Scenario Interpretation based on Primary Situations for Automatic

Turning at Urban Intersections

David Perdomo Lopez

, Rene Waldmann

, Christian Joerdens

and Ra

ul Rojas

Automated Driving, Volkswagen Group Research, Wolfsburg, Germany

Department of Mathematics and Computer Science, Freie Universi

at Berlin, Berlin, Germany

Keywords:

Automated Driving, Scenario Interpretation.

Abstract:

Even for a human driver, urban intersections represent probably the most difﬁcult scenarios, in which the

driver could be overloaded by understanding the trafﬁc rules, predicting the intention of other objects, etc. The

complexity of these scenarios makes the task of automated driving at intersections a very difﬁcult challenge.

Thus, we propose an approach that aims to reduce the complexity of the scenario interpretation by breaking

down the problem into a set of primary situations linked over time. Based on the combination of four primary

situations, the scenario interpretation should enable the corresponding planning that guides the ego vehicle

along a driving corridor.

1 INTRODUCTION

Driver assistance and automated driving systems have

become a very emerging ﬁeld of research in the last

decades. In (Okuda et al., 2014) the authors review

the most promising approaches and techniques used

in these kind of systems.

Considering the basic conceptual ﬂowchart in Fig.

1, the automated driving process can be simpliﬁed

into four main steps: (1) perception, (2) scenario in-

terpretation, (3) planning and (4) control. In this

basic representation, the perception module provides

the description of the surrounding world to the next

module. Then, the scenario interpretation module

achieves the comprehension of the relevant informa-

tion for the following planning and control stages.

Figure 1: Basic conceptual ﬂowchart for automated driving

based on four main steps. The environment perception (1)

represents the low level processing of sensors and a priori

data (e.g. image processing, object recognition and track-

ing, localization and mapping, etc.). The scenario interpre-

tation (2) corresponds with the understanding of the pro-

cessed data. The planning (3) makes the proper decisions

and delivers them to the control module (4), which ﬁnally

provides the adequate signals in terms of steering and ac-

celeration.

Obviously, the more inaccurate the perception is,

the more complex is the interpretation of the provided

data. But even if the perception provides accurate in-

formation about the surrounding of the ego vehicle,

the problem is not simple. The large number of possi-

ble collisions with other road users at urban intersec-

tions makes the problem a very complex challenge.

For this reason, the proposed concept aims to enable

the decision making for automated turning at urban

intersection in a simple manner.

In this paper, we ﬁrst describe the general con-

cept of scenario interpretation for automated driving.

Then, section 3 gives an overview of related work at

intersections. After that, we address the problem in

section 4. Hereafter, the proposed approach is ex-

plained in section 5. And ﬁnally, section 6 concludes

the paper.

1.1 Deﬁnition of Scenario

Interpretation

Before going into further details, two concepts have

to be deﬁned: scenario and interpretation. The au-

thors in (Geyer et al., 2014) propose a deﬁnition for

some relevant terms in the automated driving context

(situation, scene, scenario, etc.). As can be seen in

Fig. 2, the scenery is deﬁned as the combination of

all possible single static elements (e.g. road network,

number of lanes, crosswalks, position of trafﬁc lights,

speed limits, etc.). The scene contains the scenery and

the information of all dynamic objects with their cor-

Lopez, D., Waldmann, R., Joerdens, C. and Rojas, R.

Scenario Interpretation based on Primary Situations for Automatic Turning at Urban Intersections.

DOI: 10.5220/0006150300150023

In Proceedings of the 3rd International Conference on Vehicle Technology and Intelligent Transport Systems (VEHITS 2017), pages 15-23

ISBN: 978-989-758-242-4

Figure 2: Deﬁnition of the term scenario by (Geyer et al.,

2014).

responding states. The situation consist of the scene

and optional ego vehicle. In this context, the situa-

tion describes the current state, which could persist

several seconds until some conditions or criteria are

ﬁlled. On the contrary, the scenario describes differ-

ent states over the time, so that it contains at least one

situation, in which the last situation corresponds to

the last relevant situation.

