A Concept for Collaborative Incident Validation in a Self-organised

Trafﬁc Management System

Sven Tomforde

and Ingo Thomsen

Intelligent Systems, Christian-Albrechts-Universit

at zu Kiel, 24118 Kiel, Germany

Keywords:

Trafﬁc Management, Organic Trafﬁc Control, Trafﬁc Flow Analysis, Trafﬁc Incident Detection, Validation.

Abstract:

The strong and, in part, further increasing trafﬁc volumes of individual and heavy goods trafﬁc in urban

regions lead to a utilisation of the networks close to or above the capacity limit, especially during rush hours.

Trafﬁc light control is ideally trafﬁc-dependent, which can be realised either centralised or distributed as a

self-organised approach. However, these systems are typically not able to detect disruptions or incidents (such

as accidents, road works, etc.) and take them into account in the control logic. A key problem here is that

either there is no incident detection in place or it is not reliable enough. In this paper, we discuss the need for

collaborative validation of locally detected incidents in a self-organised trafﬁc control system. We show that

this can increase the reliability of detection to the point where incident-dependent switching becomes possible.

1 INTRODUCTION

The constant growth of individual and freight trafﬁc

is causing delays and congestion worldwide, espe-

cially in urban areas (Schrank et al., 2019), even if

recent studies showed that travel delays were heavily

reduced from 2019 to 2020 due to the COVID lock-

downs (Schrank et al., 2021). Since infrastructure ex-

pansion is typically not an option, a more intelligent

and efﬁcient trafﬁc management is pivotal. In the lit-

erature, different approaches can be found, ranging

from a purely centrally organised, static to hybrid to a

fully distributed (at the level of intersections) optimi-

sation of trafﬁc-dependent clearance times.

Due to the inherent advantages such as problem

locality, scalability, or avoidance of single-point-of-

failure, science has focused mainly on distributed ap-

proaches in the last two decades, partly with more

centralised elements. Such solutions can adapt green

times locally, automatically learn the best adapta-

tion strategy and establish self-organised coordination

with neighbouring intersection controllers (for pro-

gressive signal systems or route guidance). However,

these approaches are still reactive in the sense that

they are focused on measured and estimated trafﬁc

ﬂows – while ignoring the expected developments due

to actual incidents in the underlying network.

https://orcid.org/0000-0002-5825-8915

https://orcid.org/0000-0002-0850-4786

In preliminary work, we presented an approach to

automated incident detection (AID) in urban road net-

works in contrast to established approaches at high-

ways. Since local detection is fundamentally uncer-

tain, more global knowledge is needed to increase the

reliability. Therefore, the contribution of this paper is

(a) to address the challenges of local incident detec-

tion in urban networks, (b) to outline a collaborative,

self-organised validation of incidents to increase ac-

curacy and (c) to derive a ﬁtting research agenda.

The remainder of this paper is organised as fol-

lows: Section 2 gives an overview of related work by

outlining self-organised trafﬁc control and trafﬁc inci-

dent detection. Section 3 describes our system model

and the Organic Trafﬁc Control system as the basis

for our approach. Section 4 then shows our concept

for collaborative incident validation in self-organised

trafﬁc control systems with the Organic Trafﬁc Con-

trol system as example. Finally, Section 5 summarises

the paper and gives an outlook on future work.

2 BACKGROUND

This section describes the underlying related work

– speciﬁcally in the context of self-organised trafﬁc

control and automated incident detection.

316

Tomforde, S. and Thomsen, I.

A Concept for Collaborative Incident Validation in a Self-organised Trafﬁc Management System.

DOI: 10.5220/0011051100003191

In Proceedings of the 8th International Conference on Vehicle Technology and Intelligent Transport Systems (VEHITS 2022), pages 316-323

ISBN: 978-989-758-573-9; ISSN: 2184-495X

2.1 Self-organised Trafﬁc Control

Trafﬁc lights in urban areas are usually operated by

a trafﬁc control centre. The most prominent sys-

tems are SCOOT (Robertson and Bretherton, 1991),

SCATS (Sims and Dobinson, 1980), MOVA (Vin-

cent et al., 1990), and UTOPIA/-SPOT (Mauro and

Taranto, 1990). These systems typically rely on a

centralised control loop that adapts the behaviour of

distributed intersection controller (IC), based on a

given cost function, which may include different as-

pects, like travel times, emissions, or public trans-

port priority. Despite being centralised, these sys-

tems come with at least some self-adaptive and self-

organising (SASO), i.e., being adaptive and policy-

driven. The adaptation mechanism works on top of

a parametrisable system conﬁguration. For classiﬁca-

tion and comparison of approaches, see (Studer et al.,

2015). In addition to these popular approaches, sev-

eral systems focusing on SASO and learning capa-

bilities have been proposed: A multi-agent approach

based on fuzzy control was presented in (Gokulan and

Srinivasan, 2010), distributed W-learning was used to

optimise a phase-oriented signal control in(Dusparic

and Cahill, 2009), and (Oliveira and Camponogara,

2010) used a model with predictive control. As an

alternative to phase-based systems, a ﬂuid-dynamic

model has been discussed in (Helbing et al., 2005)

