Adaptive Interaction Field Framework for Risk-Aware Navigation of

Driverless Minibus in Pedestrian Zones

Qazi Hamza Jan and Karsten Berns

RPTU Kaiserslautern-Landau, Erwin-Schr

odinger-Straße 52, 67663, Kaiserslautern, Germany

Keywords:

Mapping, Autonomous Driving, Pedestrian Zone, Risk Assessment.

Abstract:

In complex pedestrian zones, the navigation of driverless minibuses faces signiﬁcant challenges due to vary-

ing environmental structures and pedestrian behavior. These zones range from organized pathways to open,

unstructured spaces with minimal navigational cues. To address this, dynamic interaction ﬁelds are developed

around the driverless minibus, adapting in size and shape to real-time movements. To achieve a similar rep-

resentation as vehicle, interaction ﬁelds are developed that incorporate pedestrian unawareness. These virtual

ﬁelds facilitate safer and more intuitive interactions between vehicles and pedestrians by incorporating real-

time pedestrian awareness and activity data. The proposed model assesses risk by aggregating grid values

from overlapping zones between pedestrians and driverless minibus, computing potential encounters based on

spatial positions and awareness levels. A gradient-based heat map visualizes risk, highlighting areas where

interaction with pedestrians is needed. This adaptive approach enables the decision-making module to initiate

appropriate responses, such as escape maneuvers or interaction mode activation, based on risk thresholds. The

interaction ﬁeld further classiﬁes risk into ambient, direct, or critical levels, guiding the system’s reactions.

This framework enhances safety protocols and situational awareness in diverse urban environments. The vehi-

cle was able to drive and interaction in a better way with enabled interaction ﬁelds. Based on these risk values,

various interaction modules were activated, facilitating meaningful and context-aware interactions with pedes-

trians.

1 INTRODUCTION

The inclusion of driverless minibuses in pedestrian

zones is becoming an exciting area for the au-

tomotive industry and researchers (Jan and Berns,

2023),(Nesheli et al., 2021),(Milan

es et al., 2020).

With the expansion of pedestrian zones and narrow

spaces, driverless minibuses offer quick and ﬂexible

transport between buildings in shared spaces. Safety

becomes a primary concern when navigating through

crowds. In one of our previous works (Jan and Berns,

2021), a hardwired safety system was conﬁgured to

prevent collisions. This fail-safe system directly en-

abled emergency braking in case of routine malfunc-

tions. To ensure smooth navigation and avoid abrupt

safety braking due to pedestrians’ presence within the

ﬁeld, risky pedestrians need to be informed about the

vehicle and give way in advance.

Risky pedestrians are deﬁned by their unaware-

ness of the driverless minibus. This unawareness

could stem from several reasons, such as being en-

gaged in conversation, talking/texting on the phone,

or walking in the same direction ahead of the vehicle.

In such cases, it is necessary to interact with pedes-

trians to alert them using various interaction mod-

ules (Jan et al., 2023), (Rasouli and Tsotsos, 2019).

There exists a spectrum of structural variation in

pedestrian zones based on the intended use and level

of pedestrian and vehicle integration. This enforces

different walking and driving behaviors in pedestri-

ans and vehicle, respectively. To understand the riski-

ness of pedestrians around the vehicle, we introduce a

novel method of understanding riskiness of pedestri-

ans in maps by creating virtual dynamic ﬁelds around

the vehicle. Such dynamic ﬁelds, also known as ve-

hicle interaction ﬁelds serve dual purpose, to improve

safety and enable useful interaction between vehicle

and pedestrians. This helped in better driving and

efﬁcient interaction. To understand the riskiness of

pedestrians in the map, their riskiness is presented as

dynamic ﬁelds. Next section presents the inspiration

and concept of interaction ﬁeld for driverless mini.

Section 3 provides the implementation of the concept.

Integrated results from vehicle-pedestrian interaction

382

Jan, Q. H. and Berns, K.

Adaptive Interaction Field Framework for Risk-Aware Navigation of Driverless Minibus in Pedestrian Zones.

