Maneuver-based Visualization of Similarities between Recorded Trafﬁc

Scenarios

Thilo Braun, Lennart Ries, Moritz Hesche, Stefan Otten and Eric Sax

FZI Research Center for Information Technology, Haid-und-Neu-Str. 10-14, 76131 Karlsruhe, Germany

Keywords:

Scenario-based Testing, Scenario Extraction, Graph Visualization.

Abstract:

Since automated driving functions are safety-critical systems, extensive validation and veriﬁcation is neces-

sary. Scenario-based testing is a promising approach for this challenge. For selection of relevant scenarios, col-

lected data and knowledge models are potential sources. In this paper we introduce a concept to use recorded

trajectory and map data, abstracted to maneuvers, to describe the scenarios and visualize them intuitively. This

enables a data-driven scenario-mining process to ﬁnd relevant scenarios for the testing of automated driving

functions. To compare the scenarios, a similarity measure based on the manuevers is designed and the sce-

narios and their similarities are represented as a graph. Graph-visualization methods, already successfully

applied in other domains, structure the collected data for further analysis. The concept is exemplary applied

to an urban trafﬁc dataset.

1 INTRODUCTION

Automated driving functions on public roads have

the potential to fundamentally change our transporta-

tion in the future. But validation and veriﬁcation are

major concerns to ensure a sufﬁciently high level of

safety and acceptance (PEGASUS, 2019). As the

functionality and operational design domain for au-

tomated driving functions increases, the combination

of safety-critical systems in an unstructured open-

world lead to new challenges in safety argumenta-

tion. Distance-based testing, as practiced to date, is no

longer feasible to ensure safe systems because several

billions of test kilometers would have to be driven for

a valid test coverage (Wachenfeld and Winner, 2015)

(Kalra and Paddock, 2016). Even with the usage of

simulation tools, the required amount of test kilome-

ters can not be achieved (Pfeffer, 2020). Therefore, a

more efﬁcient way to increase test coverage is needed.

Current research projects and industry are focusing

on scenario-based testing in this regard (PEGASUS,

2019)(Neurohr et al., 2020)(VVM, 2020)(Bagschik

et al., 2017). This approach exploits the redundancy

of trafﬁc scenarios that occur during distance-based

testing: Instead of testing the same scenarios several

times, as they occur randomly in distance-based test-

ing, they are selected from a scenario catalog. Since

one test of the scenario is enough to ensure the safety

of the driving function in this scenario, the test cov-

erage increases more efﬁciently. The scenario cata-

log should contain all relevant scenarios for the oper-

ational design domain of the automated driving func-

tion to ensure a certain performance metric. However,

one question is unanswered yet: How to build the sce-

nario catalog and how to select the scenarios for exe-

cution and testing of automated driving functions?

Recorded real-world driving data from drones, traf-

ﬁc infrastructure, or vehicles include helpful informa-

tion to support these challenges. They are the most

realistic source for scenarios and contain informa-

tion about their occurrence probabilities. Since the

amount of data from recorded trips is too extensive

for manual analysis and the unprocessed data does not

provide information in directly useable form, meth-

ods are needed to extract the relevant information.

The recorded data builds a scenario space with high

redundancy. A structured and intuitive representa-

tion of the recorded scenarios, e.g. as maneuvers,

enables efﬁcient data mining in this scenario space.

Different techniques can help to extract relevant in-

formation from the data: Grouping similar scenar-

ios can reduce the high redundancy. E.g., all sce-

narios where a vehicle is driving alone on a straight

road can be considered a scenario group. Also, the

relationships between these groups contain relevant

information. From a maneuver perspective, follow-

ing another vehicle on a straight road is more simi-

lar to the previously described scenario, than driving

236

Braun, T., Ries, L., Hesche, M., Otten, S. and Sax, E.

Maneuver-based Visualization of Similarities between Recorded Trafﬁc Scenarios.

