journals.orclever.

com/ejrnd

Research Article

AI-Enhanced Automotive Navigation: Enriching Driving

Experiences through Intelligent Contextual Information
Delivery
Huseyin KARACALI1*, Nevzat DONUM2*, Efecan CEBEL3*

1TTTechAuto Turkey, Software Architect, Orcid ID: https://orcid.org/0000-0002-1433-4285, E-mail:

[email protected]
2 TTTechAuto Turkey, Embedded Software Engineer, Orcid ID: https://orcid.org/0000-0002-8293-8267, E-

mail: [email protected]
3 TTTechAuto Turkey, Embedded Software Engineer, Orcid ID: https://orcid.org/0000-0002-2027-0257, E-

mail: [email protected]
* Correspondence: [email protected]

(First received January 21, 2024 and in final form May 20, 2024)

Reference: Karacalı, H., Donum., N., Cebel, E. Cryptographic Enhancement of Named Pipes for
Secure Process Communication. The European Journal of Research and Development, 4(2), 110-
129.

Abstract
Improving in-car features and increasing the driving experience has become one of the main goals in the
automotive industry. Driver-specific adjustments, automatic operations, increased functionality of
infotainment systems and the development of automotive navigation systems, as discussed in this article,
have been factors that improve driving quality. This paper presents a novel approach to navigation system
development tailored for automotive use, aiming to enrich the driving experience through the integration
of artificial intelligence (AI) technologies. The proposed system operates by leveraging AI algorithms to
gather pertinent information about historical destination landmarks with temporal significance once a
driver selects it via the navigation interface. Moreover, the system not only provides information about the
selected destination, but also provides information about all historically important locations on the route,
depending on the location over GPS. Subsequently, this information is intelligently synthesized and
conveyed to the driver in a voice-assisted manner, enriching their journey with insightful details and
enhancing situational awareness. By seamlessly integrating AI-driven contextual information delivery into
the navigation paradigm, this system aims to not only facilitate smoother navigation but also elevate the
overall driving experience. Through a synthesis of AI, natural language processing, and location-based
services, this innovation represents a significant step towards more intuitive and enriching automotive
navigation systems.

Keywords: Navigation, automotive, in-car features, location-based services

Online ISSN: 2822-2296 journals.orclever.com/ejrnd 110

The European Journal of Research and
Development, 4(2), 2024 https://doi.org/10.56038/ejrmd.v4i2.433

1. Introduction
The automotive industry is undergoing a transformative phase, where the focus has
shifted from conventional transportation to enhancing the overall driving experience.
This evolution is driven by a growing demand for in-car features that not only provide
comfort and convenience but also integrate seamlessly into the driver's lifestyle. Central
to this evolution are advancements such as driver-specific adjustments, automated
operations, and the continuous improvement of infotainment systems. Moreover, the
development of sophisticated automotive navigation systems has emerged as a critical
area for enhancing the quality of driving experiences.

Amidst these advancements, the integration of AI technologies stands out as a pivotal

enabler of innovation in the automotive sector. AI-powered navigation systems, in
particular, hold immense promise in revolutionizing the way drivers interact with their
vehicles and the road. By leveraging AI algorithms, these systems have the potential to
offer personalized, contextually relevant guidance and insights, thereby enhancing both
the efficiency and enjoyment of driving.

This paper presents a new approach to developing a navigation system specifically

designed for automotive use, focusing on enriching the driving experience through the
integration of AI technologies. The basis of the proposed system is the use of AI
algorithms to collect data on historical destination landmarks of temporal significance.
By understanding the historical context of destinations, the navigation system can
provide drivers with comprehensive information beyond traditional route guidance.