DOI: 10.5220/0013221400003941

Paper published under CC license (CC BY-NC-ND 4.0)

In Proceedings of the 11th International Conference on Vehicle Technology and Intelligent Transport Systems (VEHITS 2025), pages 382-389

ISBN: 978-989-758-745-0; ISSN: 2184-495X

Table 1: Translation of four interaction phases to three zones for driverless minibus. Implicit and subtle interactions from the

phased interaction (Vogel and Balakrishnan, 2004) are combined into direct interaction within the smart interaction zone in

this work.

Phased Interaction Smart Interaction Zones Purpose in Vehicle for Interaction

Ambient Display Ambient General awareness, signals vehicle presence

passively to pedestrians

Implicit Interaction Direct Overt communication with pedestrians

Subtle Interaction

Personal Interaction Risk Critical response measures for pedestrian

safety

ﬁeld are shown in Section 5.

2 PEDESTRIAN ENGAGEMENT

THROUGH PHASED

INTERACTION ZONES

The interaction ﬁeld is made into different phases,

since this allows to have speciﬁc interaction for

pedestrians with speciﬁc risk values and different dis-

tance to vehicle. The phase interaction was inspired

from Vogel et. al. (Vogel and Balakrishnan, 2004).

The authors have shown a ﬂuid interaction strategy

from distant to close interaction. In the literature,

these phases are Ambient Display, Implicit Interac-

tion, Subtle Interaction, and Personal Interaction. The

ambient display initially presents general information

passively, shifting to implicit interaction as a user ap-

proaches to offer personalized notiﬁcations. During

this phase, simple gestures enable shared use while

maintaining an overview. In the personal interaction

phase, users can engage directly via touch for detailed

information. To the best of our knowledge, no exist-

ing research considers pedestrians’ unawareness as a

risk factor rather than treating them solely as conven-

tional obstacles.

This paper proposes ”Interaction Field Zones”

based on (Vogel and Balakrishnan, 2004) phased in-

teraction framework to enhance communication with

pedestrians. It translates the four phases into three

zones as shown in Table 1. The model transitions

from ambient awareness to direct engagement, estab-

lishing a natural communication channel. In the ba-

sic zone, similar to an ”Ambient Display,” the vehicle

passively signals its presence to ensure pedestrians

are aware without direct interaction. As pedestrians

approach, entering an ”ambient interaction” phase,

the vehicle dynamically adjusts its behavior based on

their speed and path to convey subtle intentions. In the

”direct interaction” phase, it uses visual or auditory

cues to communicate actions and prioritize pedestrian

safety. In complex scenarios, it switches to a ”risk”

mode with emergency measures like braking to en-

sure safety. These interaction zones improve commu-

nication, safety, and trust, adapting to varying pedes-

trian environments and aligning with human interac-

tion paradigms.

3 INTERACTION FIELD DESIGN

Establishing interaction ﬁelds for Autobus enables

real-time risk assessment, spatial awareness of pedes-

trian activity, and adaptive responses to different

pedestrian behaviors, enhancing safety and trafﬁc

ﬂow.

To develop the vehicle interaction ﬁeld for driver-

less minibus, it is crucial to understand pedestrian be-

havior (Jan et al., 2022a), (Jan et al., 2022b), partic-

ularly the risks posed by a lack of awareness. These

risks often arise when pedestrians are distracted by

electronic devices or moving in the same direction as

the vehicle, increasing the likelihood of collisions or

delays. This work proposes a tailored model of inter-

action ﬁelds for pedestrian-vehicle interaction to en-

hance vehicle navigation and decision-making. The

goal is to reduce risks by enabling more informed and

responsive vehicle driving and interaction strategies.