DOI: 10.5220/0011140600003269

In Proceedings of the 11th International Conference on Data Science, Technology and Applications (DATA 2022), pages 236-244

ISBN: 978-989-758-583-8; ISSN: 2184-285X

over an intersection in an urban environment. In ad-

dition, there are rarely occurring exceptional corner

cases that have to be identiﬁed, such as driving ma-

neuvers executed by emergency vehicles against the

rules for ordinary trafﬁc participants, which must also

be considered in the test process. All of these tech-

niques need a similarity measure between the indi-

vidual scenarios. Furthermore, this similarity can be

used as an input for a graph representation for intu-

itive visualization. In other domains, this graph anal-

ysis approach is successfully used for challenges of

visualizing data of similar data structures by many au-

thors (Batagelj and Mrvar, 1998)(Pavlopoulos et al.,

2017)(Bastian et al., 2009). In literature manage-

ment (Perianes-Rodriguez et al., 2016) the relation-

ships of various publications and journals are visual-

ized. Biological (Shannon et al., 2003) and genomic

data (Thimm et al., 2004) (Theocharidis et al., 2009)

is represented as graphs for interpretation. Also so-

cial networks are analyzed with this method (Borgatti

et al., 2002). This paper examines the extent to which

the methods, already proven helpful in other domains,

can achieve a structured and intuitive representation

of the scenario space in the automotive context.

2 RELATED WORK

2.1 Scenario Sources and Description

A formal deﬁnition for scenarios is presented in the

(ISO21448, 2019): A scenario describes the tem-

poral development between several scenes in a se-

quence of scenes. Compared to a scene, which is

a snapshot without temporal expansion, a scenario

spans a certain amount of time. According to the

PEGASUS Project (PEGASUS, 2020) and (Neurohr

et al., 2020), there are two sources for scenarios:

Knowledge-based and data-based. In the knowledge-

based approach, scenarios are created using expert

knowledge and a knowledge model (Bagschik et al.,

2018)(Ponn et al., 2019) (Neurohr et al., 2021). For

the data-based approach, recorded real-world trafﬁc,

simulated trafﬁc, test drives and accident databases

build a possible set of source for the scenarios (Koop-

man and Wagner, 2018)(PEGASUS, 2020)(Hartjen

et al., 2019a)(Weber et al., 2021)(Lizenberg et al.,

2021).

(Scholtes et al., 2021) propose a 6 layer model

to describe different aspects of the scenarios: Street

level (L1), trafﬁc infrastructure (L2), temporal mod-

iﬁcations of L1 and L2 (L3), movable objects (L4),

environment conditions (L5) and digital information

(L6).

For the dynamic description of the movable ob-

jects in L4, which is the focus of this paper, a

sequence of driving maneuvers is an established

way (ASAM, 2020)(Pfeffer, 2020) (Hartjen et al.,

2019b)(Braun et al., 2021)(King et al., 2021). A

maneuver is an abstract description of the behavior

of a participant during a speciﬁed timespan (Bach,

2018). Compared to a description with trajectories,

the maneuvers are more concise and semantically in-

terpretable and therefore offer easier further process-

ing.

2.2 Scenario Similarity and Clustering

While not much literature exist on similarities be-

tween scenarios, there is ongoing research on the re-

lated method of scenario clustering. (Ries et al., 2021)

uses a DTW-comparison of recorded trajectories to

cluster scenarios. (King et al., 2021) cluster scenarios

based on a maneuver extraction. Scenarios with the

same maneuvers are considered as equal and build a

cluster. A more sophisticated measure for similarities

between non-equal scenarios is not created. (Hart-

jen et al., 2020) also deﬁne the equality based on

extracted maneuvers and investigate how often new

scenarios occurs during recordings. (Balasubrama-

nian et al., 2021) introduce a random forest activa-

tion pattern to build clusters of trafﬁc scenarios based

on iterative optimization of self-supervised networks.

(Langner et al., 2019) derives the cluster based on the

static elements of the scenarios. (Ries et al., 2020)

convert trajectories to driving states to represent sce-

narios and use them for a semantic comparison of the

recorded scenarios.

2.3 Graph-visualization

As stated in chapter 1, visualization methods for

graphs give an intuitive representation of the under-

lying data. The position of the nodes in this visual-

ization is called layout and can be calculated by dif-

ferent algorithms: Force-directed algorithms offer a

ﬂexible and domain-independent way to create lay-

outs. The underlying concept of these categories of

layout algorithm is to model the graph as a system

of particles which all exert forces on each other de-

pending on their position in the layout and their actual

relationship in the graph structure. From a general

view, force directed layouts try to ﬁnd an appropri-

ate positioning of the nodes by minimizing the corre-

sponding energy in the modelled system. A compre-

hensive comparison of existing force-directed meth-

ods is given in (Kobourov, 2012). Besides the force-

embedders, multidimensional scaling is a collective

Maneuver-based Visualization of Similarities between Recorded Trafﬁc Scenarios

237

(a) Recorded trajectories and

segmenation to scenarios.