that uses waiting vehicles as pressure and counter-

pressure for switching trafﬁc lights policies. In con-

trast to the aim of this proposal, these trafﬁc control

systems do not autonomously identify and classify in-

cidents and adapt their signalisation according to de-

tected incidents.

2.2 Trafﬁc Incident Detection

Techniques for automatic recognition of incidents, ac-

cidents, and other road events, e.g. requiring emer-

gency responses, have been the focus of research for

more than three decades. Most of the resulting al-

gorithms rely on data of loop detectors. Chronolog-

ically, AID research started with the Standard Nor-

mal Deviate algorithm (Dudek et al., 1974), subse-

quently followed by the California algorithm fam-

ily (Payne, 1975; Payne and Tignor, 1978). These

techniques are essentially following a simple de-

cision tree structure considering threshold. Sub-

sequently, approaches based on time-series analy-

sis (Ahmed and Cook, 1980), identiﬁcation of low-

volume conditions (Dudek et al., 1975), ﬁltering and

smoothing-based algorithms (Stephanedes and Chas-

siakos, 1993), a dynamic-systems-model-based al-

gorithm (Willsky et al., 1980), correlation-analysis-

based approaches (Takaba and Matsuno, 1985), the

McMaster catastrophe theory-based algorithm (Gall

and Hall, 1989), and a mathematical trafﬁc-ﬂow-

model-based algorithm (Lin and Daganzo, 1997) have

been presented. More recently, video-based ap-

proaches have been presented (Shehata et al., 2008)

and combined with semantic annotations (Kamijo

et al., 2004). In addition to these infrastructure-based

approaches, probe vehicles have been considered to

estimate trafﬁc ﬂows (Jenelius and Koutsopoulos,

2013); with some work specially dedicated to urban

environments (Feng et al., 2014) – which may serve

the incident detection.

However, these approaches all come with some

limitations: Either they are designed for highways

only or they are based on experienced travel times

through the underlying road network, and/or they do

not distinguish between different incident types (and

the corresponding reaction). Most importantly, there

is no integrated trafﬁc management solution that con-

siders detected incidents, an estimation of their sever-

ity and impact, or takes this information pro-actively

into account when deciding about trafﬁc control or

progressive signal systems, for instance.

In response to these observations, we presented

a novel clustering-based approach for AID in urban

road networks that is based on standard loop detector

technology again (Thomsen et al., 2021). Based on

responsibility zones of ICs (i.e., intersection area and

incoming sections equipped with induction loop sen-

sors), the distributed ICs are considering the time se-

ries of the detector loops and apply techniques such as

DBSCAN (Ester et al., 1996) to detect incidents on-

line in a certain time window. We showed that appro-

priate detection accuracy is given for high load con-

ditions, while the approach still suffers in weak load

conditions. Current work focuses on an improvement

of the approach and subsequent classiﬁcation of in-

cidents. This should serve as a basis for predicting

properties such as duration and impact of the incident.

3 ORGANIC TRAFFIC CONTROL

Below, we consider collaborative incident detection to

improve the detection accuracy and reliability in dis-

tributed and self-organised trafﬁc control. We present

our system model with the possible incident types and

discuss the Organic Trafﬁc Control System as a basis.

3.1 System Model

We assume urban road networks with varying topol-

ogy and decentralised nodes that are responsible for

A Concept for Collaborative Incident Validation in a Self-organised Trafﬁc Management System

317

controlling the trafﬁc light controllers (TLC) of the

intersection. Each node is responsible for the area

of this controlled intersection as well as the incom-

ing sections – where induction loop sensors are as-

sumed to be available. Further, each node is capable

of detecting trafﬁc incidents in its inbound and out-

bound sections, and it communicates with its neigh-

bours (i.e., other nodes that share a road segment).