Upon driver selection of a destination via the navigation interface, the system
dynamically retrieves information about historically significant locations along the route,
leveraging GPS-based location tracking. As locations of historical importance on the
route are approached at a certain distance, the collected data is prepared to be transferred
to the driver. Furthermore, the system employs voice-assisted information delivery to
ensure seamless integration into the driving experience, enhancing situational awareness
and minimizing distractions. The data collected by AI is transferred to the driver with the
speech feature. The driver can stop and restart this storytelling, which is under user
control, and can switch between detected historical locations. This is the control panel
provided to the user by the system, which works completely automatically with AI
support in the background.

Online ISSN: 2822-2296 journals.orclever.com/ejrnd 111

The European Journal of Research and
Development, 4(2), 2024 https://doi.org/10.56038/ejrmd.v4i2.433

There are studies in the literature that, although not similar, are close in concept and
compatible in purpose. In recent years, various studies on improving driving experience,
integrating AI into automotive, and providing online guidance about historical regions
are included in this paper. A similar technology was used in Vékony's study in 2016 and
a speech recognition mechanism was used in the car navigation industry. The text to
speech used in that work by Vékony was also used in the study in this paper [1].

Another related work is the mobile application called 'Piri Guide'. Thanks to ready-made
routes created in certain parts of the world, or while the user is actively traveling in
specified regions, it conveys information to the user about important points, historical
monuments, and structures in the vicinity. However, there is no AI integration in this
system. Since it works based on pre-recorded data, it is only valid in certain regions [2].

Throughout this paper, we delve into the methodology employed in developing the AI-
driven navigation system, emphasizing its potential to not only facilitate smoother
navigation but also elevate the overall driving experience. Through a synthesis of AI, text
to speech, and location-based services, this innovation represents a significant step
towards more intuitive and enriching automotive navigation systems.

The following sections provide a detailed review of the system's design, implementation,
and outcomes, as well as a comprehensive analysis of its functionality and effectiveness.
Additionally, the technologies and materials used throughout development are also
included. Additionally, the broader implications of the findings are discussed and
potential avenues for future research in this emerging field are outlined. Ultimately, this
article aims to contribute to the ongoing discourse on advancing automotive technology
towards smarter, user-centered solutions.

2. Materials
2.1. CPU
Specialized embedded systems differ from general-purpose computers in their tailored
design aimed at efficient operation within limited resources. Embedded CPUs are
customized to prioritize compactness, low power consumption, and integrated memory,
emphasizing quick response times and energy efficiency rather than high clock speeds.
Despite potentially lower clock speeds, these CPUs excel in performance for their
designated tasks [3]. The selection of CPU architectures significantly impacts the design
and functionality of embedded systems, driven by specific requirements and applications
[3].

Online ISSN: 2822-2296 journals.orclever.com/ejrnd 112

The European Journal of Research and
Development, 4(2), 2024 https://doi.org/10.56038/ejrmd.v4i2.433

In this investigation, the NXP i.MX 8QuadMax CPU serves as the central processing unit,
designed specifically for robust system integration in embedded applications. It utilizes
a heterogeneous multicore architecture, combining Arm Cortex-A72 and Arm Cortex-
A53 cores to efficiently manage high-performance and energy-efficient tasks [4]. For
graphic processing, it integrates a GC7000XSVX GPU capable of handling 4K resolution
videos and offering high performance for embedded graphics tasks [4]. The CPU also
offers extensive connectivity options, including dual PCIe interfaces, dual Gigabit
Ethernet, and various data communication ports, making it adaptable and connectivity-
oriented [4]. Its security features comply with industry standards, making it particularly
suitable for security-focused embedded applications.

