyze control

strategies in silico. They rely on computational mod-

els to test these strategies before deploying them in

the real world. One key requirement for trafﬁc mod-

els is realism. If the dynamics of simulated trafﬁc

are close to reality, then trafﬁc management experts

can assess the trafﬁc resulting from the deployment of

https://orcid.org/0000-0002-6865-1833

https://orcid.org/0000-0002-0943-883X

https://orcid.org/0000-0002-2380-8630

https://orcid.org/0000-0001-8621-316X

some control strategy in the real world (Siebke et al.,

2023).

Research on what-if scenario modeling is gener-

ating increasing interest within the scientiﬁc commu-

nity. The growing complexity of decision-making

tasks requires the implementation of strategies that

enable the evaluation of the impact of various choices.

This is particularly crucial in trafﬁc management,

where congestion can be signiﬁcantly affected by a

single decision, especially in the case of disruptive

events. Therefore, it is essential to have efﬁcient com-

putational methods that help predict and mitigate con-

gestion issues.

c et al. (Ku

c et al., 2023b) propose a digital

twin of the Geneva motorway, integrating real-time

trafﬁc data from motorway trafﬁc counters. These

data include vehicle counts, speeds, and categories,

updated every minute. The data are provided as input

to the SUMO simulator to provide accurate simula-

tion of trafﬁc conditions. The authors use SUMO’s

calibrator objects, which adjust trafﬁc demand in the

simulation in real time according to the input trafﬁc

data. The proposed model allows for evaluating traf-

ﬁc control strategies, such as variable speed limits, in

a cost-effective, real-time environment. The authors

Guastella, D. A., Silva-Muñoz, M., Montero-Porras, E. and Bontempi, G.

A Simulation Tool to Assess the Impact of Deviation Plans on Disruptive Events of Urban Trafﬁc.

DOI: 10.5220/0013518900003970

In Proceedings of the 15th International Conference on Simulation and Modeling Methodologies, Technologies and Applications (SIMULTECH 2025), pages 51-61

ISBN: 978-989-758-759-7; ISSN: 2184-2841

measure the GEH statistics (below 5) and compare the

real and simulated trafﬁc ﬂow to evaluate the accuracy

of the digital twin.

Naing et al. (Naing et al., 2021) propose a dy-

namic framework for short-term microscopic trafﬁc

forecasting using Long Short-Term Memory (LSTM)

neural networks. By integrating real-time data into

a trafﬁc simulation, the system dynamically updates

the LSTM parameters, enabling more accurate pre-

dictions of vehicle behavior, especially in “Just-in-

Time” decision-making contexts. To perform what-if

analysis, multiple simulations are branched from the

base simulation, each one exploring different scenar-

ios based on varying trafﬁc conditions or decisions.

The authors evaluate the proposed framework on real-

world trafﬁc data from US Highway 101, demonstrat-

ing that the dynamic LSTM-based model outperforms

traditional models such as the Krauss car-following

model, in terms of accuracy and efﬁciency.

However, interactive simulation tools speciﬁcally

designed for what-if analysis of disruptive events are

limited. Adreani et al. (Adreani et al., 2022) introduce

a novel scenario model and editor, integrated into

the open-source Snap4City.org platform, to deﬁne

general processes and what-if scenarios for analysis

across different domains. The authors present a case

study on trafﬁc ﬂow reconstruction in the city of Flo-

rence, Italy, using the scenario editor and data-driven

tools to analyze and reconstruct trafﬁc patterns in the

area of interest. This enables the performance of

what-if analysis, aiding in informed decision-making.

The trafﬁc ﬂow reconstruction relies on solving a ﬂuid

dynamic problem formulated using partial differen-

tial equations (PDEs). PDE models often assume ho-

mogeneity in trafﬁc ﬂow, neglecting real-world varia-

tions such as driver behavior, heterogeneity in vehicle

dynamics, or localized phenomena (e.g., sudden lane

changes or road closures). Moreover, authors assume

the availability of data of trafﬁc ﬂow entering/exiting

in the area of interest, which is not always possible.

Additionally, a major limitation of the current

state-of-the-art is the lack of tools to quantitatively

and visually assess the impact of mitigation strategies.

An example is the evaluation of road deviation plans,

whose effects of deployment are often unknown. We

deﬁne a deviation plan as a set of rules that dictate

how trafﬁc is redirected to avoid congestion. In a real-

world environment, this corresponds to strategic road

closures or the placement of road signals.

In this paper, we present a novel tool (called

TrafﬁcTwin) for assessing the impact of alternative

road deviation plans on vehicular trafﬁc. The main

contributions of this work are: (i) an interactive tool

to model disrupting events, such as road closures, and

the related deviation plans to divert trafﬁc and avoid

closed roads, (ii) a simulation-based method to eval-

uate deviation plans with different trafﬁc demands,

synthetic or calibrated, (iii) an interactive dashboard

to visualize the differences between scenarios. We

use the SUMO package to simulate realistic trafﬁc

patterns. The proposed tool targets experts in trafﬁc

management to deﬁne easily simulation models and

perform what-if analysis in case of disrupting events

that cause road closures. We introduce the features of

the proposed tool by showcasing a realistic case study

in the city of Brussels.