We consider ﬁve groups of possible static inci-

dents while dynamic events such as the partial closure

of a lane on a multi-lane section (e.g. heavy goods

trafﬁc) will be addressed future work. The groups are:

• Complete closing of the section between two in-

tersections – called a section closure (Fig. 1a)

• Lane closure in a multi-lane section (Fig. 1b)

• Partial lane closure in multi-lane section (Fig. 1c)

• Closure or limited use of an intersection by block-

ing one or more turnings (Fig. 1d)

• Technical defect at an intersection (e.g. loss of

function of a trafﬁc light or a detector)

At the very least, nodes have to work with the fol-

lowing information: They only know with a certain

probability that there is a possible incident. This in-

formation can also be passed on to their neighbours.

Moreover, the nodes send their data in two scenarios

– direct or with a certain delay.

Based on these model assumptions, the objective

of a collaborative validation scheme is to improve the

incident prediction of the underlying self-organised

trafﬁc control and management system based on de-

centralised communication – or to reduce its false

positive rate. To increase the conﬁdence towards the

own data, a periodic self-diagnosis of each node, the

associated TLC, and detectors is required. The com-

munication is used to specify the nature and origin of

a possible incident and to validate the disturbance by

other trafﬁc signal controllers. All nodes only have

knowledge about their own state (paired with a cer-

tain conﬁdence), while the state is either ‘everything

is okay’ or one of the following levels:

1. The node knows of an incident somewhere.

2. The node knows that there is an incident in an out-

going or incoming section.

3. The node can assign the incident to a speciﬁc sec-

tion or junction.

4. The node can assign the incident type to one of

the previous mentioned groups.

3.2 The Basis: Organic Trafﬁc Control

The Organic Trafﬁc Control (OTC) system (Proth-

mann et al., 2009) and its extensions serve as a basis

(a) Section closure

(b) Lane closure

(d) Turn closure

Figure 1: Examples of incidents simulated in the Aimsun

Next trafﬁc simulator (Aimsun, 2021) in the road section

from A to B, or in the last case, in intersection A itself,

when cars cannot turn right towards intersection B.

for automatic detection of incidents in urban road net-

works, their self-organised collaborative validation,

and ﬁnally their consideration in signalisation strate-

gies at ICs. The OTC system is a self-adaptive and

self-organised trafﬁc control system that decides lo-

cally at each intersection about the behaviour of the

underlying IC. Here, “organic” follows the ideas of

Organic Computing (M

uller-Schloer and Tomforde,

2017) and emphases the transfer of principles from

nature to technical systems: The decentralised struc-

ture, the cooperation of smaller, autonomous entities,

as well as local adaptation and learning capabilities

allow for high robustness, scalability, and ﬂexibility.

Based on the Observer/Controller paradigm (Tom-

forde et al., 2011), the OTC system adapts the green

duration of trafﬁc lights in a phase-based approach

and optimises this adaptation strategy at runtime by

means of reinforcement learning and safety-oriented

generation of novel behaviour within a simulation en-

vironment, see (Prothmann et al., 2009). This adap-

tation process is performed depending on the cur-

rently active cycle time of the trafﬁc controller, i.e.,

an adapted control strategy is active for three cycles

(typically 60 to 120 sec) before it can become sub-

ject to adaptations again. As decision basis for any

adaptations, the current trafﬁc ﬂows for all turning

movements passing the intersection (in

vehicles

hour

and es-

VEHITS 2022 - 8th International Conference on Vehicle Technology and Intelligent Transport Systems

318

timated from detector readings) are considered. The

estimated waiting times are then used as feedback to

improve the behaviour over time. OTC is further able

to establish progressive signal systems in a fully self-

organised manner (Tomforde et al., 2008) and to pro-

vide route recommendations to drivers which reﬂect

the current state of the trafﬁc network (Prothmann

et al., 2012). Based on OTC, further contributions in-

vestigated are robust trafﬁc demand prediction (Som-

mer et al., 2013), integration of these predictions in

the control strategies, and infrastructure-based antici-

patory route guidance (Sommer et al., 2016).

OTC is self-organised in a way that all nodes oper-

ate independently and collaborate to achieve system-

wide goals, such as reduction of waiting times, num-

ber of stops, emissions, etc. Therefore, it establishes

a multi-layered adaptation and learning system on top

of a standard TLC. Figure 2 illustrates the concep-

tual design. Here, Layer 0 represents the System un-

der Observation and Control (SuOC), which is the ac-

tual TLC and the interfaces to detectors and neigh-

bouring nodes. This TLC (i.e., its green duration)

is re-conﬁgurable at runtime, which is done by the

layer above. Consequently, Layer 1 adapts dynami-

cally to the state of the environment (assessed using

the sensors) and its controller which uses a Learn-

ing Classiﬁer System (LCS, here a variant of Wilson’s

XCS (Wilson, 1995)). This LCS chooses rules from

a rule set to modify the trafﬁc signalisation appropri-

ately at runtime. Finally, Layer 2 is activated when

Layer 1 is confronted with an environment for which

it has no suitable rule or only inappropriate knowl-

edge. In this case, a trafﬁc simulation software (Aim-

sun Next, see (Aimsun, 2021)) is used to validate new

rules which are generated using an evolutionary algo-

rithm.