3.1 Vehicle Interaction Field

Not all areas around a vehicle are relevant for pedes-

trian interaction while driving. Due to open spaces

and crowds, interactions should primarily focus on in-

dividuals in the vehicle’s drivable area, referred to as

”interactees,” even if they are not actively reacting to

Adaptive Interaction Field Framework for Risk-Aware Navigation of Driverless Minibus in Pedestrian Zones

383

Figure 1: The plot shows top view of pedestrian trajectories

and directions during encounters with the driverless minibus

during testing in the campus, revealing a consistent pattern

of avoidance. This predictable behavior can inform the de-

sign of interaction ﬁelds that minimize disruptions under

normal conditions.

the vehicle. This requires identifying and segment-

ing regions of interest based on interaction types. To

achieve this, interaction ﬁelds are created around the

vehicle to strategically limit and streamline the ac-

tivation of interaction modules, enhancing their efﬁ-

ciency.

The interaction ﬁeld conﬁguration is based on

pedestrian trajectories relative to the vehicle. As

shown in Figure 1, pedestrians approaching the ve-

hicle often display a clear pattern of yielding, es-

pecially in tight spaces. Research by Schneemann

et. al. (Schneemann and Gohl, 2016) highlights that

pedestrians typically adjust their path to avoid the ve-

hicle, indicating a collective behavior of spatial nego-

tiation and collision avoidance.

The interaction ﬁelds around the driverless

minibus, shown in Figure 2, are designed to enable

graduated communication and risk assessment based

on pedestrian proximity. Inspired by the behavior ob-

served in Figure 1, these ﬁelds extend further in the

direction of travel and are shorter in the opposite di-

rection, reﬂecting the dynamics of pedestrian-vehicle

interactions and risk assessment principles. The de-

sign of the ﬁeld was based on the following princi-

ples:

• The forward extension of the ﬁelds accounts for

the vehicle’s stopping distance, which varies with

speed.

• A longer ﬁeld in the travel direction provides suf-

ﬁcient space for safe stopping.

• Side ﬁelds are shorter since pedestrians in these

areas are more likely to have seen the vehicle pass,

reducing risk.

Figure 2: Schematic of the vehicle with interaction ﬁelds:

the ’Ambient’ ﬁeld (light red) serves as an initial awareness

zone, extending forward for early pedestrian detection; the

’Direct’ ﬁeld (darker red) enables active communication, al-

lowing dynamic responses; and the ’Risk’ ﬁeld (darkest red)

covers the closest zone for immediate collision prevention.

This stratiﬁcation supports graduated safety measures based

on proximity.

According to the zones clariﬁed in Table 1, the

three zones of interaction are:

• Ambient Interaction: The outermost ﬁeld pro-

vides general awareness by passively signaling

the vehicle’s presence to distant pedestrians, fo-

cusing on those crossing from the front. It aligns

with the ”Ambient Display” concept and involves

no collision risk.

• Direct Interaction: This ﬁeld facilitates active

communication, especially with pedestrians who

haven’t noticed the vehicle. It enables early de-

tection and allows the vehicle to adjust its speed

or path, issuing auditory alerts if necessary to en-

hance safety.

• Risk Interaction: The closest zone triggers crit-

ical responses when a pedestrian enters, includ-

ing emergency braking and visual alerts. It is de-

signed for immediate action to prevent collisions

in high-risk situations.

3.2 Pedestrian Interaction Field

To integrate pedestrian models into vehicular in-

teraction ﬁelds, representing pedestrians as ellipses

is effective, as suggested by W. Limprasert et

al. (Limprasert et al., 2013). Ellipses capture pedes-

trian size and direction, simplifying computations and

enabling quicker decisions. They adapt to changes in

speed and direction, aligning with natural human mo-

tion and enhancing pedestrian movement prediction.

This approach balances accuracy and computational

efﬁciency, improving safety in pedestrian-vehicle in-

teractions. Figure 3 illustrates this representation.

An ellipse is deﬁned by its geometric parameters,

which correlate with pedestrian movement relevant to

vehicle interaction ﬁelds. The standard equation for

an ellipse, centered at the origin and aligned with the

coordinate axes, is:

= 1 (1)

VEHITS 2025 - 11th International Conference on Vehicle Technology and Intelligent Transport Systems

384

Figure 3: A graphical depiction of a pedestrian interaction

ﬁeld as a gradient-ﬁlled ellipse, with ’X’ and ’Y’ represent-

ing horizontal and vertical coordinates. The gradient indi-

cates pedestrian awareness, decreasing from the center out-

ward. Arrows show the major axis (forward movement) and

minor axis (lateral space). The unawareness value increases

toward the periphery.

where a and b are the semi-major and semi-minor

axes. For an ellipse centered at (x

, y

) and rotated

by an angle θ, this equation is adapted to represent

pedestrian positioning and orientation.