2 THE TrafﬁcTwin

ARCHITECTURE

TrafﬁcTwin is composed of the following compo-

nents, as shown in Figure 1:

• the data sources, currently supporting static ﬁles

containing the trafﬁc demand deﬁnition (when us-

ing calibrated trafﬁc), and the deﬁnition of devia-

tion plans;

• the server component, responsible for the simula-

tion of urban trafﬁc (using SUMO), data aggrega-

tion and management;

• the message broker, based on RabbitMQ (Rab-

bitMQ, 2024), for handling the data exchange be-

tween client and server;

• the client component, allowing trafﬁc manage-

ment expert to deﬁne virtual trafﬁc, road closures

and deviation plans;

• the dashboard component, allowing presenting

the result of modiﬁcation in urban infrastructures

through graphical plots.

.csv

Data sources Server Message broker Client UI

Figure 1: Main components of TrafﬁcTwin. The sys-

tem consists of four components (from left to right): data

sources, a server that simulates urban trafﬁc using SUMO

and manages data, a message broker (RabbitMQ) facilitat-

ing data exchange and a client interface for trafﬁc experts to

deﬁne virtual scenarios.

In the following, we describe the core components

of the proposed tool.

SIMULTECH 2025 - 15th International Conference on Simulation and Modeling Methodologies, Technologies and Applications

2.1 Client Component (User Interface)

The graphical user interface allows three main func-

tionalities: (i) the deﬁnition of deviation plans, each

including the closed roads and the speciﬁcation of

how the vehicles must be diverted to avoid closed

roads, (ii) trafﬁc modeling, where the user can specify

the level of congestion in simulated trafﬁc, and (iii)

dashboard, where it is possible to analyze the impact

of roadworks in trafﬁc congestion. In the following,

we describe the functionalities of each part of the user

interface.

Deﬁnition of Deviation Plans

The panel for managing disrupting events (Figure 2)

enables the deﬁnition of road closures and deviation

plans. Herein, a deviation plan consists of a set of

roads where vehicles cannot drive, a time interval

deﬁning the temporal horizon of the plan, a source

road, and a set of destination roads to which vehicles

are redirected during road closures.

Figure 2: Roadworks planning in the proposed tool. On the

left a map is shown, with roads represented as blue lines.

Also, yellow lines represent closed roads, green lines the

source roads, and red lines represent the destination roads.

Simulated vehicles cannot enter the roads highlighted in

gray. In the sidebar to the right, a “Deviation Plan Deﬁni-

tion” is set, where users can close speciﬁc roads, deﬁne time

intervals for closures, and specify rerouting plans (source

and destination roads).

We implement the deviation plans in SUMO by

properly conﬁguring rerouters. The rerouters allows

vehicles’ routes to be modiﬁed as they travel over spe-

ciﬁc roads (the source roads in the deviation plan),

based on the available options and the current trafﬁc

conditions (Lopez et al., 2018). When a new desti-

nation is selected, the fastest route is calculated from

the vehicles’ current position to the new destination

(see red line in Figure 2). However, it is not always

possible to determine a new alternative route. This

is mainly due to the conditions of the road network.

In this case, the rerouter has no effect and the vehi-

cle’s original route is maintained. If the edge ahead

is closed due to roadwork, the vehicle waits until the

edge is reopened, leading to congestion. However,

after a predeﬁned time (in SUMO), the vehicle is re-

moved from the simulation.

The following XML code shows the deﬁnition of

a rerouter for use with SUMO.

<rerouter id="example" edges="-19 -17 -25

9">

<closingReroute id="-9" disallow="

all"/>

</interval>

</rerouter>

Using the previously deﬁned rerouter, vehicular

trafﬁc is prohibited on the edge with ID -9 and redi-

rected to the adjacent edges with IDs -17, -25, and 9,

respectively. The closure of edge -9 occurs between

time instants 0 and 3600 seconds. The TrafﬁcTwin

software allows to deﬁne deviation plans through an

easy-to-use interface. In this way, it is not neces-

sary for non-expert users to have speciﬁc knowledge

on how to encode rerouters for simulating deviation

plans.

The TrafﬁcTwin user interface (Figure 3) allows

conﬁguring the trafﬁc demand used by SUMO to

replicate the realistic dynamics of trafﬁc.

Figure 3: Trafﬁc simulation modeling interface in the pro-

posed tool. The tool displays on the right the road network

(blue lines), the closed roads (in yellow), the source roads

(in green), the destination roads (in red). The sidebar on

the right contains settings for conﬁguring the trafﬁc model,

including options for synthetic trafﬁc demand, calibrated

models, or custom trafﬁc demand inputs. Users can spec-

ify a time range for the simulation, set the data resolution,

and adjust the congestion level (low, medium, heavy, or cus-

tom).

Currently, the proposed tool supports the deﬁni-

tion of the synthetic trafﬁc demand (specifying the

number of vehicles per hour), and the use of trafﬁc

A Simulation Tool to Assess the Impact of Deviation Plans on Disruptive Events of Urban Trafﬁc

models generated from external calibration tools.

When using a calibrated model, the trafﬁc demand

is generated from realistic trafﬁc data. The trafﬁc de-

mand consists of a set of vehicles to be inserted into

the simulation, with each vehicle deﬁned by a time

instant (when it enters the simulation) and the set of

roads it drives on to reach its destination (Guastella

et al., 2025).