Another deﬁnition is used in (Domsch and Negele,

2008). Domsch et al. propose a deﬁnition for terms

like: driver-situation, trafﬁc situation, scenario, etc..

A driving situation is described with the parameters:

static (road network, trafﬁc rules, priority, etc.), dy-

namic (objects, trafﬁc lights phases, etc.) and diverse

(weather, road conditions, etc.). Moreover, Ulbrich et

al. (Ulbrich et al., 2015) present a coherent review

and comparison of these terms and propose their own

deﬁnitions.

On the other hand, the term interpretation refers

to the act of explaining the meaning of something. In

this sense, the perception module makes a description

of the surrounding of the vehicle and the scenario in-

terpretation module gives a proper meaning to this in-

formation. Accordingly, the scenario interpretation at

intersection involves the following tasks:

• Filtering relevant information

• Using the information of the road network with

corresponding logical correspondences

• Predicting the intention of other vehicles

• Handling occlusions

• Achieving risk assessment

• Considering logical trafﬁc rules

• Handling the right of way

• Handling localization uncertainty

• Etc.

For example, in the ﬁrst example of Fig. 3 (A) the

ego vehicle (in blue) is turning to the left and another

vehicle (in red) is approaching the intersection. It be-

comes obvious that it is crucial to know on which lane

the other car is driving to determine a possible colli-

sion with the ego vehicle: if the red car is driving on

Figure 3: Examples of scenario interpretation at intersec-

tions. (A) Intention prediction of an oncoming vehicle

with inaccurate position: the ego vehicle (in blue) is mak-

ing a left turn maneuver (blue path) and other vehicle (in

red) is approaching the intersection. The yellow blob sur-

rounding the other vehicle represents the uncertainty of its

measured position. In case that the other vehicle is driving

forward (red path), there is an intersection of both paths.

Otherwise (black dotted path), there is no collision between

both vehicles. (B) Intention prediction of an oncoming

vehicle with accurate position: the ego vehicle is turn-

ing left and the other vehicle (which is already in the in-

tersection and its position is accurate enough) could drive

forward or turn left. (C) Handling occlusion while ap-

proaching an intersection: the ego vehicle is approaching

the intersection and aims to turn to the right. Due to an ob-

stacle (e.g. another vehicle), the occlusion impede to detect

a crossing pedestrian at the right side of a zebra crossing.

The green and red colored regions indicate the perceptible

and non perceptible areas respectively.

its most left lane, it is just allowed to turn to the left,

so that a collision with the ego vehicle is not expected.

Alternatively, if the other car is not driving on its most

left lane, its path has a conﬂict with the ego’s driving

corridor. Thus, if the position of other vehicles (or

ego vehicle) is not accurate enough (e.g. due to loca-

tion uncertainty), the scenario interpretation module

has to manage the uncertainty of the information in

order to understand how critical is the situation.

But even considering a perfect accuracy of the po-

sition of both vehicles, a proper intention prediction

could be crucial (depending on the road network and

its turning possibilities). As shown in 3 (B), the ego

vehicle is turning left and the other car could perform

two maneuvers: driving forward or turning left. In

this case, the accuracy of the state of the other car

(e.g. yaw, velocity, etc.) is crucial to achieve a proper

intention prediction.

Moreover, an important task of the scenario inter-

pretation module is to handle occlusions. In this way,

it is not only important to understand the provided in-

VEHITS 2017 - 3rd International Conference on Vehicle Technology and Intelligent Transport Systems

formation, but also to take into account which infor-

mation is missing. For example, in Fig. 3 (C) the

ego vehicle (blue) is approaching the intersection and

an obstacle (a parked car in white) impede to detect

a pedestrian. For this given scenario, the ﬁrst pedes-

trian (behind the obstacle) is not detected due to the

occlusion, but a proper scenario interpretation should

be able to interpret the occlusion as a critical miss-

ing information. Consequently, it is unclear if more

pedestrians approach the crosswalk or not.