(b) Segmented scenarios as a graph with

similarities as edges.

lar scenarios are close to each other).

Figure 1: Concept for extraction of scenarios, building a graph and a visualization based on similarities.

term for methods whose goal is to visualize the sim-

ilarity relations of objects by the suitable arrange-

ment of points in low-dimensional space (Cox and

Cox, 2008). While multidimensional scaling is often

used as a method for dimension reduction, (Gansner

et al., 2004) and (Klimenta, 2012) consider the appli-

cation of this method as a layout for (fully connected)

graphs.

3 CONCEPT FOR SCENARIO

SPACE VISUALIZATION

The concept contains the following steps (Fig. 1):

Abstraction of the trajectories to maneuvers se-

quences, segmentation and extraction of scenarios,

calculation of similarities between the scenarios and

graph visualization based on the similarities.

3.1 Abstraction through Representation

as Maneuver Matrix

A modiﬁed version of the maneuver list of (Hartjen

et al., 2019a) is used to describe the behavior of the

trafﬁc participants. Each participant gets assigned

a maneuver state for each maneuver type for each

timestep. For demonstration purposes, this paper will

mainly focus on the maneuver type “vehicle state”,

which describes the longitudinal motion, and the ma-

neuver type “turn”, which describes the behavior on

intersections. Nevertheless, the concept is applicable

to any other maneuver types. The state of maneuver

type “vehicle state” can be “cruise”, “decelerate”, or

“standstill”. The state of maneuver type “turn” can be

“turn left” or “turn right” for the direction when cross-

ing an intersection or “no turn”, if the participant does

not perform a turn maneuver. A maneuver matrix M,

containing the timesteps as columns and the extracted

maneuvers of each type as rows as shown in Fig. 2,

serves as a formalized abstract representation for fur-

ther processing.

Building on the trajectories and map data, the ma-

neuvers are identiﬁed with a rule-based algorithm also

used in previous work (King et al., 2021). Decelera-

tion, standstill and cruise maneuvers are calculated by

analyzing the vehicle speed, turn maneuvers by com-

paring the map with the position and yaw-angle of the

trafﬁc participant.

3.2 Scenario Segmentation and

-Extraction

Since the recorded driving data can contain many

hours of driving and includes multiple scenarios, the

data is segmented to extract the scenarios. Depending

on the focus of the analysis, meaningful deﬁnitions

of start and end points can vary. We use a ﬂexible

maneuver-based segmentation algorithm: The start

and end of the scenario is is determined by the change

of the state of a single maneuver type, as exemplarily

shown in Fig. 2. To get the maneuver matrix of the

scenario, the corresponding columns are cut out.

cruise decelerate standstill

no turn turn right no turn

Figure 2: Example abstraction of trajectories to “vehicle

state” and “turn” maneuvers. Segmentation of the scenarios

, S

and S

using “turn” maneuvers.

DATA 2022 - 11th International Conference on Data Science, Technology and Applications

238

turn

σ(S

, S

) = σ

+ σ

turn

Figure 3: Relationships between the maneuvers of the maneuver types “vehicle state” (vs) and “turn” (turn), their maneuver

sequences M, maneuver-similarities σ

and scenario-similarity σ for two scenarios.

3.3 Maneuver-based Scenario

Similarity

A generally valid deﬁnition of scenario similarities is

not possible, since different elements of the scenario

are in focus depending on the context. Therefore, the

methods to calculate such a similarity also vary. This

paper introduces a deﬁnition based on the maneuver

matrices of the scenarios. To determine the similar-

ity of the matrices, the maneuver types are compared

independently in a ﬁrst step. The scenario similar-

ity is then a weighted sum of individual simmiliarities

w.r.t. to the different maneuver types used for repre-

sentation of the scenarios (Equ. 1, Fig. 3). Using the

deﬁnitions

• S

: Scenario with id A

• σ(S

, S

): similarity between the scenarios S

and S

• t: maneuver type (e.g. turn or vehicle state),

•

: maneuver at index i of scenario A

• M

= {

, . . . ,

}: maneuver sequence of

maneuver type t of scenario S

• σ

, M

): similarity for the maneuver type t

between scenarios S

and S

• c

: weighting factor for different maneuver types,

the scenario similarity is calculated by:

σ(S

, S

) =

∑

· σ

, M

) (1)

We use the pattern matching technique of se-

quence alignment on the maneuver sequences. For the

calculation of the sequence aligment the Needleman-

Wunsch algorithm (Needleman and Wunsch, 1970a)

is used. Sequence alignment was originally developed

for amino acid comparison (Needleman and Wunsch,

1970b), but is also used in other domains (Abbott and

Forrest, 1986) (

Cavojsk

y and Drozda, 2019). The

similarity of the two sequences is determined by the

number of matches, mismatches and gaps between

two sequences of categorical data, like the maneuver

sequences (Fig. 4). The scores for matches, mis-

matches and gaps can be parameterized for various

analyses.

Figure 4: Match, Mismatch and Gap for sequence align-

ment of two categorical sequences.

A normalization step is performed for the results

of the sequence alignment, to keep the maximum sim-

ilarity independent of the length of the maneuver se-

quences:

norm

0, 5 · (n

+ n

)

, (2)

where n

is the number of maneuvers in scenario

3.4 Graph-representation and

Visualization

The similarities between the scenarios form the ba-

sis for a graph-representation of the extracted scenar-

ios. Each node in the graph represents an extracted

scenario, each edge the calculated similarity between

these two scenarios. For visualization of the sce-

narios, state-of-the-art layout algorithms are used as

described in section 2.3. In our experience, multi-

dimensional scaling (Cox and Cox, 2008) gives the

best results for the scenario graphs. Nevertheless,

other layout methods, such as Fruchterman-Reingold

(Fruchterman and Reingold, 1991) and Kamada-

Kawai (Kamada et al., 1989), can be used.

4 EVALUATION

4.1 Dataset

For an exemplary evaluation, an intersection of the

INTERACTION dataset (Zhan et al., 2019) is used. It

contains recordings of urban trafﬁc with various inter-

actions and was created using a drone. In addition to

the recorded trajectories, map data of the intersection

is included and used for the maneuver identiﬁcation.

Maneuver-based Visualization of Similarities between Recorded Trafﬁc Scenarios

239

Figure 5: Considered intersection of the INTERACTION

dataset (Zhan et al., 2019).

Fig. 5 shows a snapshot of the intersection from the

drones perspective. The recording area of the selected

intersection includes four intersecting road segments

and thus a comparatively large variety of possible traf-

ﬁc scenarios. The intersection records include a total

of 732 individual road users over approximately 45

minutes. All graphs are visualized using multidimen-

sional scaling as layout.

4.2 Grouping by Turn Maneuvers

Figure 6: Scenario graph visualization for turn maneuver

sequences for all scenarios while crossing the intersection.

In the visualization in Fig. 6 each point represents a

scenario. A scenario here contains the trajectories of

a participant over the whole intersection, so no seg-

mentation based on maneuvers is performed. For the

similarity, only the turn maneuvers are considered, so

in terms of the terminology introduced in section 3.3

this corresponds to c

= 0 and c

turn

= 1. For valida-

tion purpose of the layout algorithm, the color repre-

sent the turn maneuver sequence of the corresponding

scenario. Table 1 shows the turn maneuver sequence

of selected groups, Fig. 7 the recorded trajectories

with the same color coding.

Table 1: Turn maneuver sequences of scenario groups in

Fig. 6.

group turn maneuver sequence

blue no turn

red no turn turn right no turn

pink no turn turn left no turn

brown no turn turn right

green turn right no turn

turquoise no turn turn left

purple turn left no turn

black no turn turn left no turn turn right no turn

Figure 7: Trajectories of scenarios in Fig. 6 on the intersec-

tion.

To validate the layout algorithm, the positions of

scenarios with equal turn maneuver sequences, repre-

sented by the color, are examined. As expected, sim-

ilar scenarios have smaller distance to each other and

form separated groups.