4 COLLABORATIVE,

SELF-ORGANISED INCIDENT

VALIDATION

Based on the OTC approach and the presented system

model – including the incident detection approach

from (Thomsen et al., 2021) – this section derives

the challenges for establishing a collaborative inci-

dent validation procedure. This outlines a research

agenda in addition to the basic concept.

Trafﬁc conditions as observed by detectors are the

result of vehicles traversing the network. In conse-

quence, the patterns observed at consecutive detec-

tor stations are not independent. Individual vehicles

may temporarily be delayed due to parking (includ-

Figure 2: Overview of the multilevel OTC architecture.

ing long-term stays in parking lots) – but they do not

completely appear or disappear. This also implies that

the observed trafﬁc conditions are strongly related to

those of the preceding detector station a short period

before. We consequently, aim at utilising this infor-

mation to (i) further decrease the false alarm rate and

(ii) detect disturbed sensors.

The goal of the collaborative self-organised inci-

dent validation approach is to turn a self-adaptive in-

cident detection system (here: The OTC-based ap-

proach from (Thomsen et al., 2021)) into a cooper-

ative solution. This initially implies using the knowl-

edge of preceding and succeeding ICs for the valida-

tion purposes of incident notiﬁcations (Challenge 1).

Afterwards, we require novel techniques for assessing

the observation success of detector stations based on

the same information (Challenge 2). Finally, we need

to explore how ICs can learn the accuracy and impact

of the detection success of their neighbours using a re-

inforcement approach (Challenge 3). This will be the

basis for the actual response to alarms in OTC (and

which must be learned, not just communicated since

the impact are not visible for the neighbour).

Our starting point is preliminary work for estab-

lishing PSS (Tomforde et al., 2008). Here, informa-

tion exchange protocols based on communication and

processing techniques to consider the knowledge of

other ICs have been developed. We need to adapt

the communication protocols and develop novel tech-

niques to make use of sensor information and classi-

ﬁcation knowledge from neighbouring ICs.

A Concept for Collaborative Incident Validation in a Self-organised Trafﬁc Management System

319

Challenge 1: Collaborative Validation of

Stream-based Incident Classiﬁcations

To begin with, the existing unicast-based communica-

tion protocols in OTC have to be enabled to exchange

information about trafﬁc volumes running over shared

roads and the underlying aggregated detector signals.

Based on these communication capabilities, we pro-

pose to investigate a pull-based mechanism where ICs

can ask their neighbours about observations and clas-

siﬁcations of trafﬁc behaviour. Here, we distinguish

two cases: an event-driven approach (e.g., in case of

abnormal trafﬁc behaviour or if incidents have been

signalised) and a self-testing loop (i.e., in cycles to

verify the correct functioning of a particular sensor).

For the event-driven approach, the IC has to iden-

tify the relevant neighbour with respect to the detec-

tor location (upstream or downstream). Taking the

estimated travel time (i.e., following the approach in

(Prothmann et al., 2012) based on Webster’s formula

(Webster, 1959)) to/from this intersection controller

into account, the patterns of trafﬁc ﬂow can be com-

pared with the neighbour’s information. If the com-

parisons show a correlation of effects, the incident

alarm is conﬁrmed. Otherwise, it should be delayed –

and an additional validation step has to be done in the

other direction of the trafﬁc stream. We further have

to investigate which features can be used in addition

to the trafﬁc volume to improve the validation effect

(e.g., slope, curvature, and variability of the data).

In turn, the self-testing approach aims at assessing

the behaviour of the own sensors. Therefore, the traf-

ﬁc behaviour of all road segments approaching a de-

tector station of a turning, the corresponding detector

data, the detector data of the road segment taking up

the turning’s trafﬁc stream, and the information of the

IC where this road segment leads must be evaluated.

In general, the smoothed and averaged trafﬁc volumes

of all three involved intersections have to account for

to the same level. Based on a comparison of these

ﬁgures, deviations can be detected – and they can be

related to an individual source of information (i.e., a

detector station or a neighbour). As a result, either a

neighbour can be triggered that the received informa-

tion is conspicuous or the impact of the disturbance

of the own detector is analysed in detail (i.e., repeat-

ing this analysis based on historical data backwards in

time until deviations are no longer signiﬁcant).