When deﬁning synthetic trafﬁc demand, Traf-

ﬁcTwin allows the user to specify the number of ve-

hicles per hour to be included in the simulation. In

the resulting trafﬁc demand, each vehicle is assigned

a random origin and destination, and its route is de-

termined using the Dijkstra algorithm (Utomo et al.,

2023).

Dashboard for Simulation Analysis

TrafﬁcTwin includes an interactive dashboard for

conducting a comparative analysis of simulation out-

comes. The dashboard visualizes different trafﬁc at-

tributes obtained from two simulations: one in which

deviation plan conﬁgurations are active and another

where they are not. By comparing the results of the

two simulations, trafﬁc management experts can as-

sess the impact of roadworks on trafﬁc congestion

without any technical knowledge about the function-

ing of the simulation tool. Figure 4 shows the dash-

board, developed as a web application using Dash

Figure 4: Interface for analyzing the simulated trafﬁc with

and without roadworks.

Table 1 lists the attributes that can be analyzed

with the dashboard.

2.2 Server Component

The server component is responsible for performing

trafﬁc simulations by using the open-source simulator

https://dash.plotly.com. Last visited: February 6, 2025

Table 1: Trafﬁc attributes that can be analyzed with the

dashboard available in TrafﬁcTwin.

Attribute (unit) Description

Edge Occupancy (%)

Occupancy of the edge/lane in %. A value of

100 would indicate vehicles standing bumper

to bumper on the whole edge (minGap=0).

Edge Density (#veh/km) Vehicle density on the edge

Waiting Time (s)

The total number of seconds vehicles were

considered halting (speed <speedThreshold).

Summed up over all vehicles

Average Speed (m/s)

The mean speed on the edge/lane within the

reported interval.

Time Loss (s)

The total number of seconds vehicles lost due

to driving slower than desired (summed up

over all vehicles)

Travel Time (s)

Time needed to pass the edge/lane.

Sampled Seconds (s)

The number of vehicles that are present on the

edge/lane in each second summed up over the

measurement interval (may be subseconds if a

vehicle enters/leaves the edge/lane).

SUMO. This tool allows emulating the complex dy-

namics of trafﬁc, including stop-and-go patterns, traf-

ﬁc lights, and junctions. In SUMO, the road network

is modeled as a graph, where the edges represent the

roads and the vertices represent the junctions. Vehi-

cles in SUMO are deﬁned by an identiﬁer, departure

time, and route through the road network. A vehicle

can be described in more detail, including properties

such as departure and arrival times, the lane to use,

and maximum and minimum speeds. Also, each ve-

hicle can be associated with a class that deﬁnes its

type (for instance, private vehicles, taxis, coaches,

trucks, or bicycles). Each class has different proper-

ties that determine how vehicles move within the road

network.

We assume that the user can generate synthetic

trafﬁc or use a pre-existing (calibrated) trafﬁc model.

The former case is useful when no realistic trafﬁc

model is available, or when the user wants to deter-

mine the impact of different deviation plans under un-

known trafﬁc conditions. The latter case is useful for

evaluating deviation plans in a realistic trafﬁc condi-

tion, that is, in a simulation model where vehicles fol-

low realistic routes (similar to real trafﬁc ﬂow).

We use the randomTrips tool provided

with SUMO to generate synthetic trafﬁc demand.

randomTrips generates a set of random routes (ran-

dom departure and destination points) for a given

road network. Each route is associated with at least

one vehicle. To generate realistic trafﬁc demand,

the randomTrips tool allows, through the parame-

ter --insertion-rate to specify the number of ve-

hicles per time interval. In this way, it is possible to

simulate the variation of trafﬁc during peak hours.

The randomTrips tool ﬁrst generates a set of

origins and destinations for the speciﬁed number of

vehicles, with this number provided as input. The

randomTrips tool uses the Dijkstra algorithm to ﬁnd

routes connecting the origin and destination for all ve-

SIMULTECH 2025 - 15th International Conference on Simulation and Modeling Methodologies, Technologies and Applications

hicles. The parameters used to generate random traf-

ﬁc for each simulation are the insertion rate and the

fringe value: the former refers to the number of ve-

hicles per hour to be introduced into the simulation;

the latter determines the probability that an edge at

the boundary of the modeled road network is selected

as the origin or destination of a vehicle. This allows

modeling vehicles whose origin or destination is over

the modeled road network.

We input the following parameters to generate

random trafﬁc (Guastella et al., 2023):

• network: the road network deﬁnition;

• begin: begin time of the simulation;

• end: end time of the simulation;

• insertion rate: the expected number of vehicles

per hour. By varying this parameter, it is possible

to control the amount of trafﬁc in the simulation.

This is modeled in the “Trafﬁc Modeling” inter-

face.

The output of randomTrips is a list of vehicles,

each associated with a route. The set of vehicles is in

XML format. The calibrated trafﬁc models supported

by TrafﬁcTwin are in the same XML format, contain-

ing a list of vehicles associated with the starting time

and an edge list that constitutes their route. This can

be generated starting from real trafﬁc count data and

using standard tool such as RouteSampler (Behrisch

and Hartwig, 2022), included with SUMO.