Additionally, similarities between the different

groups can be analyzed by looking at the relative po-

sitions of the groups: Scenarios of the purple and

turquoise groups, which have maneuver sequences

mixed of the blue and pink groups, are positioned in

between of these two groups. Also, the scenario in

the black group, that contains “turn left” and “turn

right” maneuvers intersparsed by “no turn” maneu-

vers, is positioned in the middle of the red and blue

groups, that contain either a “turn left” or a “turn

right” maneuver separated by “no turn” maneuvers.

Furthermore, a symmetry of right and left turn sce-

narios can be observed: The red and pink group, the

purple and the turquoise group and the brown and the

green group, which all contain equivalent maneuver

sequences regarding the turn direction, are mirrored

on an axis. The scenario of group f, containing a “turn

left” and a “turn right” maneuver, and the middle of

the blue group, containing no turn maneuvers, is part

of this axis.

The visualization also gives a fast impression of

the frequency of the scenarios: there are almost equal

numbers of left and right turn scenarios, but most of

the participants do not perform a turn maneuver.

DATA 2022 - 11th International Conference on Data Science, Technology and Applications

240

4.3 Grouping by Turn and Vehicles

State Maneuvers

Figure 8: Scenario graph visualization for a combination

“turn” and “vehicle state” maneuver sequences during turn

scenarios.

Table 2: “Vehicle state” maneuver sequences scenario

groups in Fig. 8.

group vehicle state maneuver sequence

blue cruise

red standstill decelerate

green decelerate cruise

purple cruise decelerate

black decelerate standstill cruise

pink cruise decelerate cruise

brown cruise decelerate standstill cruise

others various longer combinations of vehicle state maneuvers

The scenarios in Fig. 8 are segmented by turn ma-

neuver type and ﬁltered for “turn left” or “turn right”

maneuver during the scenario. Hence, only trajecto-

ries during a left or right turn are considered. The

turn maneuver with a weight of c

turn

= 0.1 and the

vehicle state with a weight of c

= 0.9 form the cal-

culation of the similarities. The color represents the

vehicle state maneuver series, the shape the turn di-

rection (turn right: diamond, turn left: circle).

Similar to the previous graph, scenarios with the

same maneuver sequence are mapped close to each

other and form groups. The combination of both, ve-

hicle state and turn maneuvers, lead to symmetrical

positions of the scenarios. “Turn right” maneuvers

(diamond) are mapped on the right, “turn left” maneu-

vers (circle) on the left, groups with the same vehicle

state maneuver sequences (same color) are opposite

each other. Most of the scenarios have maneuver se-

quences with a maximum of two maneuvers (blue, red

and green groups). The length of the maneuver se-

quence grows from bottom to top of the graph. The

size of the groups decreases with the length of the ma-

neuver sequence. The share of scenarios without de-

celeration or standstill maneuvers (blue groups) in the

total number of scenarios is lower for left turn scenar-

ios, which is plausible considering the trafﬁc rules for

the intersection.

Figure 9: Scenario graph visualization for a combination

“turn” and “vehicle state” maneuver sequences for all sce-

narios.

Fig. 9 shows an extension of the previous Fig. 8,

where additional to the “turn right” an “turn left” sce-

narios also all scenarios with the “no turn” maneuver

are included. So the entire recorded data is visual-

ized in the graph. The calculations for the similarity

remains the same as in the previous graph, the color

represents the combination of turn and vehicle state

maneuver series.

Some large, well-separated groups of similar scenar-

ios can be identiﬁed, formed by scenarios with short

maneuver sequences. Like in the previous graphs,

groups with similar maneuver sequences are arranged

close to each other.

In regions with scenarios of larger maneuver se-

quences, fewer groups are formed. In addition, there

are a number of outliers that cannot be assigned to any

group.

4.4 Beneﬁts for Scenario-based Testing

This visible redundancy and outlier detection brings

several beneﬁts for the use of the scenarios in the

scenario-based testing workﬂow: Instead of working

on the complete set of scenarios, each group can be

represented by one member. Rare, and therefore pos-

sibly unknown scenarios, can be easily identiﬁed in

Maneuver-based Visualization of Similarities between Recorded Trafﬁc Scenarios

241

a visual manner or following automated processing.