Challenge 2: Collaborative

Self-assessment

The concept of the previous Challenge 1 is based on a

bilateral comparison of detector data and aggregated

trafﬁc ﬂow estimations. In this Challenge 2, we in-

crease the focus towards a network-wide collective

self-assessment of incident information and detector

plausibility. Therefore, we need to further investigate

if and how the validation effect can be improved by

considering longer streams than just pairs of ICs. This

implies higher uncertainty due to detector-inherent

differences and trafﬁc splitting into lanes at each in-

tersection (or between lanes on road segments), but it

also allows to follow trafﬁc streams for sequences of

detector stations. As a result, we aim at estimating the

expected trafﬁc volume and use this as ground truth

for computing how the individual detector deviates

from the expected stream. This approach has its ad-

vantages in cases where, e.g., two consecutive detec-

tor stations are disturbed simultaneously. In general,

we have to estimate a time series of detector readings

where the travel times between detector stations refer

to the time steps of the time series. We propose to ap-

proximate the time series linearly and use this infor-

mation to detect outliers. This process can be further

improved by considering more features than just the

trafﬁc volumes.

However, the effort of such an approach is dra-

matically higher than the bilateral approach of Chal-

lenge 1: All-to-all communication between several

ICs burdens the underlying communication network.

In order to (i) keep the effort at a feasible level and

(ii) narrow the overall trafﬁc stream estimation prob-

lem down to a certain part of the network, we pro-

pose to follow an event-based approach again: All in-

cident alarms that cannot be validated by the corre-

sponding neighbour serve as an event to start the dis-

tributed mechanism. Consequently, a responsible IC

is available to manage the process. The mechanism

itself runs conceptually in iterations: We propose to

increase the horizon hop-wise in both directions in

each iteration. This can then be augmented by a sub-

sequent outlier detection approach as outlined above

to identify incorrect detector information or to ﬁnally

explain the deviations by normal variations from the

underlying stream information (e.g., if the stream is

characterised by highly heterogeneous readings).

Challenge 3: Reinforced Reliability

Estimation of Neighbouring Intersection

Controllers

The approach as outlined by the previous two chal-

lenges means that ICs notify their neighbours of de-

tected incidents by starting the distributed validation

process (both, the bilateral and the multilateral vari-

ant). In Challenge 4, we propose to develop tech-

niques to take this information into account when de-

VEHITS 2022 - 8th International Conference on Vehicle Technology and Intelligent Transport Systems

320

ciding about adapting the current control strategies.

However, such a reaction is only successful if the un-

derlying information is quickly available and highly

reliable. The reliability is affected by characteristics

that do not depend on the incident detection tech-

niques. For instance, different aspects such as the

topology of the intersection, parking areas between

intersections, or large taxi and bus stopping areas may

result in temporal abnormality. Some of these char-

acteristics have a constant impact while others (e.g.,

large bus areas) have a situation-dependent impact

(e.g., the time of the day). Challenge 3 focuses on

learning the reliability of the incident alarms of an IC,

depending on the current situation.

Since this information is not directly available,

we have to learn the corresponding reliability at run-

time. The basic idea of this Challenge 3 is (a) to com-

bine the observations of neighbouring ICs and (b) to

use their situation-dependent reliability estimation for

the IC under consideration in an ensemble-based ap-

proach again. For the ﬁrst step (a), we propose to fol-

low a similar approach as already used in OTC for the

online signalisation adaptation: We make use of rein-

forcement learning capabilities. The concept utilises

a variant of an XCS that maps a trafﬁc situation (mea-

sured at the neighbouring intersection) to a reliability

estimation of the incident alarm of the considered IC

– and then learns the accuracy and ﬁtness of the reli-

ability estimation over time. This has to be done for

all neighbouring intersections. For the second step

(b), we propose to use the reliability estimations of

each neighbour as input and compute an aggregated

measure. Initially, each neighbour has one vote and

we will have to investigate if it is possible to im-

prove the behaviour by adapting these votes (i.e., the

weights assigned to the input of a neighbour). An-

other aspect of the challenge in this context is that

for the bilateral approach only one or two neighbours

are involved in most cases (based on the design of the

approach). Consequently, the ensemble itself faces

a dynamic constellation of participants. We need to

further investigate how this affects the success of the

estimation.

Challenge 4: Consideration of Incident

Information in Control Strategies

Based on OTC, we have to investigate how reliable in-

cident information can be considered in the different

aspects of the controller decisions: (a) for adapting

the signalisation, (b) for maintaining PSS, and for (c)

modiﬁcation of route guidance information.