2.3 Communication Between Client and

Server

We use RabbitMQ (RabbitMQ, 2024), an open-

source message broker that acts as an intermediary

between the client (user interface) and the server

through a message queuing technique. Message queu-

ing is a method of communication between applica-

tions or components. With message queueing, client

and server can remain independent while processing

their individual tasks. Messages are typically small

requests, responses, status updates, or simple pieces

of information. A message queue temporarily stores

these messages, allowing applications to send and re-

ceive them as needed. One advantage of this approach

is that systems are loosely coupled-they do not need

to know the physical location of other components;

a simple name is sufﬁcient to reach them. This en-

ables systems to evolve independently without affect-

ing one another, as the reliable delivery of messages

is managed by the broker.

RabbitMQ uses an Erlang-based implementa-

tion of the Advanced Message Queuing Protocol

(AMQP), an open standard protocol that deﬁnes how

a system can exchange messages (RabbitMQ, 2024).

The protocol deﬁnes a set of rules that the communi-

cating system must follow. It also deﬁnes the repre-

sentation of messages and commands exchanged be-

tween several systems.

In the proposed tool, we use two message queues

to facilitate communication between the client and the

server. One queue is used to transfer requests from

the client to the server, mainly involving the execution

of simulations and uploading OpenStreetMap (OSM)

ﬁles. Communications from the server to the client

transfer the output of the simulations.

The use of message queues enables the conﬁgu-

ration and subsequent execution of numerous simula-

tions on the server.

3 EXPERIMENTAL RESULTS

This section presents a case study where TrafﬁcTwin

is used to assess alternative deviation plans designed

to mitigate congestion when different roads are closed

due to roadworks (or other events). The case study fo-

cuses on the area of the Ixelles municipality in Brus-

sels, Belgium.

We begin by deﬁning the road network, model-

ing the roadworks (by closing roads to vehicular traf-

ﬁc), and establishing the deviation plans and trafﬁc

demand (Section 3.1). We deﬁne two conﬁgurations,

each containing different deviation plans for the same

set of roadworks (closed roads). The conﬁgurations of

deviation plans differ in the choice of possible edges

where vehicles can transit in order to deviate their tra-

jectory to avoid closed edges. Let us consider the two

deviation plans in Figure 5. In Figure 5a we deﬁne

two possible alternatives for diverting trafﬁc. In De-

viation Plan 1, vehicles can go left, right, or straight.

However, allowing vehicles to turn left could lead to

congestion on minor streets near the roadwork site.

For this reason, we have deﬁned a second Deviation

Plan, where vehicles can only go straight or turn right.

Similarly to the previous case, the deviation plan 2 in

Figure 5b prevents vehicles from entering a partially

closed road. Deviation plans can be set up to model

various situations requiring trafﬁc diversion, such as

tunnel closures, roadworks, accidents, or strikes.

In Section 3.2, we verify the effectiveness of the

ﬁrst conﬁguration by comparing the trafﬁc statistics

with and without roadworks. In Section 3.3, we com-

pare the ﬁrst conﬁguration to the alternative one to

determine if a different deviation plan would be more

effective in reducing the impact of roadworks on ve-

hicular trafﬁc (e.g. in terms of congestion).

A Simulation Tool to Assess the Impact of Deviation Plans on Disruptive Events of Urban Trafﬁc

Closed road

Deviation Plan 1

Deviation Plan 2

(a)

Closed road

Deviation Plan 1

Deviation Plan 2

(b)

Figure 5: Two alternative deviation plans for the roadworks

(indicated as white and red rectangles) in Rue Malibran

(Figure 5a) and Rue Lesbroussarts (Figure 5b). In the main

deviation plan, vehicles can follow the directions indicated

by both red and blue arrows, whereas in the alternative de-

viation plan vehicles follow only the direction indicated by

blue arrows.

3.1 Scenario Modeling

We use OSM to extract the road network in the con-

cerned area and load it in TrafﬁcTwin. Figure 6 shows

a roadwork plan provided by the municipality of Ix-

elles. Figure 7 shows the location of the roads (around

Rue Scarron) where we simulate the presence of road-

works.

Figure 6: Detailed roadwork plan for Rue Scarron (Ixelles,

Brussels, Belgium), outlining the restricted areas during

construction activities. This is provided by Brussels Mobil-

ity, the government agency responsible for managing and

developing transport infrastructure in the Brussels-Capital

Region of Belgium.

Chaussée

d’Ixelles

Rue du

Viaduc

Rue Malibran

Rue Scarron

Rue

Lesbroussarts

Figure 7: Location of the roadworks in the area of Ixelles

(Belgium), implemented using the proposed tool.

To assess the impact of the deviation plan on ve-

hicular trafﬁc, we present the results of two scenar-

ios: one with roadworks, considering a conﬁguration

of deviation plans (Figure 7), one without roadworks.

Then, we evaluate an alternative conﬁguration of de-

viation plans to assess if this allows for minimizing

the congestion.

We consider a 1-hour trafﬁc simulation and we

generate random trafﬁc using the randomTrips tool

available with SUMO. By considering random traf-

ﬁc, we can evaluate the effects of road closures under

different hypothetical trafﬁc congestion levels.