2.2. Android Automotive Operating System

The Android Automotive Operating System (AAOS) is an adapted version of the
Android platform tailored explicitly for vehicle use. It offers an intuitive interface for in-
car infotainment and navigation systems, facilitating seamless integration of smartphone
features and services into the vehicle environment [5]. AAOS encompasses a diverse
array of functionalities, including media playback, navigation, communication, and
connected services, supporting various connectivity options like Bluetooth, Wi-Fi, and
cellular connectivity for effortless interaction with other devices and services [6]. Its
flexible architecture allows vehicle manufacturers to customize and tailor the user
experience according to their specific requirements. Furthermore, AAOS serves as a
platform for developers to design and distribute automotive-centric applications, thereby
augmenting the system's capabilities. In summary, the Android Automotive Operating
System presents a robust and adaptable solution for automotive infotainment and
navigation systems, delivering a rich and immersive experience for both drivers and
passengers.

Figure 1 represents the abstract layer architecture of Android Automotive Operating

System (AAOS). The top layer is related with the applications that anything the user can
interact with such as navigation, music, and messaging. The service layer provides core
services that the applications need to run, such as window management, resource
management, and security. Hardware Abstraction Layer (HAL) is a separator between
the operating system and the hardware. It allows the operating system to work with
different types of hardware without having to be rewritten for each type of hardware.
The last layer, Board Support Package (BSP) includes the Linux kernel to run Android
Automotive OS.

Online ISSN: 2822-2296 journals.orclever.com/ejrnd 113

The European Journal of Research and
Development, 4(2), 2024 https://doi.org/10.56038/ejrmd.v4i2.433

Figure 1: Abstract layer architecture of Android Automotive Operating System [7]

2.3. Organic Maps

Organic Maps is a free and open-source mobile application offering offline navigation
services with a focus on privacy and community-driven data [8]. Developed as a fork of
the popular ‘Maps.me’ application, it stands out with its commitment to user data
protection and its reliance on crowd-sourced data from OpenStreetMap (OSM) [9].

In addition to map display, its key features include transferring real-time traffic data,
providing navigation features by providing public transport routes, being open-source,
and being available offline [8]. Moreover, the application adheres to strict privacy
principles. It avoids user tracking, data collection, and third-party integrations [10].

Online ISSN: 2822-2296 journals.orclever.com/ejrnd 114

The European Journal of Research and
Development, 4(2), 2024 https://doi.org/10.56038/ejrmd.v4i2.433

Figure 2: Organic Map logo [8]

In this study, the integration of artificial intelligence (Gemini AI), which will provide data
collected about historical regions, and text to speech algorithms, which will voice this
data to inform the driver, has been made into the Organic Maps application. It was chosen
as the navigation basis of this study due to its open-source development and wide scope.

2.4. Gemini AI
In the ever-evolving landscape of AI, Google's Gemini AI emerges as a groundbreaking
technology, pioneering a new era of "multimodal AI." Unlike traditional models confined
to individual modalities like text or images, Gemini seamlessly integrates information
across text, images, and audio, unlocking richer context analysis and reasoning
capabilities.

Departing from the limitations of unimodal approaches, Gemini's core strength lies in its
ability to process and understand information from various modalities simultaneously
[11]. This enables the model to handle complex scenarios requiring comprehension of
diverse information sources, paving the way for advancements in various fields like
visual question answering, multimodal search, and machine translation. Furthermore,
the end-to-end training approach optimizes information flow across modalities, fostering
comprehensive understanding [12].

In this study, data about historical locations up to the destination are obtained online
with Gemini AI. Thanks to the integration of the AI algorithm into the Organic Maps
navigation application, locations on the route were detected and these locations were sent

Online ISSN: 2822-2296 journals.orclever.com/ejrnd 115

The European Journal of Research and
Development, 4(2), 2024 https://doi.org/10.56038/ejrmd.v4i2.433

sequentially as input to Gemini AI. It was preferred because of its fast response time, its
popularity close to the date of the study, its ease of configuration, and its ease of
implementation into projects.

2.5. TextToSpeech
The TextToSpeech (TTS) application programming interface (API) in the Android SDK
empowers developers to seamlessly integrate speech synthesis capabilities into their
applications. This enables devices to "speak" textual content, enhancing accessibility, user
interaction, and information delivery. Text-to-Speech conversion by ensuring proper
pronunciation and inflection, fine-tuning speech characteristics, and queue management,
offline speech synthesis, and third-party TTS engines are the key features of this API [13].