For aspect (a), we have to investigate possibilities

to modify the decision system of the adaptation cy-

cle. Alternatives include: (i) extending the situation

description as basis for the adaptation loop with in-

cident information (drawback: is not part of initial

rules, increases the search space), (ii) artiﬁcially de-

creasing the trafﬁc volumes of the situation descrip-

tion towards an outgoing section if the incident is part

of this road segment (drawback: affects the learning

mechanism, the exact value for reduction has to be

determined), or (iii) modify trafﬁc volumes by extrap-

olating the (estimated) impact of the incident (draw-

back: affects the learning mechanism). However, in

all cases, we have to assess the implications on the

learning feedback and the resulting self-improving

adaptation behaviour.

For aspect (b), the PSS algorithm needs to be mod-

iﬁed – which select partners based on the current traf-

ﬁc ﬂow volumes and coordinates the signalisation of

those ICs that serve the trafﬁc streams with the highest

number of vehicles (i.e., negotiating common cycle

times and offsets). Incident information can be used

to alter these mechanisms in different ways, including

the following:

1. Anticipatory switching to alternative PSS since

the incident is expected to decrease the trafﬁc vol-

umes to be served by the current PSS.

2. Favouring coordination schemes that are expected

to alleviate the impact of the incident (e.g., those

that faster release trafﬁc affected by the incident).

3. Prefering PSS that avoid reported incidents.

Since each IC decides autonomously, the incident

information (such as reliability, severity, expected im-

pact, and type) has to be considered in the decision

process when choosing partners and signalising the

need for a PSS update. An incident-aware approach

has to estimate the beneﬁt of changing the PSS in re-

sponse to the incident, where “beneﬁt” is computed

in terms of cars being served and the uncertainty as-

signed to this result.

For aspect (c), we rely on existing work, i.e. two

variants of fully decentralised route guidance mech-

anisms imitating the Distance-Vector Routing (DVR)

and the Link-State Routing (LSR) protocols (Proth-

mann et al., 2012). Both are based on broadcast-

ing local trafﬁc data: ICs exchange either a topol-

ogy graph of the controlled intersection (i.e., includ-

ing in- and outgoing roads, neighbour information,

and destination information) augmented with the cur-

rent trafﬁc conditions (i.e., occurring delays and ex-

pected travel times – LSR variant) or propagate short-

est paths through the network (DVR variant). In the

LSR variant, shortest paths have to be determined for

each incoming section based on, e.g., Dijkstra’s al-

gorithm, while the DVR variant directly provides this

A Concept for Collaborative Incident Validation in a Self-organised Trafﬁc Management System

321

information. The resulting shortest path information

(i.e., the list of considered destinations together with

the next hop/turning advice at the intersection and the

expected travel time) are displayed via Variable Mes-

sage Signs (VMS) for each road approaching the in-

tersection. Here, incident awareness requires modiﬁ-

cations of these concepts.

Assuming a static acceptance rate, i.e. a given per-

centage of drivers that will follow the recommenda-

tions displayed via VMS, leads to the question of how

a variation of the acceptance rate within simulations

has an impact on the outcome (following the method

suggested in (Bazzan and Kluegl, 2005)). Hence, we

have to investigate how (i) the routing protocol and

(ii) the computation of route recommendations has

to be modiﬁed to consider incident information. For

the LSR variant, this initially means to further anno-

tate the topology graph representation to be commu-

nicated via broadcasts. Therefore, edges of the graph

representation can be annotated with additional val-

ues representing the incident status, the estimated im-

pact (i.e., severity and duration), and the reliability

of this information. After broadcasting this topology

information, each IC can build a complete graph rep-

resentation of the underlying road network on its own

and can compute shortest paths. However, we have

to investigate how these computations are impacted

by incident information: (a) individual roads suffer-

ing from incidents have to be avoided, (b) intersec-

tions affected by incidents have to be avoided, and

ferred. For instance, this can be done by introducing

static penalty values for links and nodes, introducing

varying penalty values (e.g., in a gradient approach

surrounding the incident area), or removing links and

nodes from the graph representation. We have to anal-

yse and compare these concepts. As an alternative,

a multi-objective variant of Dijkstra’s algorithm may

be developed. For the DVR variant, similar consider-

ations as for deriving shortest paths have to be done

when updates of routing entries arrive via broadcast

messages.

5 CONCLUSIONS

In this work, we argued that trafﬁc incident detec-

tion in urban road networks is different to that at

highways. Following recent results, clustering ap-

proaches allow for appropriate reliability of detected

events – which can be performed locally at intersec-

tions. However, to incorporate incident information

in signalisation strategies (for green duration modiﬁ-

cation, progressive signal systems, and even for route

guidance), the uncertainty has to be reduced. Conse-

quently, we propose to utilise of local dependencies

in self-organised road networks as trafﬁc ﬂows pass

several intersections. This spatio-temporal behaviour

can be used to collaboratively compare and analyse

ﬂow information – and to either conﬁrm incident in-

dicators or to reject them in a self-organised manner.