1.2 Scenario Interpretation at Urban

Intersections

Due to the complexity of scenarios at urban intersec-

tions, it becomes obvious that a proper scenario in-

terpretation is required. In recent years several meth-

ods have been proposed to tackle this problem. Vacek

et al. (Vacek et al., 2007) present an approach for

a case- and rule-based situation interpretation using

description logic. The raw data from the sensors is

stored and transformed into a higher level represen-

tation. The different expected behavior of other ve-

hicles generates the linkage of other cases over time

with corresponding probabilities for every different

situation. Since the number of different options be-

comes very large at intersections, the computational

cost for the description logic reasoning constitutes the

main drawback of this approach. Logic description is

also used by Huelsen et al. (H

ulsen et al., 2011) to de-

scribe an ontology that represents the road networks,

objects, their relations and the corresponding trafﬁc

rules. The goal is to reason relations, objects, traf-

ﬁc rules (e.g. hasRightOfWay or hasToYield) using

inference services. Even keeping only necessary in-

formation for reasoning, the main drawback of this

approach are the high computational costs. There-

fore, this approach is insufﬁcient for real-time com-

putation.

Geyer et al. (Geyer et al., 2011) present a method

based on the cooperation between the driver and the

system with the Conduct-By-Wire (CBW) concept.

Depending on the current driving situation, and the

required information, the so-called gates are identi-

ﬁed. A driving situation is described with three types

of parameters: Static (road network, trafﬁc rules, pri-

ority, etc), dynamic (objects, trafﬁc lights phases, etc.)

and diverse (weather, road conditions, etc.). The sys-

tem analyzes the required information at the gates.

Consequently, different automation levels are set to

make easier the cooperation between the system and

the driver. To determine which information is needed,

a occupancy map and entry directions at the intersec-

tion are set. The CBW approach was also used by

Schreiber and Negele (Schreiber et al., 2010) to de-

velop of a maneuver catalog from the driver point

of view. The focus is to analyze what the driver is

expected to do. This information is combined with

a set of maneuvers that should cover every possible

trafﬁc and driving maneuver. The authors in (Alonso

et al., 2011)present two methods for priority conﬂict

resolution (priority charts and priority levels) using a

vehicle-to-vehicle (V2V) communication system as a

requirement. The ﬁrst method uses vectors to describe

the turning possibilities of all vehicles and their corre-

sponding priority signs. Then, an auxiliary table con-

taining all possible vectors associated with Boolean

values is used to indicate if the ego vehicle has to

move or stop. This table contains 111 different cases

without considering the trafﬁc signs combinations (3

for one vehicle, 27 for two vehicles, and 81 for three

vehicles). On the other hand, the second proposed

method aims to determine whether the ego vehicle can

continue or must wait by interpreting the different pri-

ority levels (using an auxiliary truth table to detect po-

tential conﬂicts with other vehicles). The authors pro-

pose a ﬂowchart to handle the right of way problem.

These two proposed methods depend on an speciﬁc

topology (in this case a two road intersection). More-

over, V2V communication is required. Although the

focus of (Lotz and Winner, 2014) is not to turn auto-

matically at urban intersections, the authors propose

a maneuver-based planning for automated vehicles.

Based on the desired maneuver (or set of maneuvers

over the time) the proposed system plans the proper

lane change by approaching the intersection. The ap-

proach was tested in a multi-lane road network with-

out other road users.