• Center (x

, y

): Indicates the pedestrian’s current

position, serving as the geometric center of the el-

lipse.

• Major axis (2a): The longest axis, aligned

with the pedestrian’s primary direction of move-

ment, indicating forward motion and intended

path length.

• Minor axis (2b): The shorter axis, perpendicular

to the major axis, representing lateral movement

and the pedestrian’s width in the ﬁeld.

• Angle of rotation (θ): Deﬁnes the orientation of

the ellipse relative to the coordinate frame, align-

ing the major axis with the pedestrian’s movement

direction for accurate path projection.

• Color Gradient: The gradient within the ellipse

represents the pedestrian’s level of awareness, de-

creasing from the center to the edge, with un-

awareness increasing outward. These parame-

ters help autonomous vehicles dynamically adjust

navigation for pedestrian safety.

4 INTEGRATING INTERACTION

FIELDS AS AN ENTITY OF

ASPECT MAPS

The design of interaction ﬁelds uses aspect maps, a

framework developed by (Zolynski, 2018) that ap-

plies cognitive processes in robotics. Unlike tra-

ditional object-oriented methods, it employs paral-

lel data ﬂow networks inspired by biological visual

systems. Aspect maps are modular structures that

represent speciﬁc aspects of the environment, inte-

grating data from sources like sensors, storage, and

expert knowledge. They unify and abstract data,

enabling question-solving, information extrapolation,

and contradiction detection. This approach empha-

sizes modular, reusable solutions with multilevel ab-

stractions and supports early data uniﬁcation, improv-

ing processing, transparency, and reliability. Interac-

tion ﬁelds for vehicles and pedestrians are represented

within this framework for seamless correlation and in-

tegration.

4.1 Vehicle Interaction Field Mapping

The interaction ﬁeld around the vehicle is shaped us-

ing B

ezier curves, as shown in Figure 4, offering an

efﬁcient design. The cubic B

ezier curve with four

control points is expressed as:

B(t) = (1 − t)

+ 3(1 − t)

+ 3(1 − t)t

(2)

where:

• B(t) denotes a point on the curve.

• P

, P

, and P

are the control points.

• t ranges from 0 to 1.

At t = 0, the curve starts at P

, and at t = 1, it ends

at P

, passing smoothly through intermediate control

points.

The construction of interaction ﬁelds around the

vehicle involves a multi-step process using B

ezier

curves and gradient maps to deﬁne and reﬁne spatial

boundaries. Each step progressively shapes the in-

teraction zones, ensuring efﬁcient vehicle navigation

and pedestrian safety. Figure 4 illustrates the steps

involved in creating the interaction ﬁeld, as detailed

below:

1. B

ezier Curves (Step 1): The interaction ﬁeld is

generated by extending two B

ezier curves from

the vehicle’s front to its left and right sides. These

curves shape the interaction boundaries on both

sides of the vehicle, as illustrated in the top left

and bottom left insets. The B

ezier curves act

Adaptive Interaction Field Framework for Risk-Aware Navigation of Driverless Minibus in Pedestrian Zones

385

Figure 4: The ﬁgure illustrates the step-by-step process for

creating a vehicle interaction ﬁeld design.

as ﬂexible boundaries that are created on a grid,

deﬁning the ﬁeld’s spatial extent.

2. Filtered Boundaries (Step 2): The left and right

ﬁltered boundaries are determined on a grid, with

speciﬁc regions designated as null (indicated in

red). These boundaries establish the limits of in-

teraction ﬁelds, separating null zones (zero inter-

action) from active zones (one interaction).