We proposed a research agenda comprising four

major challenges: (a) collaborative validation of

stream-based incident classiﬁcation, (b) collaborative

self-assessment of sensory equipment, (c) learning

the reliability of estimations, and (d) Consideration

of incident information in signalisation strategies. We

provided conceptual approaches to tackle these chal-

lenges – based on an integration in the Organic Trafﬁc

Control system and the available cluster-based inci-

dent detection approach. Our current and future work

focuses on the subsequent implementation of the con-

cepts and the subsequent tackling of these challenges.

ACKNOWLEDGEMENTS

This research was supported by the Deutsche

Forschungsgemeinschaft, DFG, in the context of the

project “Zwischenfall-bewusstes resilientes Verkehrs-

management f

ur urbane Straßennetze (InTURN)” un-

der grant TO 843/5-1. We acknowledge this support.

REFERENCES

Ahmed, S. and Cook, A. (1980). Time series models for

freeway incident detection. Transp. Eng. J of the Am.

Soc. of Civ. Eng., 106(6):731–745.

Aimsun (2021). Aimsun Next 20 User’s Manual, Aimsun

Next 20.0.3 edition.

Bazzan, A. and Kluegl, F. (2005). Case studies on the braess

paradox: Simulating route recommendation and learn-

ing in abstract and microscopic models. Transp. Res.

Part C: Emerging Tech., 13(4):299 – 319.

Dudek, C., Messer, C., and Nuckles, N. (1974). Incident de-

tection on urban freeways. Transp. Res. Rec., 495:12–

24.

Dudek, C., Weaver, G., Ritch, G., and Messer, C. (1975).

Detecting freeway incidents under low-volume condi-

tions. Transp. Res. Rec., 533:34–47.

Dusparic, I. and Cahill, V. (2009). Using distributed w-

learning for multi-policy optimization in decentralized

autonomic systems. In Proc. of 6th Int. Conf. on Au-

tonomic Computing, pages 63–64. ACM.

Ester, M., Kriegel, H.-P., Sander, J., and Xu, X. (1996).

A density-based algorithm for discovering clusters in

large spatial databases with noise. In kdd, pages 226–

231. AAAI Press.

VEHITS 2022 - 8th International Conference on Vehicle Technology and Intelligent Transport Systems

322

Feng, Y., Hourdos, J., and Davis, G. (2014). Probe vehicle

based real-time trafﬁc monitoring on urban roadways.

Transp. Res. Part C: Emerging Tech., 40:160–178.

Gall, A. and Hall, F. (1989). Distinguishing between inci-

dent congestion and recurrent congestion: a proposed

logic. Transp. Res. Rec., (1232).

Gokulan, B. and Srinivasan, D. (2010). Distributed geomet-

ric fuzzy multiagent urban trafﬁc signal control. IEEE

Trans. on Int. Transportation Sys., 11(3):714–727.

Helbing, D., L

ammer, S., and Lebacque, J. (2005). Self-

organized control of irregular or perturbed network

trafﬁc. Optimal control and dynamic games, pages

239–274.

Jenelius, E. and Koutsopoulos, H. (2013). Travel time esti-

mation for urban road networks using low frequency

probe vehicle data. Transp. Res. Part B: Methodolog-

ical, 53:64–81.

Kamijo, S., Harada, M., and Sakauchi, M. (2004). An in-

cident detection system based on semantic hierarchy.

In Proc. of 7th Int. Conf. on Int. Trans. Sys. (ITS’04),

pages 853–858. IEEE.

Lin, W. and Daganzo, C. (1997). A simple detection scheme

for delay-inducing freeway incidents. Transp. Res.

Part A: Policy and Practice, 31(2):141–155.

Mauro, V. and Taranto, C. D. (1990). Utopia. Control,

computers, communications in transportation.

uller-Schloer, C. and Tomforde, S. (2017). Organic

Computing-Technical Systems for Survival in the Real

World. Springer.

Oliveira, L. D. and Camponogara, E. (2010). Multi-agent

model predictive control of signaling split in urban

trafﬁc networks. Transp. Res. Part C: Emerging Tech.,

18(1):120–139.

Payne, H. and Tignor, S. (1978). Freeway incident-

detection algorithms based on decision trees with

states. Transp. Res. Rec., (682).

Payne, H. J. (1975). Freeway incident detection based upon

pattern classiﬁcation. In Proc. of IEEE Conf. on Deci-

sion and Control, volume 14, pages 688–692. IEEE.