2 PROBLEM DESCRIPTION

The described problem is focused on understanding

the perceived information of the environment at urban

intersections. In this sense, the interpretation should

enable to plan the proper vehicle motion for turn-

ing at urban intersections with different precedence

states. Unfortunately, there is no international regula-

tion that controls the trafﬁc ﬂow at intersections in a

unique manner for all the possible scenarios all over

the world. Therefore, this work considers the regu-

lation described in the Vienna Convention on Road

Signs and Signals (for Europe, 2006) and the German

regulation (f

ur Straßen und Verkehrswesen (FGSV),

2010) in particular. A coherent way to address the

problem is to describe an intersection in a simple

manner. Therefore, we try to make a conceptual de-

scription of the scenario answering three questions

Scenario Interpretation based on Primary Situations for Automatic Turning at Urban Intersections

Figure 4: Simple classiﬁcation to describe an intersection.

(cf. 4):

• Which maneuver is making the ego vehicle?

• In which way is the trafﬁc ﬂow controlled?

• What is the topology of the intersection?

Therefore, the trafﬁc ﬂow at intersections can be

controlled in three different ways: by the right of way

rule, with vertical signs or trafﬁc lights. The ego ve-

hicle intention and the control of the trafﬁc ﬂow, yield

different scenarios and potential conﬂicts with other

vehicles or vulnerable road users (VRU). Fig. 5 il-

lustrates the different possible scenarios considering

a simple intersection topology. This classiﬁcation is

an improved version of the method proposed by Fas-

tenmeier (Fastenmeier et al., 1995):

In this sense, every possible state of the intersec-

tion deﬁnes the control of the trafﬁc ﬂow (from an ego

perspective, i.e. it indicates how ego vehicle should

handle the right of way). The following intersection

states are considered:

Denied: a common circular red trafﬁc light has been

detected. The ego vehicle has to stop as long as

the trafﬁc light color is red.

Permitted: a common circular green light has been

detected, so that the ego vehicle is allowed to turn.

However, the ego vehicle has to give way to on-

coming vehicles while turning left and VRU have

priority in parallel conﬂicts.

Protected: a green arrow trafﬁc light has been de-

tected. According to (f

ur Straßen und Verkehr-

swesen (FGSV), 2010), the path of the ego vehicle

to complete the turning maneuver has no conﬂicts

with other road users.

Permitted on Red: a static sign with a green arrow

has been detected beside a red trafﬁc light. Even

if the trafﬁc light indicates red, the ego vehicle is

allowed to turn if there is no potential collision

with other crossing/oncoming vehicles.

Right before Left: a vertical sign indicates that the

rule right before left has to be applied or no verti-

cal sign controls the trafﬁc ﬂow, and consequently,

this rule is applied by default.

With Precedence: a priority road sign has been de-

tected, so that other crossing vehicles are required

to give way to the ego vehicle.

Give Way: a give way is used to notify the ego vehi-

cle that it has to give way to other crossing vehi-

cles.

Stop: the detected stop sign implies that the ego ve-

hicle is required to stop (even if any other vehicle

is crossing) and give way to other vehicles.

The other possibility to describe an intersection is

to consider its topology. Other authors (Gersten-

berger, 2015) have analyzed in detail the most com-

mon topologies to determine the relation between the

different topologies and the trafﬁc accidents. Consid-

ering the large number or possible different topolo-

gies, we take for granted that a scenario interpreta-

tion based on speciﬁc topologies is not appropriate.

Therefore, the proposed solution aims to achieve a

left and right turn maneuver independently on the in-

tersection topology.

3 PROPOSED APPROACH

The proposed approach aims to make the interpreta-

tion of the scenario (and further planning) easier by

breaking it down into primary situations. Therefore,

we ﬁrst describe in detail how we deﬁne a scenario

based on primary situations and ﬁnally we explain

how the ego vehicle is guided to complete the turn-

ing maneuver.