3. Gradient Map Creation (Step 3): A gradient

map is created, as shown in the middle of the im-

age, indicating different zones of pedestrian in-

teraction with varying intensities. The gradient

progresses from red (low risk) near the vehicle

to green (high risk) at the periphery, representing

awareness levels within the interaction ﬁeld.

4. Thresholding and Composite Maps (Step 4 and

5): The gradient map undergoes a thresholding

process using the right and left ﬁltered boundaries

sequentially. During each thresholding step, red

zones (zero values) are assigned gradient values,

while non-colored zones are marked as NaN, de-

noting inactive areas. The sequential ﬁltering re-

ﬁnes the map, ultimately leading to a composite

map representing ﬁnalized interaction boundaries.

5. Final Interaction Field (Step 6): The uppermost

image shows the culmination of this process, pre-

senting the ﬁnal interaction ﬁeld map around the

vehicle. This map indicates varying interaction

zones based on proximity and provides the neces-

sary spatial structure for the vehicle’s navigation

system.

4.2 Pedestrian Interaction Field

Mapping

The vehicle interaction ﬁeld concept requires an anal-

ogous representation of pedestrians to understand

their unawareness and associated risks. Unlike the

vehicle interaction ﬁeld, this method uses ellipses,

as shown in Figure 3, to model pedestrian dimen-

sions and movement dynamics accurately. In Fig-

ure 5, pedestrian is depicted as yellow-green ellipse

centered based on their actual position relative to the

vehicle, which is positioned at the origin (0,0). This

alignment ensures precise spatial representation and

enhances the model’s ability to assess interactions and

collision risks. The ellipses’ orientation further indi-

cates pedestrian movement direction.

(a) Pedestrian walking in front of the vehicle.

(b) Aspect map representation of pedestrian awareness in

grip map (top view), where the minibus center is located at

(0, 0).

Figure 5: Grid representation of pedestrian space: Pedes-

trian occupancy as unawareness is shown as a yellow-green

ellipsoid, while unoccupied areas are marked in red. This

visualization quantiﬁes pedestrian dynamics and spatial in-

teraction on the grid.

Figure 6 presents a zoomed-in view of the pedes-

trian representation on the grid map, comparing

awareness levels between zero and one. The grid cells

transition from green (indicating higher awareness) to

VEHITS 2025 - 11th International Conference on Vehicle Technology and Intelligent Transport Systems

386

red (indicating lower awareness) based on grid val-

ues. This color scheme not only aids visualization but

also enhances the model’s ability to quantitatively as-

sess and represent pedestrian awareness, highlighting

areas of potential vulnerability and improving safety

insights in vehicle-pedestrian interactions.

(a) Pedestrian awareness with unawareness value 0, shown

as an isolated high-intensity area on a red background indi-

cating maximum awareness.

(b) Pedestrian awareness with unawareness value 1, show-

ing a gradient from green (low awareness) to red within the

ellipsoid on the grid map.

Figure 6: (a) Unawareness value of 0 shows a localized

high-awareness area in green, surrounded by a red back-

ground indicating maximum awareness. (b) Unawareness

value of 1 demonstrates a gradient from red (lower aware-

ness) to green (higher awareness) within the ellipsoid on the

grid.

To represent the collective awareness of multi-

ple pedestrians, their respective values are aggre-

gated. Figure 7 illustrates two pedestrians with re-

duced awareness near the vehicle, indicated by ar-

eas of higher intensity. When pedestrians’ unaware-

ness regions overlap, the values are summed, result-

ing in more saturated colors that reﬂect increased risk.

This approach captures and emphasizes heightened

risks from reduced awareness in overlapping pedes-

trian groups.

(a) Front camera image showing two pedestrians walking

ahead of the Autobus.

(b) Top view grid map showing the spatial occupancy of

both pedestrians.

Figure 7: The grid map visualizes two overlapping pedes-

trian spaces as ellipses on a red background. The intersec-

tion creates a value gradient, with the central yellow region

indicating the highest overlap. This composite view is es-

sential for analyzing pedestrian density, movements, and in-

teractions.