Integration and development of this API requires several steps. Firstly, the TTS API is
added into the project by including necessary permissions and dependencies. Then, an
instance for TextToSpeech is created and potential speech data download. The speak
method from this API specifies the text to speech by synthesis. Finally, the events should
be handled like successful synthesis completion or errors to manage speech playback [14].

2.6. Android Studio

Android Studio serves as the primary integrated development environment (IDE)
specifically crafted for Android app development. This sophisticated tool, endorsed by
Google, offers an extensive array of features and resources to streamline the app
development workflow. Android Studio amalgamates efficiency, flexibility, and robust
functionality, making it a preferred choice among developers embarking on Android
application projects [15]. Its user interface is intuitive and user-friendly, facilitating
seamless navigation and access to various tools, thus promoting productivity within the
development workspace.

The IDE incorporates an advanced code editor that supports multiple programming
languages, with a primary focus on Java and Kotlin for Android development. These
features enrich the coding experience by providing functionalities like code completion,
syntax highlighting, and real-time error checking [15].

The development process of the AG application was conducted using Android Studio,
the officially endorsed IDE for Android app development. Android Studio was selected
owing to its status as the official IDE, offering a comprehensive suite of features and

Online ISSN: 2822-2296 journals.orclever.com/ejrnd 116

The European Journal of Research and
Development, 4(2), 2024 https://doi.org/10.56038/ejrmd.v4i2.433

resources to expedite the development of Android applications [15]. Employing Android

Studio ensures smooth integration with the latest Android SDKs, APIs, and platform
updates, guaranteeing optimal performance and utilization of the latest Android OS
features [15].

The choice of Java as the programming language for the AG application is driven by the
necessity for platform independence and the widespread adoption of Java in Android
development. Java's principle of "write once, run anywhere" aligns with the objective of
ensuring that the AG application operates seamlessly across various devices and
operating systems, providing a consistent user experience. The extensive support and
libraries available within the Java community simplify the development process of the
AG application, enhancing its reliability.

In conclusion, the AG application was developed utilizing Android Studio as the

designated IDE, leveraging its official status and robust features tailored for Android
development. Adopting Java as the programming language ensures platform
independence and benefits from the extensive support within the Android development
community. This strategic approach aims to create a potent and user-friendly AG
application capable of delivering a consistent experience across diverse platforms.

3. Method
This work was carried out entirely on a previous open-source navigation application,
Organic Maps, and therefore, it was first aimed to obtain the source codes of this project
and compile it for the Android-based operating system. Current source codes were
downloaded from the project's shared repository web page. In order to develop this
project, there must be a study in which we will have the source codes, because artificial
intelligence, text voice-over, and other remaining modules will be added to these source
codes and the study will be completed. The navigation application, whose source codes
were downloaded, was first opened as an Android Studio development environment,
and compiled for the Android Automotive operating system.

The next stage of the study is to examine these source codes to obtain the necessary
information. The purpose of this review is to determine in which method and how the
project uses the data, the current location and the latitude and longitude locations of the
selected destination, which will be given as a prompt to the AI model in the future and
enable it to receive the necessary information. Thanks to the gemmulation method of the
LocationHelper class, the current location of the vehicle can be read in terms of latitude

Online ISSN: 2822-2296 journals.orclever.com/ejrnd 117

The European Journal of Research and
Development, 4(2), 2024 https://doi.org/10.56038/ejrmd.v4i2.433

and longitude. These read values are converted from decimal to degree, minute, and
second as shown in Figure 3 for artificial intelligence to better understand. Obtaining this
data was appointed as the first step of development as it will be used in the prompt that
will be given to AI in the future.