E.g. the participant in Fig. 6 which drives a trajectory

that is not allowed by the trafﬁc rules of the consid-

ered intersection, is shown isolated in the graph. Fur-

ther, correlations between the maneuvers in the sce-

narios can be identiﬁed. E.g. “turn left” maneuvers

occur more often in a combination with standstill than

“turn right” maneuvers (Fig. 8). The frequency of

the detected scenarios can be a basis for prioritization

of test scenarios. Analysis and visualization of the

scenario space make the frequency of scenarios intu-

itively visible and is thus a support for speciﬁcation

and test engineers.

5 CONCLUSION AND FUTURE

WORK

Scenario-based testing is supposed to play a sig-

niﬁcant role in the testing of highly automated

vehicles, but the building of a scenario catalog and

the selection of the relevant scenarios for efﬁcient

testing is up to now an unanswered question. In this

paper, we support to master these challenges with

the use of recorded trafﬁc data. For this purpose,

the data is represented in a scenario-graph structure.

The extraction of maneuvers builds the basis for

a semantic interpretation of the data. To describe

similarities between the scenarios, a metric based

on the maneuvers is introduced. Interpreting the

similarities as the weights of a graph, state-of-the-art

methods for graph visualizations create an intuitive

representation for the scenario catalog. The concept

was exemplary applied to a dataset of urban trafﬁc

on an intersection. The created graphs show how

this representation can present a rich variety of

information in a compact way, including information

on which scenarios have occurred, how the scenarios

can be grouped, with what frequency these groups

occur and whether there are any outliers that need

special attention or represent previously unknown

scenarios. This analysis supports the building of

a scenario catalog from data and increases the test

coverage for highly automated vehicles efﬁciently

by selecting suitable test scenarios. In addition, the

method can be used ﬂexibly, since different maneu-

vers and weights can be selected for composing a

suitable similarity measure depending on the use case.

Future work will focus on various maneuvers and

datasets, also including data from simulation tools.

Additional similarity metrics, using maneuver or tra-

jectories without abstraction, will be evaluated. Based

on the positions of the scenarios in the visualization

of a graph, a clustering algorithm could automatically

ﬁnd clusters of the scenarios and outliers.

ACKNOWLEDGEMENTS

The research leading to these results is funded by the

German Federal Ministry for Economic Affairs and

Climate Action” within the project “Veriﬁkations-

und Validierungsmethoden automatisierter Fahrzeuge

im urbanen Umfeld”. The authors would like to thank

the consortium for the successful cooperation.”

REFERENCES

Abbott, A. and Forrest, J. (1986). Optimal matching meth-

ods for historical sequences. The Journal of Interdis-

ciplinary History, 16(3):471–494.

ASAM (2020). OpenSCENARIO. https://www.asam.net/

standards/detail/openscenario/. Accessed 2022-03-01.

Bach, J. (2018). Methoden und Ans

atze f

ur die Entwick-

lung und den Test pr

adiktiver Fahrzeugregelungsfunk-

tionen. PhD thesis, KIT.

Bagschik, G., Menzel, T., and Maurer, M. (2018). Ontol-

ogy based scene creation for the development of au-

tomated vehicles. In 2018 IEEE Intelligent Vehicles

Symposium (IV), pages 1813–1820. IEEE.

Bagschik, G., Menzel, T., Reschka, A., and Maurer, M.

(2017). Szenarien f

ur Entwicklung, Absicherung und

Test von automatisierten Fahrzeugen. In 11. Workshop

Fahrerassistenzsysteme. Uni-DAS e. V, pages 125–

135.

Balasubramanian, L., Wurst, J., Botsch, M., and Deng, K.

(2021). Trafﬁc scenario clustering by iterative opti-

misation of self-supervised networks using a random

forest activation pattern similarity.

Bastian, M., Heymann, S., and Jacomy, M. (2009). Gephi:

an open source software for exploring and manipu-

lating networks. In Proceedings of the International

AAAI Conference on Web and Social Media.

Batagelj, V. and Mrvar, A. (1998). Pajek-program for large

network analysis. Connections, 21(2):47–57.

Borgatti, S. P., Everett, M. G., and Freeman, L. C. (2002).

Ucinet for windows: Software for social network anal-

ysis. Harvard, MA: analytic technologies, 6.

Braun, T., Ries, L., K

ortke, F., Turner, L. R., Otten, S., and

Sax, E. (2021). Collection of requirements and model-

based approach for scenario description. In VEHITS,

pages 634–645.