5 INTEGRATED

VEHICLE-PEDESTRIAN

INTERACTION FIELD

The analysis of merged values within interaction

ﬁelds is essential for understanding pedestrian-

vehicle dynamics, particularly with a driverless

minibus. In Figure 8, a pedestrian’s entry into the in-

teraction ﬁeld is depicted, marked by a black ellipti-

cal outline on the grid map, while the vehicle’s ﬁeld is

shown with a gray dotted line. A schematic of driver-

less minibus is overlaid for orientation, indicating its

position and direction.

The grid map uses a color gradient from red to yel-

low to represent increasing perceived risk, with yel-

low highlighting areas of higher merged awareness

values. As the pedestrian enters the interaction ﬁeld,

the model assesses their location and corresponding

awareness levels to compute the likelihood and sever-

ity of a potential encounter with the vehicle. This

Adaptive Interaction Field Framework for Risk-Aware Navigation of Driverless Minibus in Pedestrian Zones

387

Figure 8: The grid map illustrates the spatial risk assess-

ment between a pedestrian and the vehicle. The pedestrian’s

position is outlined in black, while the vehicle’s interaction

ﬁeld is marked with a gray dotted line. A schematic of the

vehicle provides orientation. The heat map, with a red-to-

yellow gradient, indicates increasing perceived risk, with

yellow denoting higher merged awareness values. These

values help evaluate the likelihood and severity of a poten-

tial encounter, enhancing understanding of how the pedes-

trian’s location inﬂuences the system’s risk assessment.

visualization allows the system to quantify and eval-

uate how the pedestrian’s spatial positioning affects

the system’s risk assessment, thereby enhancing the

understanding of dynamic pedestrian-vehicle interac-

tions.

Associated pedestrian risk is quantiﬁed by sum-

ming the grid values of the pedestrian’s ellipsoid and

the vehicle’s interaction ﬁeld. The grid is modeled

as a two-dimensional matrix, where each entry rep-

resents a speciﬁc position. The overlapping area, de-

noted as E for the pedestrian’s ellipsoid and V for the

vehicle’s interaction ﬁeld, is used to calculate the total

risk. The formula sums the values within this overlap,

capturing the combined effect of both entities on the

grid.

Let E be the set of grid positions occupied by the

pedestrian’s ellipsoid and V be the set of grid posi-

tions occupied by the vehicle’s interaction ﬁeld.

S =

∑

(i, j)∈E∩V

i j

) (3)

Where:

• S is the total sum of the overlapped grid values.

• E

i j

is the grid value at position (i, j) due to the

pedestrian’s ellipsoid.

• V

i j

is the grid value at position (i, j) due to the

vehicle’s interaction ﬁeld.

• E ∩ V represents the set of grid positions where

the ellipsoid and interaction ﬁeld overlap.

The ellipsoid grid values represent pedestrian un-

awareness, while the vehicle interaction ﬁeld values

signify potential risk. The combined sum of these

values at overlapping positions provides a composite

metric to assess risk in the interaction zone based on

predetermined thresholds for various ﬁelds.

These aggregated grid values are then sent to the

decision-making process, which evaluates the data to

determine if escape maneuvers are possible or if ac-

tivating the interaction module is necessary. This

decision-making phase integrates risk assessments to

enhance safety measures in scenarios requiring active

intervention.

Additionally, the calculated risk is utilized by

the interaction module, which classiﬁes it into

three types of interactions—ambient, direct, or risk-

based—using a threshold mechanism. This classiﬁca-

tion guides the module’s response to different levels

of pedestrian-vehicle interactions, ensuring an adap-

tive approach to managing potential hazards.

6 CONCLUSION

In conclusion, the concept of this work prioritizes ef-

fective risk assessment through adaptive interaction

ﬁelds aligned with the vehicle’s direction of travel.

This strategic alignment minimizes inappropriate in-

teractions, especially in less critical peripheral zones.

The pedestrian interaction area is oriented based on

the pedestrian’s walking path, establishing a dynamic

safety buffer that provides the driverless minibus with

ample time to respond to pedestrian movements.