3.1 Scenario Interpretation based on

Primary Situations

A scenario contains mainly three important concepts:

the linkage of the expected primary situations, the ego

intention and the information of the intersection state,

which deﬁnes how the trafﬁc ﬂow is controlled from

an ego perspective. In this way, the primary situa-

tions are deﬁned by potential conﬂicts with other road

users and further relevant. A single primary situation

VEHITS 2017 - 3rd International Conference on Vehicle Technology and Intelligent Transport Systems

Figure 5: Classiﬁcation of possible scenarios at a simple intersection topology considering the desired maneuver and right of

way. Every column represents a different maneuver of ego vehicle with its path (blue). Every row corresponds to a different

manner to control the of right of way. All possible path of other vehicles with a potential collision with ego vehicle is colored

depending on its priority. The other vehicles (or VRU) with a red path have priority with respect to ego vehicle. The other

vehicles with yellow paths are required to give way to ego. The dotted arrows indicate paths of other vehicles without an

intersection with ego’s path.

Figure 6: Nine primary target points for automatic turning

left or right at intersections.

requires at least a target point, its corresponding re-

gion(s) of interest (ROI) and the relevant objects in-

side this ROI(s). This concepts are explained down

bellow. Every target point indicates the position and

velocity of the ego vehicle along a determine driving

path (or trajectory). It represents the stages where a

single primary situation is expected, so that the set

of target points deﬁnes the maneuver of the vehicle.

As shown in Fig. 6, the set of primary target points

for turning at intersections can be automatically cal-

culated based on the road network information..

Furthermore, every ROI describes a geometric

area (as 2D polygon) that has to be observed for every

primary situation. It represents the area where rel-

evant objects are expected. For example, if a zebra

crossing was detected in front of the ego vehicle, the

ROI for this situation should represent the area where

pedestrians could be. In other words, the ROIs indi-

cate the areas where objects (other vehicles or VRU)

could appear and produce potential conﬂicts with the

ego vehicle. Therefore, the ROIs can be calculated

automatically considering the road network informa-

tion (i.e. considering the paths intersection of the ego

Figure 7: Primary situations (A, B, B1, B2, C, C1, D) with

corresponding region of interest (yellow semi-transparent

rectangles) for a generic intersection and ego position (col-

ored in blue).

vehicle and other road users).

As it is illustrated in Fig. 7, we propose four pri-

mary situations with their corresponding ROIs (A, B,

C, D):

A: a perpendicular conﬂict with VRU (e.g. a cross-

walk, zebra crossing or bike lane) is in front of the

ego vehicle. The corresponding ROI consists of

the conﬂict and its surround area where potential

objects are expected. For example, if the ego ve-

Scenario Interpretation based on Primary Situations for Automatic Turning at Urban Intersections

hicle is approaching the intersection and there is

a perpendicular bike lane in front of it, the ROI

is automatically calculated considering the area

that the ego vehicle aims to overdrive and the sur-

rounding area along the bike lane where a possi-

ble bicycle could be. Obviously, the size of the

area along the bike lane is determined by the ego

velocity (the faster ego vehicle drives, the larger

should be the size of the area along the bike lane).

B: the ego vehicle has a conﬂict with a left-cross

lane (e.g. at a T-form intersection without right-

crossing lanes). Here, the ROI is calculated con-

sidering the information of the road network along

the possible left-cross lanes. B1 is not considered

a primary situation its own, but a mirrored version

of B, in which the cross lane comes from the right

side. In addition, B2 corresponds to a combina-

tion of B and B1 (e.g. at a X-form intersection).

C: the ego vehicle has a conﬂict with a parallel cross-

walk, zebra crossing or bike lane. Perpendicu-

lar and parallel conﬂicts with VRUs by turning

at intersections have to be handled in a different

manner compared to situation A. For example, at

an intersection controlled with trafﬁc lights, when

the state is permitted, the ego vehicle has prece-

dence with respect to the VRUs crossing a per-

pendicular crosswalk. On the contrary, the ego

vehicle has no precedence with respect to VRUs

crossing a parallel crosswalk (this is graphically

explained in Fig. 5).

D: the ego vehicle has a conﬂict with an oncoming

vehicle.

The linkage of the primary situations represents

the order in which consecutive single primary situ-

ations are expected and the relations between them.