We introduce stochasticity to better replicate real-

ity in simulation scenarios by varying i) the departure

times of vehicles (uniform distribution over [1, 60]

seconds) and ii) the edge weights so that the Dijkstra

algorithm returns different routes for the same origin-

destination pair. Weight perturbation allows us to take

into account drivers’ individual preferences.

SIMULTECH 2025 - 15th International Conference on Simulation and Modeling Methodologies, Technologies and Applications

This stochastic approach results in different con-

gestion outcomes with each simulation run. To cap-

ture this variability, we perform 100 simulation for

each trafﬁc congestion level.

Finally, we analyze the simulation output. We be-

gin by edge-wise measurements returned by SUMO,

then we analyze aggregated information on individual

vehicle trips. The ﬁrst analysis aims to show that, on

average, the congestion level increases when multiple

roadworks are present. The second analysis aims to

show that, on average, vehicles take longer to reach

their destinations when roadworks are in place. This

information is expected to support trafﬁc management

experts in assessing the impact of deviation plans in a

virtual setting before real-world deployment.

3.2 Deviation Plan Assessment

(With/Without Roadworks)

This Section presents the assessment of the ﬁrst devi-

ation plan. Figure 8 shows the Pareto chart (Wilkin-

son, 2006) clustering the edges according to the av-

erage density (in #veh/km), considering the simula-

tions with and without roadworks. The line represents

the cumulative percentage of the average density. The

density measures the average spacing of vehicles per

km, obtained by dividing the number of vehicles by

the road segment length.

For 500 and 2000 vehicles, there’s no remarkable

difference in average density after road closures. Con-

trarily, when 5000 vehicles are present, the average

density is lower compared to scenarios without road

closures. Although this outcome may seem counterin-

tuitive, it can be explained by the elimination of cer-

tain bottlenecks at intersections that typically cause

congestion. Additionally, deviation plans may reroute

trafﬁc away from areas affected by roadworks. When

vehicles are diverted, their destinations may change

as a result of the deviation plans, further contributing

to a reduction in vehicle density within the road net-

work.

Figure 9 shows the Pareto chart clustering the

edges by average speed (in m/s), from simulations

with and without roadworks. For scenarios with 500

and 2000 vehicles, the absolute average speed dif-

ference obtained by the simulations with and with-

out roadworks is not signiﬁcant. However, when con-

sidering 5000 vehicles per hour, the average speed is

higher in the case of roadworks. As for the trafﬁc

density measure, this can be caused by the elimina-

tion of certain bottlenecks at intersections that typi-

cally cause congestion.

Figure 10 shows the road network, with edges col-

ored based on the absolute difference in average speed

[1.0, 39.0)

[39.0, 76.9)

[76.9, 114.9)

[114.9, 152.8)

[152.8, 190.8)

[190.8, 228.7)

100

200

300

400

500

600

80%

90%

100%

With Roadworks

Without Roadworks

Density

Number of Edges

Cumulative Percentage

(a) 500 vehicles

[4.5, 61.0)

[61.0, 117.5)

[117.5, 174.0)

[174.0, 230.5)

[230.5, 287.0)

[287.0, 343.5)

100

200

300

400

500

80%

90%

100%

With Roadworks

Without Roadworks

Density

Number of Edges

Cumulative Percentage

(b) 2000 vehicles

[66.0, 113.7)

[113.7, 161.4)

[161.4, 209.1)

[209.1, 256.8)

[256.8, 304.6)

[304.6, 352.3)

100

200

300

400

500

40%

50%

60%

70%

80%

90%

100%

With Roadworks

Without Roadworks

Density

Number of Edges

Cumulative Percentage

Figure 8: Average density of vehicles per edge, measured

in #veh/km, using 500 (Figure 8a), 2000 (Figure 8b), and

5000 vehicles (Figure 8c). Lower values are better.

between simulations with and without roadworks, in

the 5000 vehicles scenario.

A positive speed difference indicates that the road-

works have a major impact in congestion, as in this

case the average speed is lower. Contrarily, a negative

difference indicates a higher speed in the simulation

when roadworks are applied. This can be attributed

to the conﬁguration of deviation plans, which allow

for avoiding congested areas of the environment, re-

ducing average trafﬁc and resulting in higher vehicles

speed. We also observe a greater difference in bound-

ary roads, particularly in the northern part of the en-

A Simulation Tool to Assess the Impact of Deviation Plans on Disruptive Events of Urban Trafﬁc

[7.5, 8.3) [8.3, 9.2) [9.2, 10.0) [10.0, 10.9) [10.9, 11.7) [11.7, 12.6)

100

200

300

400

500

80%

With Roadworks

Without Roadworks

Speed (m/s)

Number of Edges

Cumulative Percentage

(a) 500 vehicles

[7.3, 8.2) [8.2, 9.0) [9.0, 9.9) [9.9, 10.7) [10.7, 11.6) [11.6, 12.4)

100

200

300

400

500

80%

90%

With Roadworks

Without Roadworks

Speed (m/s)

Number of Edges

Cumulative Percentage

(b) 2000 vehicles

[4.9, 6.1) [6.1, 7.3) [7.3, 8.5) [8.5, 9.7) [9.7, 10.9) [10.9, 12.0)

100

150

200

250

20%

30%

40%

50%

60%

70%

80%

90%

100%

With Roadworks

Without Roadworks

Speed (m/s)

Number of Edges

Cumulative Percentage

Figure 9: Average speed of vehicles (m/s) in the simulations

with and without roadworks, using 500 (Figure 9a), 2000

(Figure 9b), and 5000 vehicles (Figure 9c). Higher values

are better.

vironment. Although this area is not directly affected

by any roadworks, this highlights how simulation can

reveal emerging trafﬁc dynamics that may not be eas-

ily identiﬁed through analytical methods.