Figure 3: Log messages for read current location

The next and always repeated step in the study is based on the system exchanging data
with AI and presenting the coordinates to the AI in a certain structure and receiving
information from it. Therefore, the first technical process for this is to apply the AI
integration to the project. Gemini AI's application programming interface (API), which
has up-to-date data and has a much higher response time than other AI robots, was used
to be integrated into the project. Transactions made using AI cannot be used without
starting navigation. Since running this system before a route is created for navigation
would be unnecessary workload, the connection with AI is activated after a route is
planned. To create the route, a destination must be selected on the application and click
on the 'Route To' button. The application creates a driving route from the current location
(obtained in the previous step) to the selected destination, as traditional navigation
applications do. With this operation performed with the onNavigationStarted function,
several interface operations are also performed.

Creating the route and starting navigation also enables the integrated AI system to be
activated. The function that starts the artificial intelligence is also called within the
function that starts this navigation. This function is the part where the system exchanges
data with all AI, starting from the integration of AI. The first part of this data exchange is
the current location of the vehicle (in degree, minute and second), which we previously
obtained from the application, is sent to the AI with a prompt like this: “"list the historical
places near these coordinates:" + transformedCoordinates + ". write every place side by
side with '*' between them. do not explain them. do not start your response with '*'"”. As
seen in the prompt, it is desired to list historical points close to the given coordinate. It is
very important that the received message be in a certain format so that it can be parsed
and listed by the system, so these historical locations are required to be separated from
each other with the * sign. It is also specifically mentioned to the AI side that it does not
give any unnecessary information and does not make any statements as it is only
intended to share the names of historical landmarks. In this section, it is aimed to obtain

Online ISSN: 2822-2296 journals.orclever.com/ejrnd 118

The European Journal of Research and
Development, 4(2), 2024 https://doi.org/10.56038/ejrmd.v4i2.433

and list the names of historical points currently nearby the vehicle. Figure 7 represents
the response of the AI for the prompt given above. The near historical places around the
current position (named as transformedCoordinates in the project) are listed and
separated with the * sign.

Figure 4: Location name parser finite automaton

Thanks to the developed algorithm, the response shown in Figure 7 is separated item by
item using a regular expression and assigned to a string array by using a parser that
designed as shown in Figure 4. Designed to comply with the given prompt, this parser
can terminate at an error (or empty) or successful parsing node. If any character other
than a letter is encountered in the start node, an abnormal situation is detected and the
parser returns an error to the relevant functions. In the parsing process, which proceeds
in its normal flow, a new location will appear where the asterisk (*) sign is seen, so it
returns to the start node. The same location data continues to be received until the asterisk
sign appears. If a null character (\0) comes after a letter, the parser has completed its
operation successfully and the historical locations are appropriately transferred into the
string array. Thus, a list of historical places close to the given location is obtained. Then,
the data that will be asked to be sent to us by AI is the distance. It is of great importance
how far the mentioned historical locations are from our current location, because in order
for the system to transfer these historical locations to the driver, they must be closer than

Online ISSN: 2822-2296 journals.orclever.com/ejrnd 119

The European Journal of Research and
Development, 4(2), 2024 https://doi.org/10.56038/ejrmd.v4i2.433

a certain threshold value. This list of historical locations, kept in a string array, is sent to
the AI via a different prompt. This prompt is as follows: "give me coordinates of these
locations:" + Arrays.toString(locationList) + " in decimal format. write every coordinate
side by side without space. write '*' after every coordinate pair. give the coordinates in
decimal form. write ',' between every coordinate pair. do not leave any space. example
output: '37.22,122.3*37.12,122.12'". As can be seen in this prompt, decimal type
coordinates of the locations whose names are given in the string array are requested. The
main purpose of requesting these latitude and longitude values arises from the need to
measure the distance between the instantaneous vehicle position and these historical
regions. These historical region names, kept under the name locationList, are sent to AI.
Just like these names, a parsing process will be applied, so no spaces are required.