Accordingly, the situations can be vertically or hori-

zontally linked: a vertical link indicates that the next

primary situation is expected at a certain distance

(and, depending on the variation of the ego velocity

over time). But a horizontal link denotes that another

different primary situation could also be expected at

the same distance. The level of the vertical and hor-

izontal linkage is denoted by i and j respectively. In

this sense, a scenario (S) denotes the connections of

primary situations (PS

i j

S = {PS

, ..., PS

;

, ..., PS

;

...;

, PS

(1)

where M is the number of different vertical linkage

levels and Ni is the number horizontal linkage levels

Figure 8: Concept of a scenario based on linkages of pri-

mary situations. (A) and (B) illustrates the concept and a

given example respectively. Every ellipse (PS

i j

) represents

a primary situation with its corresponding relevant infor-

mation. The blue and green arrows indicate the vertical (i)

and horizontal ( j) linkage level respectively. The yellow

rectangles represent the information of the ROIs and the de-

tected objects. In this example, the ﬁrst situation PS

linked vertically with the primary situations PS

and PS

which are horizontally linked between them (green arrow)

and once again vertically linked with PS

for every corresponding vertical level (i). The concept

is illustrated in Fig. 8.

For example, the ego vehicle is approaching the

intersection with the intention of turning to the left

without precedence. Firstly (i.e. vertical linkage level

1), a perpendicular crosswalk is detected, so that the

ﬁrst situation PS

corresponds to the primary situa-

tion A. In this case, the yellow rectangle represents

the ROI around the crosswalk and the detected ob-

jects. Lets say that the information of the road net-

work is not accurate (or trustworthy) enough, and

consequently the interpretation module can not guar-

anty that the intersection has a T- or X-form. In this

case two possibilities are expected: PS

= B (left-

cross lane) and PS

= B2 (left and right-cross lane).

Furthermore, independently on which primary situa-

tion is the most probable at the vertical linkage 2, the

next expected situation (PS

) is D (i.e. we expect on-

coming vehicles inside the corresponding ROI). And

so on, the scenario is interpreted by generating the

linkage of primary situations over time and continu-

ously updating these linkages. For this given exam-

ple, as ego vehicle is getting closer to the intersection,

the road network information becomes more accurate,

so that the vertical linkages become also more robust.

In other words, the linkages are updated taking into

account the information of the perception module and

optimizing the connections between the primary situ-

ations.

3.2 Planning the Maneuver

Once the scenario is generated, the target points have

to be set according to the available information (e.g.

relevant objects or intersection state), so that the ego

VEHITS 2017 - 3rd International Conference on Vehicle Technology and Intelligent Transport Systems

Figure 9: Flowchart for automated turning based on primary situations. It explains step by step the connection of the expected

primary situations and their according target points. The last target point guides the ego vehicle to the end of the turning

maneuver (rectangles with green border). A given example is highlighted in red (left turn without precedence).

vehicle is guided to achieve the whole turning maneu-

ver. Since the trajectory of the ego vehicle is part of

the scenario, the target points are set along a driving

path. The ﬂowchart highlighted in Fig. 9 explains

which information is needed to set the proper target

points for every primary situation. For the sake of

clarity, the diagram has been kept simple by consid-

ering only a very basic topology. In other words, the

consideration of more complex topologies (e.g. han-

dling T- or X-form intersections) is omitted to ease its

representation and understanding. Furthermore, the

difference between an intersection state stop or give

way, and the intersection state permitted on red are

also omitted.

Once the ego vehicle approaches the intersection

and has done the proper lane change(s), the ﬁrst re-

quired information is how plausible the existence of

trafﬁc lights and vertical signs are. This determines

the ﬁrst main branching of the ﬂowchart, which is

based on how the trafﬁc ﬂow is controlled, namely by

trafﬁc lights, vertical signs or the right before left rule.