The results presented so far refer to network

edges. In what follows, we present individual trip

statistics.

Figure 11 compares the average duration of trips

(shorter trips are obviously preferable) with and with-

out roadworks. This is the time required by vehicles

to go from origin to destination.

Chaussée

d’Ixelles

Rue du

Viaduc

Rue Malibran

Rue Scarron

Rue

Lesbroussarts

Figure 10: Average absolute speed difference obtained by

the simulation with and without the roadworks.

Without roadworks, trip duration is generally

shorter. When roadworks are present, trips take more

time as vehicles face delays and detours due to road

closures. The comparison highlights the signiﬁcant

impact roadworks have on travel time, causing con-

gestion and longer trips.

Figure 12 compares the average trip lengths with

and without roadworks. Vehicles adjust their routes

when they encounter rerouters, placed on roads near

the closed ones. As expected, this leads to longer

paths to bypass the closures and reach the destination.

However, not all destinations are reachable due to

road closures, primarily because of the limited scope

of the considered environment. For this reason, it is

possible to include in the simulator a probability for

drivers to change the destination (in our experiments

this value was set to 0.5). Note that if a vehicle does

not change its destination and the deviation plan does

not allow ﬁnding an alternative route (because both

road network conﬁguration and deviation plan), the

vehicle remains stationary in front of the closed road

until it is reopened. In this case, SUMO teleports ve-

hicles that wait too long to avoid grid-locks.

3.3 Alternative Deviation Plans

Assessment

This Section focuses on the comparison of two al-

ternative deviation plans. Figure 13 compares the

respective average edge density and vehicle speed.

This result is obtained as the average of each attribute

after 100 simulations, using the same set of routes.

For both conﬁgurations, we show the results obtained

from 1 hour of simulation using 5000 vehicles per

hour.

The density plot does not reveal signiﬁcant differ-

ences, while the speed comparison indicates that the

average speed is higher when using the ﬁrst deviation

SIMULTECH 2025 - 15th International Conference on Simulation and Modeling Methodologies, Technologies and Applications

[1.0, 250.1)

[250.1, 499.3)

[499.3, 748.4)

[748.4, 997.6)

[997.6, 1246.7)

[1246.7, 1495.9)

100

120

10%

20%

30%

40%

50%

60%

70%

80%

90%

100%

With Roadworks

Without Roadworks

Duration (s)

Number of Trips

Cumulative Percentage

(a) 500 vehicles

[1.0, 804.3)

[804.3, 1607.6)

[1607.6, 2410.9)

[2410.9, 3214.1)

[3214.1, 4017.4)

[4017.4, 4820.7)

200

400

600

800

60%

70%

80%

90%

100%

With Roadworks

Without Roadworks

Duration (s)

Number of Trips

Cumulative Percentage

(b) 2000 vehicles

[1.0, 1006.9)

[1006.9, 2012.7)

[2012.7, 3018.6)

[3018.6, 4024.4)

[4024.4, 5030.3)

[5030.3, 6036.1)

200

400

600

800

1000

1200

1400

30%

40%

50%

60%

70%

80%

90%

100%

With Roadworks

Without Roadworks

Duration (s)

Number of Trips

Cumulative Percentage

Figure 11: Average duration (in seconds) of the trips, us-

ing 500 (Figure 11a), 2000 (Figure 11b), and 5000 vehicles

(Figure 11c). Lower values are better.

plans. This suggests that the ﬁrst plan is more effec-

tive in reducing congestion. The same analysis in-

dicates that, in some cases, the second conﬁguration

outperforms the ﬁrst, notably in the range [7, 8.3]. By

examining the locations of the edges where the aver-

age vehicle speed falls within this range (Figure 14),

it appears that most edges are near the roadwork sites.

This suggests that using the second plan would result

in higher average speeds on edges in the proximity of

the roadworks.

[0.0, 883.7)

[883.7, 1767.5)

[1767.5, 2651.2)

[2651.2, 3535.0)

[3535.0, 4418.7)

[4418.7, 5302.5)

10%

20%

30%

40%

50%

60%

70%

80%

90%

100%

With Roadworks

Without Roadworks

Route Length (m)

Number of Trips

Cumulative Percentage

(a) 500 vehicles

[0.0, 923.2)

[923.2, 1846.4)

[1846.4, 2769.6)

[2769.6, 3692.8)

[3692.8, 4616.0)

[4616.0, 5539.2)

100

200

300

10%

20%

30%

40%

50%

60%

70%

80%

90%

100%

With Roadworks

Without Roadworks

Route Length (m)

Number of Trips

Cumulative Percentage

(b) 2000 vehicles

[0.0, 918.2)

[918.2, 1836.5)

[1836.5, 2754.7)

[2754.7, 3673.0)

[3673.0, 4591.2)

[4591.2, 5509.5)

200

400

600

800

1000

30%

40%

50%

60%

70%

80%

90%

100%

With Roadworks

Without Roadworks

Route Length (m)

Number of Trips

Cumulative Percentage

Figure 12: Average length (in meters) of the routes, us-

ing 500 (Figure 12a), 2000 (Figure 12b), and 5000 vehicles

(Figure 12c). Lower values are better.