Figure 5: Location coordinate parser finite automaton

In addition, it is requested that there be a "," sign between latitude and longitude, and a *
sign is also required between coordinates. Similar to the previous parsing process, a
parsing is applied to the coordinates received by the AI and these coordinates are
assigned to a string array as shown in Figure 5. The parser algorithm prepared for
coordinates is similar to the previous one, but slightly more complex. There is a comma

Online ISSN: 2822-2296 journals.orclever.com/ejrnd 120

The European Journal of Research and
Development, 4(2), 2024 https://doi.org/10.56038/ejrmd.v4i2.433

separating both latitude and longitude, and a dot separating the decimal part for both
values. Also, as in the previous parser, the asterisk is used to separate coordinates. The
parser, which begins at the start node, returns the function as an error or empty list unless
it encounters a numerical value. In normal operation, the decimal part of the latitude
value is read until the dot sign is seen, and the fractional part of the latitude is read after
the dot. Since there will be a comma between latitude and longitude, when the parser
detects the comma, the same operation for latitude is applied to longitude. With the
asterisk sign, the coordinate for a location is read successfully and the automata returns
to the start node to read the coordinate of the new location. When there is a null character
instead of an asterisk, the parser successfully completes the coordinate reading process.
As a result of this process, both the historical locations in the vicinity and the latitude and
longitude of those locations are kept in separate string arrays. The AI's response to this
prompt is shown in Figure 8 below.

As shown in Figure 8, the received coordinate data is stored in a string array. These
coordinates are used to calculate how far the current vehicle location is from historical
locations specified by the AI. Using the coordinate distance finding function, the distance
to each historical place is measured and these distance values are recorded in an array.

At this stage of the study, sharing of coordinates and historical or cultural location data
between the AI integrated with the navigation application is terminated for a while. In
this part, it is expected that the study will provide constant control of the data it receives
and transfer data about locations to the driver when appropriate conditions occur. It
includes distance control, collecting summary information about the relevant location by
the AI engine, and transferring this collected text to the driver with the help of Text-to-
Speech. The function that performed the AI operations up to this point is exited and a
different function is called. Location names kept in the form of an array, distance data to
the locations kept in the form of an array, the integrated AI model, and the executer object
that enables the task to run are given as arguments to this function. Distance
measurements kept in an array are examined sequentially by this called function. Starting
from the first element of the array, it is checked whether the vehicle is closer than 5
kilometers to its current location. In this case, if the distance is more than 5 kilometers,
the next element of the distance array is checked without any action. All values are
examined iteratively unless there is a distance of less than 5 kilometers.

On the other hand, if one of the important places specified by the AI is closer than the
specified threshold value distance, that is, if any element of the distance array is less than

Online ISSN: 2822-2296 journals.orclever.com/ejrnd 121

The European Journal of Research and
Development, 4(2), 2024 https://doi.org/10.56038/ejrmd.v4i2.433

5 kilometers, the function that will exchange data with the AI is called again with the
prompt given below and a summary description text about the relevant location name is
requested: “"tell me about " + locationList[locationCounter]”. Since the names array of
historically important locations and the distance array are in the same order, the name of
the nearby point can be found with the same index of the element detected in the distance
array. Thus, thanks to this prompt, general summary information about the selected
location is requested by the AI. Figure 9 shows an example of the AI's answer to this
prompt.