In the given example (highlighted with red arrows in

the ﬂowchart 9 any trafﬁc light is detected. Then, a

primary situation A is expected depending on the ex-

istence of a perpendicular conﬂict with VRUs. Then,

the ﬂowchart is divided into two branches depending

how the trafﬁc ﬂow is controlled, i.e., by vertical signs

or the right before left rule. In the given example a

yield sign was detected, and the ego vehicle intends

to turn left, so that both left and right possible cross-

ing vehicles have the right of way (primary situation

B2). Consequently, if the a collision with ego vehi-

cle and some crossing vehicle from both sides inside

the corresponding ROIs is predicted, the target point 2

forces ego vehicle to stop in front of the conﬂict area

so long as no collision is expected. Then, the next

primary situation D implies setting the target point 3,

in such a way as to avoid a collision with oncoming

vehicles. But in case that any collision is predicted

(e.g. because the are no oncoming vehicles in the cor-

responding ROI), the next target point 4 (primary sit-

uation C ) is set as long as the conﬂict area is not pass-

able. Otherwise, the left turn maneuver is completed

with the target point 5. An important advantage of this

concept is that a target point does not only indicate a

position, but also a desired velocity. This allows han-

dling occlusions in a simple manner. For example, in

case that the ego vehicle is making a right turn (see

Fig. 10) and an obstacle (e.g. other car) impedes to

observe the ROI completely, the scenario interpreta-

tion should not take for granted that the whole ROI is

free. In fact, a pedestrian could be in those parts of the

ROI that are not perceptible (in this example behind

the white car). As human drivers, we would probably

reduce the velocity because we are not sure if some

pedestrians are crossing.

Scenario Interpretation based on Primary Situations for Automatic Turning at Urban Intersections

Figure 10: Occlusion example illustrated over the time (n,

n + 1 and n + 2). The ego vehicle (blue) is making a right

turn maneuver (blue path) and other vehicle (white) im-

pedes to observe the ROI (yellow rectangle) completely.

The perceptible and not perceptible areas are colored in

green and red respectively. The illustrated pedestrian and

its path (dotted red arrow) corresponds to a placed virtual

object.

The idea is to imitate cognitive human reaction in

a very simple way: setting a virtual object. This ob-

ject is placed representing the worst case (i.e. a pedes-

trian is crossing so that a collision with the ego vehicle

will occur). In the given example, a virtual pedestrian

is placed at the time n and n + 1. This placed object

causes that the prediction module set the target point

6 with a very low velocity (e.g. 5 Km/h), because a

pedestrian could be behind the white car.

4 CONCLUSIONS

In this paper a scenario interpretation approach for au-

tomated driving at intersections has been introduced.

The main goal of the proposed method is to make

the interpretation of the scenario and the further deci-

sion making easier. This has been achieved by break-

ing down the problem into four primary situations (or

combinations of them). After explaining the mean-

ing of the term scenario interpretation, an overview

of the related work was given. Then, we described

the problem taking into account the intention on the

ego vehicle and the intersection state that controls the

trafﬁc ﬂow. This analysis identiﬁed the potential con-

ﬂicts with other road users in a simple manner. In

this way, the scenario consists of linkages of expected

primary situation over time, in which just the rele-

vant information is needed. These primary situations

are deﬁned by the potential conﬂicts with other road

users and further information (target point, ROI, rele-

vant objects). Moreover, a ﬂowchart to complete the

left and right turn maneuver was presented. This dia-

gram represents the combination of primary situations

over time facilitating to turn left/right automatically.

Compared to state-of-the-art solutions, a very impor-

tant advantage of our system is that it may be applied

independently of the intersections topology. Further-

more, it offers the possibility of handling occlusions

in a simple way.

Future research will focus on optimizing the pro-

cess of predicting the intersection state that controls

the right of way at the intersection. Moreover, re-

search work to analyze the computational cost of the

proposed approach have to be done. In this sense,

a detailed evaluation of the proposed approach and

its functionality over the time for real scenarios will

be achieved and compared with other techniques.

These future validation involves scenarios with dif-

ferent VRUs, weather conditions and topologies.