Cavojsk

y, M. and Drozda, M. (2019). Comparison of user

trajectories with the needleman-wunsch algorithm. In

International Conference on Mobile Computing, Ap-

plications, and Services, pages 141–154. Springer.

Cox, M. A. and Cox, T. F. (2008). Multidimensional scal-

ing. In Handbook of data visualization, pages 315–

347. Springer.

DATA 2022 - 11th International Conference on Data Science, Technology and Applications

242

Fruchterman, T. M. and Reingold, E. M. (1991). Graph

drawing by force-directed placement. Software: Prac-

tice and experience, 21(11):1129–1164.

Gansner, E. R., Koren, Y., and North, S. (2004). Graph

drawing by stress majorization. In International Sym-

posium on Graph Drawing, pages 239–250. Springer.

Hartjen, L., Philipp, R., Schuldt, D. F., Friedrich, P. D. B.,

and Howar, P. D. F. (2019a). Classiﬁcation of driving

maneuvers in urban trafﬁc for parametrization of test

scenarios. In 9. Tagung Automatisiertes Fahren.

Hartjen, L., Philipp, R., Schuldt, F., and Friedrich, B.

(2020). Saturation effects in recorded maneuver data

for the test of automated driving. Highway Res. Rec.

Hartjen, L., Schuldt, F., and Friedrich, B. (2019b). Seman-

tic classiﬁcation of pedestrian trafﬁc scenarios for the

validation of automated driving. In 2019 IEEE In-

telligent Transportation Systems Conference (ITSC),

pages 3696–3701. IEEE.

ISO21448 (2019). ISO/PAS 21448: 2019 Road Vehicles

Safety of the Intended Functionality. Norm, Interna-

tional Organization for Standardization.

Kalra, N. and Paddock, S. M. (2016). Driving to safety:

How many miles of driving would it take to demon-

strate autonomous vehicle reliability? Transportation

Research Part A: Policy and Practice, 94:182–193.

Kamada, T., Kawai, S., et al. (1989). An algorithm for draw-

ing general undirected graphs. Information processing

letters, 31(1):7–15.

King, C., Braun, T., Braess, C., Langner, J., and Sax, E.

(2021). Capturing the variety of urban logical sce-

narios from bird-view trajectories. In VEHITS, pages

471–480.

Klimenta, M. (2012). Extending the usability of multidi-

mensional scaling for graph drawing. PhD thesis,

Citeseer.

Kobourov, S. G. (2012). Spring embedders and force di-

rected graph drawing algorithms.

Koopman, P. and Wagner, M. (2018). Toward a framework

for highly automated vehicle safety validation. SAE

International Journal of Engines, 11(2018-01-1071).

Langner, J., Grolig, H., Otten, S., Holz

apfel, M., and Sax,

E. (2019). Logical scenario derivation by clustering

dynamic-length-segments extracted from real-world-

driving-data. In VEHITS, pages 458–467.

Lizenberg, V., Alkurdi, M. R., Eberle, U., and K

oster, F.

(2021). Intelligent co-simulation framework for co-

operative driving functions. In 2021 IEEE 17th In-

ternational Conference on Intelligent Computer Com-

munication and Processing (ICCP), pages 109–115.

IEEE.

Needleman, S. B. and Wunsch, C. D. (1970a). A gen-

eral method applicable to the search for similarities

in the amino acid sequence of two proteins. Journal

of Molecular Biology, 48(3):443–453.

Needleman, S. B. and Wunsch, C. D. (1970b). A gen-

eral method applicable to the search for similarities

in the amino acid sequence of two proteins. Journal

of molecular biology, 48(3):443–453.

Neurohr, C., Westhofen, L., Butz, M., Bollmann, M. H.,

Eberle, U., and Galbas, R. (2021). Criticality Anal-

ysis for the Veriﬁcation and Validation of Automated

Vehicles. IEEE Access, 9:18016–18041.

Neurohr, C., Westhofen, L., Henning, T., de Graaff, T.,

ohlmann, E., and B

ode, E. (2020). Fundamental

considerations around scenario-based testing for au-

tomated driving. In 2020 IEEE Intelligent Vehicles

Symposium (IV), pages 121–127. IEEE.

Pavlopoulos, G. A., Paez-Espino, D., Kyrpides, N. C., and

Iliopoulos, I. (2017). Empirical comparison of visu-

alization tools for larger-scale network analysis. Ad-

vances in bioinformatics, 2017.