Prothmann, H., Branke, J., Schmeck, H., Tomforde, S.,

Rochner, F., H

ahner, J., and M

uller-Schloer, C.

(2009). Organic trafﬁc light control for urban road net-

works. Int. J. Auton. Adapt. Commun. Syst., 2(3):203–

225.

Prothmann, H., Tomforde, S., Lyda, J., Branke, J., H

ahner,

J., M

uller-Schloer, C., and Schmeck, H. (2012).

Self-organised routing for road networks. In Self-

Organizing Systems - 6th IFIP TC 6 International

Workshop, IWSOS 2012, Delft, The Netherlands,

March 15-16, 2012. Proceedings, pages 48–59.

Robertson, D. and Bretherton, D. (1991). Optimizing net-

works of trafﬁc signals in real time – the SCOOT

method. IEEE Trans. on Veh. Tech., 40(1):11–15.

Schrank, D., Albert, L., Eisele, B., and Lomax, T. (2021).

2021 URBAN MOBILITY REPORT.

Schrank, D., Eisele, B., and Lomax, T. (2019). 2019 UR-

BAN MOBILITY REPORT.

Shehata, M., Cai, J., Badawy, W., Johannesson, R., and

Radmanesh, A. (2008). Video-based automatic inci-

dent detection for smart roads: The outdoor environ-

mental challenges regarding false alarms. IEEE Trans.

on Int. Transp. Sys., 9(2):349–360.

Sims, A. and Dobinson, K. (1980). The Sydney coordinated

adaptive trafﬁc (SCAT) system – Philosophy and ben-

eﬁts. IEEE Trans. on Veh. Tech., 29(2):130–137.

Sommer, M., Tomforde, S., and H

ahner, J. (2013). Using

a neural network for forecasting in an organic traf-

ﬁc control management system. In 2013 Workshop

on Embedded Self-Organizing Systems, ESOS’13, San

Jose, CA, USA, June 25, 2013.

Sommer, M., Tomforde, S., and H

ahner, J. (2016). Forecast-

augmented route guidance in urban trafﬁc networks

based on infrastructure observations. In Proceedings

of the International Conference on Vehicle Technol-

ogy and Intelligent Transport Systems, VEHITS 2016,

Rome, Italy, April 23-24, 2016, pages 177–186.

Stephanedes, Y. and Chassiakos, A. (1993). Freeway inci-

dent detection through ﬁltering. Transp. Res. Part C:

Emerging Technologies, 1(3):219–233.

Studer, L., Ketabdari, M., and Marchionni, G. (2015). Anal-

ysis of adaptive trafﬁc control systems design of a de-

cision support system for better choices. J Civil Envi-

ron Eng, 5(195):2.

Takaba, S. and Matsuno, H. (1985). Trafﬁc incident detec-

tion using correlation analysis. In SCS 1985 Summer

Comp. Sim. Conf., pages 529–534.

Thomsen, I., Zapfe, Y., and Tomforde, S. (2021). Urban

trafﬁc incident detection for organic trafﬁc control: A

density-based clustering approach. In Proceedings of

the 7th International Conference on Vehicle Technol-

ogy and Intelligent Transport Systems, VEHITS 2021,

Online Streaming, April 28-30, 2021, pages 152–160.

Tomforde, S., Prothmann, H., Branke, J., H

ahner, J., Mnif,

M., M

uller-Schloer, C., Richter, U., and Schmeck, H.

(2011). Observation and control of organic systems.

In Organic Computing—A Paradigm Shift for Com-

plex Systems, pages 325–338. Springer.

Tomforde, S., Prothmann, H., Rochner, F., Branke, J.,

ahner, J., M

uller-Schloer, C., and Schmeck, H.

(2008). Decentralised progressive signal systems

for organic trafﬁc control. In 2008 Second IEEE

International Conference on Self-Adaptive and Self-

Organizing Systems, pages 413–422. IEEE.

Vincent, R., Peirce, J., and Webb, P. (1990). Mova trafﬁc

control manual. MOVA reports.

Webster, F. (1959). Trafﬁc Signal Settings - Technical Paper

No 39. Road Research Laboratory, London, UK.

Willsky, A., Chow, E., Gershwin, S., Greene, C., Houpt,

P., and Kurkjian, A. (1980). Dynamic model-based

techniques for the detection of incidents on freeways.

IEEE Transa. on Automatic Control, 25(3):347–360.

Wilson, S. W. (1995). Classiﬁer Fitness Based on Accuracy.

Evolutionary Computation, 3(2):149–175.

A Concept for Collaborative Incident Validation in a Self-organised Trafﬁc Management System

323