In some cases, congestion is caused by vehicles

that cannot reach their destinations due to road clo-

sures. Suppose a road is closed, and this road is re-

quired for a vehicle to reach its destination. If no

alternative route is available, the vehicle will remain

stationary in front of the closed edge, waiting for it

to reopen. This situation leads to congestion, which

the simulator resolves by teleporting the vehicle af-

ter a predeﬁned time. A teleported vehicle is removed

from the network. It is then moved along its route, but

no longer being on the street, and reinserted into the

A Simulation Tool to Assess the Impact of Deviation Plans on Disruptive Events of Urban Trafﬁc

[106.8, 148.7)

[148.7, 190.6)

[190.6, 232.5)

[232.5, 274.3)

[274.3, 316.2)

[316.2, 358.1)

100

200

300

400

500

90%

100%

Deviation Plan 1

Deviation Plan 2

Density

Number of Edges

Cumulative Percentage

(a) Average edge density (#veh/km). Lower values are bet-

ter.

[4.5, 5.8) [5.8, 7.0) [7.0, 8.3) [8.3, 9.5) [9.5, 10.8) [10.8, 12.0)

100

150

200

250

20%

30%

40%

50%

60%

70%

80%

90%

100%

Deviation Plan 1

Deviation Plan 2

Speed (m/s)

Number of Edges

Cumulative Percentage

(b) Average speed (m/s). Higher values are better.

Figure 13: Average edge density (Figure 13a), speed (Fig-

ure 13b) obtained by the comparison of two alternative de-

viation plans.

Chaussée

d’Ixelles

Rue du

Viaduc

Rue Malibran

Rue Scarron

Rue

Lesbroussarts

speed

>5.8&<7

Figure 14: Location of the edges which average speed falls

in the range [7, 8.3] using the two deviation plan conﬁgura-

tions.

network as soon as this becomes possible

. To solve

this issue, a rerouted vehicle has a 50% probability of

changing its destination if it cannot reach its original

destination road.

We evaluate the number of teleported vehicles as

https://sumo.dlr.de/docs/Simulation/Why Vehicles

are teleporting.html. Last visited: February 6, 2025

a congestion metric to compare the effectiveness of

two deviation plan conﬁgurations by quantifying (1)

the number of vehicles that cannot reach their desti-

nation due to a closed road and (2) the number of ve-

hicles teleported due to congestion phenomena aris-

ing from the closed road and the deviation plan con-

ﬁguration. Table 2 compares the average number of

teleported vehicles using both deviation plan conﬁgu-

rations. Each value is calculated as the average num-

ber of vehicles teleported in the 100 simulations per-

formed for each conﬁguration.

Table 2: Average number of teleported vehicles using the

two deviation plan conﬁgurations. The lower the average

value, the more vehicles ﬁnd an alternative route to reach

their destination without being teleported.

Deviation Plan 1 Deviation Plan 2

Average Std Dev Average Std Dev

500 Vehicles 0.68 0.004 0.61 0.008

2000 Vehicles 0.74 0.039 0.43 0.063

5000 Vehicles 6.4 0.58 6.13 0.59

From the results shown in Table 2, we can observe

that, on average, the number of teleported vehicles in

both deviation plan conﬁguration is low. This indi-

cates that the majority of vehicles are able to ﬁnd an

alternative route to reach their destination.

4 CONCLUSION AND FUTURE

WORKS

This paper introduces TrafﬁcTwin, a tool to assist traf-

ﬁc management experts in decision-making tasks, by

enabling the design and assessment of alternative de-

viation plans.

The tool is based on a client-server architecture.

The server component is responsible for simulat-

ing trafﬁc, while the client component enables traf-

ﬁc management experts to deﬁne deviation plans and

trafﬁc demand through a graphical user interface. We

also developed a dashboard to visualize plots that can

be used by trafﬁc management experts to evaluate the

impact of disruptive events.

The TrafﬁcTwin software is based on the SUMO

package. This simulator software typically has a steep

learning curve, which poses a barrier to non-expert

users. TrafﬁcTwin enables researchers, urban plan-

ners, and decision-makers to use SUMO’s simula-

tion capabilities without needing extensive technical

knowledge. This positions TrafﬁcTwin as a key tool

for making SUMO accessible to non-expert users.

TrafﬁcTwin should be intended as an enabler for

a more general digital twin solution to assist trafﬁc

ofﬁcers in decision-making. The challenges still to

SIMULTECH 2025 - 15th International Conference on Simulation and Modeling Methodologies, Technologies and Applications

address are: (i) continuously calibrating trafﬁc us-

ing real-time count data, and (ii) employing machine

learning methods to suggest the best deviation plans

that minimize the impact of roadworks. For challenge

(i), the key difﬁculty lies in balancing off-line and

online (simulation-based) methods for trafﬁc calibra-

tion using real-world data. Although online methods

are typically more accurate due to their reliance on

realistic vehicle dynamics, they are computationally

intensive, making the development of trafﬁc digital

twins more challenging. For the (ii), the aim is to use

machine learning techniques to extract insights from

various trafﬁc models. By analyzing trafﬁc dynamics

over time, these methods can provide recommenda-

tions on where disruptive events are likely to have the

greatest impact, potentially causing signiﬁcant con-

gestion effects.