REFERENCES

Alonso, J., Milan

es, V., P

erez, J., Onieva, E., Gonz

alez, C.,

and De Pedro, T. (2011). Autonomous vehicle control

systems for safe crossroads. Transportation research

part C: emerging technologies, 19(6):1095–1110.

Domsch, C. and Negele, H. (2008). Einsatz von referenz-

fahrsituationen bei der entwicklung von fahrerassis-

tenzsystemen; 3. Tagung Aktive Sicherheit durch

Fahrerassistenz, April, pages 07–08.

Fastenmeier, W. et al. (1995). Autofahrer und Verkehrssi-

tuation. Neue Wege zur Bewertung von Sicherheit

und Zuverl

assigkeit moderner Strassenverkehrssys-

teme. Number 33.

for Europe, U. N. E. C. (2006). Convention on Road Signs

and Signals of 1968. Geneva, Switzerland.

ur Straßen und Verkehrswesen (FGSV), F. (2010).

Richtlinien f

ur Lichtsignalanlagen (RiLSA).

Gerstenberger, M. (2015). Unfallgeschehen an Knoten-

punkten: Grundlagenuntersuchung zu Ursachen und

Ansatzen zur Verbesserung durch Assistenz. PhD

thesis, M

unchen, Technische Universit

at M

unchen,

Diss., 2015.

Geyer, S., Baltzer, M., Franz, B., Hakuli, S., Kauer, M.,

Kienle, M., Meier, S., Weißgerber, T., Bengler, K.,

Bruder, R., et al. (2014). Concept and development of

a uniﬁed ontology for generating test and use-case cat-

alogues for assisted and automated vehicle guidance.

Intelligent Transport Systems, IET, 8(3):183–189.

Geyer, S., Hakuli, S., Winner, H., Franz, B., and Kauer, M.

(2011). Development of a cooperative system behav-

VEHITS 2017 - 3rd International Conference on Vehicle Technology and Intelligent Transport Systems

ior for a highly automated vehicle guidance concept

based on the conduct-by-wire principle. In Intelligent

Vehicles Symposium (IV), 2011 IEEE, pages 411–416.

IEEE.

ulsen, M., Z

ollner, J. M., and Weiss, C. (2011). Trafﬁc in-

tersection situation description ontology for advanced

driver assistance. In Intelligent Vehicles Symposium

(IV), 2011 IEEE, pages 993–999. IEEE.

Lotz, F. and Winner, H. (2014). Maneuver delegation and

planning for automated vehicles at multi-lane road

intersections. In Intelligent Transportation Systems

(ITSC), 2014 IEEE 17th International Conference on,

pages 1423–1429. IEEE.

Okuda, R., Kajiwara, Y., and Terashima, K. (2014). A sur-

vey of technical trend of adas and autonomous driving.

In VLSI Technology, Systems and Application (VLSI-

TSA), Proceedings of Technical Program-2014 Inter-

national Symposium on, pages 1–4. IEEE.

Schreiber, M., Kauer, M., Schlesinger, D., Hakuli, S., and

Bruder, R. (2010). Veriﬁcation of a maneuver cata-

log for a maneuver-based vehicle guidance system. In

Systems Man and Cybernetics (SMC), 2010 IEEE In-

ternational Conference on, pages 3683–3689. IEEE.

Ulbrich, S., Menzel, T., Reschka, A., Schuldt, F., and Mau-

rer, M. (2015). Deﬁning and substantiating the terms

scene, situation, and scenario for automated driving.

In Intelligent Transportation Systems (ITSC), 2015

IEEE 18th International Conference on, pages 982–

988. IEEE.

Vacek, S., Gindele, T., Z

ollner, J. M., and Dillmann, R.

(2007). Using case-based reasoning for autonomous

vehicle guidance. In Intelligent Robots and Systems,

2007. IROS 2007. IEEE/RSJ International Conference

on, pages 4271–4276. IEEE.

Scenario Interpretation based on Primary Situations for Automatic Turning at Urban Intersections