PEGASUS (2019). Pegasus method - an overview.

https://www.pegasusprojekt.de/ﬁles/tmpl/Pegasus-

Abschlussveranstaltung/PEGASUS-Gesamtmethode.

pdf. Accessed: 2022-03-01.

PEGASUS (2020). PEGASUS - Schlussbericht.

https://www.pegasusprojekt.de/de/pegasus-

abschlussveranstaltung. Accessed: 2022-03-01.

Perianes-Rodriguez, A., Waltman, L., and Van Eck, N. J.

(2016). Constructing bibliometric networks: A com-

parison between full and fractional counting. Journal

of Informetrics, 10(4):1178–1195.

Pfeffer, R. (2020). Szenariobasierte simulationsgest

utzte

funktionale Absicherung hochautomatisierter

Fahrfunktionen durch Nutzung von Realdaten.

PhD thesis, Karlsruher Institut f

ur Technologie (KIT).

Ponn, T., Gnandt, C., and Diermeyer, F. (2019). An

optimization-based method to identify relevant sce-

narios for type approval of automated vehicles. In Pro-

ceedings of the ESV—International Technical Confer-

ence on the Enhanced Safety of Vehicles, Eindhoven,

The Netherlands, pages 10–13.

Ries, L., Rigoll, P., Braun, T., Schulik, T., Daube, J., and

Sax, E. (2021). Trajectory-based clustering of real-

world urban driving sequences with multiple trafﬁc

objects. In 2021 IEEE International Intelligent Trans-

portation Systems Conference (ITSC), pages 1251–

1258. IEEE.

Ries, L., Stumpf, M., Bach, J., and Sax, E. (2020). Se-

mantic comparison of driving sequences by adapta-

tion of word embeddings. In 2020 IEEE 23rd Interna-

tional Conference on Intelligent Transportation Sys-

tems (ITSC), pages 1–7.

Scholtes, M., Westhofen, L., Turner, L. R., Lotto, K.,

Schuldes, M., Weber, H., Wagener, N., Neurohr, C.,

Bollmann, M. H., K

ortke, F., et al. (2021). 6-layer

model for a structured description and categoriza-

tion of urban trafﬁc and environment. IEEE Access,

9:59131–59147.

Shannon, P., Markiel, A., Ozier, O., Baliga, N. S., Wang,

J. T., Ramage, D., Amin, N., Schwikowski, B., and

Ideker, T. (2003). Cytoscape: a software environment

for integrated models of biomolecular interaction net-

works. Genome research, 13(11):2498–2504.

Theocharidis, A., Van Dongen, S., Enright, A. J., and Free-

man, T. C. (2009). Network visualization and analysis

of gene expression data using biolayout express 3d.

Nature protocols, 4(10):1535–1550.

Thimm, O., Bl

asing, O., Gibon, Y., Nagel, A., Meyer, S.,

uger, P., Selbig, J., M

uller, L. A., Rhee, S. Y., and

Maneuver-based Visualization of Similarities between Recorded Trafﬁc Scenarios

243

Stitt, M. (2004). Mapman: a user-driven tool to dis-

play genomics data sets onto diagrams of metabolic

pathways and other biological processes. The Plant

Journal, 37(6):914–939.

VVM (2020). V&V Methoden. https://www.vvm-projekt.

de/konzept. Accessed: 2022-03-01.

Wachenfeld, W. and Winner, H. (2015). Die Freigabe des

autonomen Fahrens. In Autonomes Fahren, pages

439–464. Springer Vieweg, Berlin, Heidelberg.

Weber, N., Thiem, C., and Konigorski, U. (2021). A needle

in a haystack–how to derive relevant scenarios for test-

ing automated driving systems in urban areas. arXiv

preprint arXiv:2109.03648.

Zhan, W., Sun, L., Wang, D., Shi, H., Clausse, A., Nau-

mann, M., K

ummerle, J., K

onigshof, H., Stiller, C.,

de La Fortelle, A., and Tomizuka, M. INTERACTION

Dataset: An INTERnational, Adversarial and Cooper-

ative moTION Dataset in Interactive Driving Scenar-

ios with Semantic Maps. arXiv:1910.03088 [cs, eess].

DATA 2022 - 11th International Conference on Data Science, Technology and Applications

244