Furthermore, we will investigate the integration of

our tool with Unity 3D to provide a more realistic

view of the urban environment. This will allow traf-

ﬁc management experts to evaluate control policies

within a highly realistic digital twin before deploying

them in the real world. This virtual environment can

also be used by citizens to understand the impact of

these policies on trafﬁc, thereby promoting alternative

forms of mobility to private vehicles.

ACKNOWLEDGEMENTS

This research work is being funded by

Paradigm.Brussels. This project was supported

by the FARI - AI for the Common Good Institute

(ULB-VUB), ﬁnanced by the European Union,

with the support of the Brussels Capital Region in

Belgium. (Innoviris and Paradigm). G. Bontempi

is also supported by the Service Public de Wallonie

Recherche under grant nr 2010235–ARIAC by

DigitalWallonia4.ai. Part of this research work is

being developed in the context of TORRES (Trafﬁc

prOcessing foR uRban EnvironmentS), a Joint R&D

Project (2022-RDIR-59b) funded by “R

egion de

Bruxelles-Capitale - Innoviris”.

REFERENCES

Adreani, L., Bellini, P., Bilotta, S., Bologna, D., Collini, E.,

Fanfani, M., and Nesi, P. (2022). Smart City Scenario

Editor for General What-if Analysis. Preprint; url:

doi.org/10.20944/preprints202402.1163.v1.

Argota S

anchez-Vaquerizo, J. (2021). Getting Real: The

Challenge of Building and Validating a Large-Scale

Digital Twin of Barcelona’s Trafﬁc with Empiri-

cal Data. ISPRS International Journal of Geo-

Information, 11(1):24.

Behrisch, M. and Hartwig, P. (2022). A comparison of

SUMO’s count based and countless demand genera-

tion tools. SUMO Conference Proceedings, 2:125–

131.

Guastella, D. A., Cornelis, B., and Bontempi, G. (2023).

Trafﬁc simulation with incomplete data: the case of

brussels. In Proceedings of the 1st ACM SIGSPATIAL

International Workshop on Methods for Enriched Mo-

bility Data: Emerging Issues and Ethical Perspectives

2023, EMODE ’23, page 15–24, New York, NY, USA.

Association for Computing Machinery.

Guastella, D. A., Morales-Hern

andez, A., Cornelis, B., and

Bontempi, G. (2025). Calibration of vehicular trafﬁc

simulation models by local optimization.

Hussain, Z., Kaleem Khan, M., and Xia, Z. (2023). In-

vestigating the role of green transport, environmental

taxes and expenditures in mitigating the transport CO2

emissions. Transportation Letters, 15(5):439–449.

c, K., Schumann, R., and Ivanjko, E. (2023a). A dig-

ital twin in transportation: Real-time synergy of traf-

ﬁc data streams and simulation for virtualizing motor-

way dynamics. Advanced Engineering Informatics,

55:101858.

c, K., Schumann, R., and Ivanjko, E. (2023b). A digi-

tal twin in transportation: Real-time synergy of trafﬁc

data streams and simulation for virtualizing motorway

dynamics. 55:101858.

Lopez, P., Wiessner, E., Behrisch, M., Bieker-Walz, L.,

Erdmann, J., Flotterod, Y.-P., Hilbrich, R., Lucken,

L., Rummel, J., and Wagner, P. (2018). Microscopic

trafﬁc simulation using sumo. In 2018 21st Interna-

tional Conference on Intelligent Transportation Sys-

tems (ITSC), pages 2575–2582. IEEE.

Naing, H., Cai, W., Hu, N., Wu, T., and Yu, L. (2021). Data-

driven Microscopic Trafﬁc Modelling and Simulation

using Dynamic LSTM. In Proceedings of the 2021

ACM SIGSIM Conference on Principles of Advanced

Discrete Simulation, SIGSIM-PADS ’21, page 1–12,

New York, NY, USA. Association for Computing Ma-

chinery.

RabbitMQ (2024). Rabbitmq.

Siebke, C., Mai, M., and Prokop, G. (2023). What Do

Trafﬁc Simulations Have to Provide for Virtual Road

Safety Assessment? Human Error Modeling in Traf-

ﬁc Simulations. In Proceedings of the IEEE Con-

ference on Intelligent Transportation Systems, vol-

ume 24, pages 1419–1436.

Utomo, D. D., Aurelia, M., Tanasia, S. M., Nurhasanah, and

Handoyo, A. T. (2023). Implementation of dijkstra

algorithm in vehicle routing to improve trafﬁc issues

in urban areas. In 2023 3rd International Conference

on Smart Cities, Automation & Intelligent Computing

Systems (ICON-SONICS), pages 73–78.

Wilkinson, L. (2006). Revising the Pareto Chart. The Amer-

ican Statistician, 60(4):332–334.

A Simulation Tool to Assess the Impact of Deviation Plans on Disruptive Events of Urban Trafﬁc