The design’s ﬂexibility is emphasized by its cus-

tomizable features, allowing the size and intensity of

the interaction ﬁelds to be adjusted for various en-

vironmental conditions. The driverless minibus im-

proved its navigation by evading pedestrians when

possible and efﬁciently interacting with them based

on their assessed risk level. This adaptability en-

sures optimal conﬁguration across diverse scenarios,

enhancing safety in both crowded urban streets and

quieter suburban areas. The integration of these tech-

nical elements optimizes pedestrian-vehicle interac-

tions, promoting proactive risk management and situ-

ational awareness.

Future work can focus on enhancing interaction

ﬁelds by integrating advanced perception systems like

gaze tracking, adapting to diverse urban and environ-

VEHITS 2025 - 11th International Conference on Vehicle Technology and Intelligent Transport Systems

388

mental contexts, and addressing the needs of vulnera-

ble pedestrian groups. Real-world validation through

longitudinal studies in complex urban environments

and multi-vehicle scenarios can reﬁne the framework,

while incorporating cultural factors and smart infras-

tructure (e.g., V2X communication) can improve con-

textual awareness. Exploring energy-efﬁcient interac-

tion designs and conducting user acceptance studies

will further optimize the system for usability, trust,

and sustainability. These advancements aim to cre-

ate safer, adaptive, and globally effective autonomous

pedestrian-vehicle interaction strategies.

REFERENCES

Jan, Q. H., Baddela, Y. S., and Berns, K. (2022a). Pedes-

trian activity recognition from 3d skeleton data using

long short term memory units. In VEHITS, pages 368–

375.

Jan, Q. H. and Berns, K. (2021). Safety-conﬁguration of

autonomous bus in pedestrian zone. In VEHITS, pages

698–705.

Jan, Q. H. and Berns, K. (2023). Autonomous shuttle

for pedestrian zones. In International Symposium on

Measurements and Control in Robotics, pages 169–

179. Springer.

Jan, Q. H., Dahiwal, P., and Berns, K. (2023). Effec-

tive visual content of ehmi for autonomous vehicles

in pedestrian zones. In International Conference on

Robotics in Alpe-Adria Danube Region, pages 203–

210. Springer.

Jan, Q. H., Wolf, P., Berns, K., Reich, J., and Wellstein,

M. (2022b). Integration of human skeleton posture

models into reaction for realizing dynamic risk man-

agement. In International Commercial Vehicle Tech-

nology Symposium, pages 109–120. Springer.

Limprasert, W., Wallace, A., and Michaelson, G. (2013).

Real-time people tracking in a camera network. IEEE

Journal on Emerging and Selected Topics in Circuits

and Systems, 3(2):263–271.

Milan

es, V., Mahtout, I., Navas, F., and Gonz

alez, D.

(2020). On the passenger acceptance of driverless

shuttles. IEEE Intelligent Transportation Systems

Magazine, 14(1):18–28.

Nesheli, M. M., Li, L., Palm, M., and Shalaby, A. (2021).

Driverless shuttle pilots: Lessons for automated tran-

sit technology deployment. Case studies on transport

policy, 9(2):723–742.

Rasouli, A. and Tsotsos, J. K. (2019). Autonomous vehicles

that interact with pedestrians: A survey of theory and

practice. IEEE transactions on intelligent transporta-

tion systems, 21(3):900–918.

Schneemann, F. and Gohl, I. (2016). Analyzing driver-

pedestrian interaction at crosswalks: A contribution to

autonomous driving in urban environments. In 2016

IEEE intelligent vehicles symposium (IV), pages 38–

43. IEEE.

Vogel, D. and Balakrishnan, R. (2004). Interactive pub-

lic ambient displays: transitioning from implicit to

explicit, public to personal, interaction with multiple

users. In Proceedings of the 17th annual ACM sym-

posium on User interface software and technology,

pages 137–146.

Zolynski, G. (2018). Cognitive maps for autonomous ma-

chines in construction and civil engineering. Verlag

Dr. Hut.

Adaptive Interaction Field Framework for Risk-Aware Navigation of Driverless Minibus in Pedestrian Zones

389