Online ISSN: 2822-2296 journals.orclever.com/ejrnd 122

The European Journal of Research and
Development, 4(2), 2024 https://doi.org/10.56038/ejrmd.v4i2.433

Online ISSN: 2822-2296 journals.orclever.com/ejrnd 123

The European Journal of Research and
Development, 4(2), 2024 https://doi.org/10.56038/ejrmd.v4i2.433

Figure 6: Flowchart of the storytelling system

The next phase in the study is to inform the driver and other passengers in the vehicle
about the important location, which is close enough and whose summary information is
received in text form. This entire text is transmitted to the integrated Text-to-Speech
module and all information is transmitted to the driver at the specified speed and in
English. The length and scope of this process may vary depending on the depth of
information provided by the AI model. When the Text-to-Speech process is completed,
the relevant location and distance data of the location are removed from the arrays.
Additionally, the function that checks the distance values is called recursively to decide
whether to read subsequent historical landmarks. This recursive call continues until there
is no unchecked location in the list.

As the last component of the system, since the vehicle may be moving, the current
coordinates of the vehicle are taken periodically every 5 minutes and all operations are
performed from the beginning, the lists are renewed and the process is repeated.

Figure 6 represents the general flow of the whole system for intelligent storytelling
development on Organic Maps navigation application for the Android Automotive OS.
The operations and transitions of the entire system according to the conditions can be
observed through this figure.

The last development made within the scope of the project is to enable the user to control
this working system on the narrative side. Three new buttons have been integrated into
the default interface of the navigation application chosen as the base. These buttons,
respectively, allow information to be transferred about the previous historical location in
the list, to stop the narration with text to speech or to start if it has been stopped, and to
transfer information about the next historical location in the list. The transition between
stop and start buttons has been transferred to the UI side in a convenient way for the user.

Historical locations provided by AI and named by parser are kept in a list called
narration. AI-generated text content about these locations is kept in exactly the same
order as in this list. Thanks to a common index value, the transitions, stop, and start
functions provided in this latest development can be used. When moving to the previous
or next one, this index value is manipulated and moved to the next in the list. If the end
of the list is reached, a warning system has been prepared to inform the user.

4. Results

Online ISSN: 2822-2296 journals.orclever.com/ejrnd 124

The European Journal of Research and
Development, 4(2), 2024 https://doi.org/10.56038/ejrmd.v4i2.433

The study and development which are designed for an in-car entertainment-information
system that is integrated with artificial intelligence, only outputs the speech of the text
produced by the AI model about the location that meets the criteria and has been selected
by the algorithm at that moment. Therefore, the results that can actually be stated in this
paper are the answers produced by the AI in the system until it reaches that speech to
text.

Figure 7: AI's response on nearby historical locations’ name prompt

After receiving the current location of the vehicle, the first output of the system can be
shown as the first output of the system to return the nearby historically important
locations in a format with a prompt to the AI model as shown in the Figure 7. The process
of listing nearby locations according to current coordinates is repeated periodically until
the navigation route is completed or finished by the driver.

The second output is the latitude and longitude coordinates of the locations listed in
Figure 7 by the AI model. These outputs presented in Figure 8 are important since the
distances between the vehicle and its current location will be measured using these
coordinates. These outputs are also returned by the AI in a certain format so that it can
be parsed.

Online ISSN: 2822-2296 journals.orclever.com/ejrnd 125

The European Journal of Research and
Development, 4(2), 2024 https://doi.org/10.56038/ejrmd.v4i2.433

Figure 8: AI's response on nearby historical locations’ coordinates prompt

As the final output of the data exchange between the AI model and the navigation
application, AI summary information is transferred about one of the locations in the list
that is determined to have a certain level of proximity. This text is the text that will be
conveyed to the driver and other passengers with speech. An example return of AI model
for San Jose Museum of Art is shown in Figure 9.

Figure 9: AI's response on nearby historical locations’ description prompt

As a final result, a snapshot can be presented where information is transferred thanks to

AI, accompanied by an actively working road route. The buttons seen at the top right are
the panel that performs transition and stopping operations and can be controlled by the
user. Additionally, the application works in both vertical and horizontal frames.

Online ISSN: 2822-2296 journals.orclever.com/ejrnd 126

The European Journal of Research and
Development, 4(2), 2024 https://doi.org/10.56038/ejrmd.v4i2.433

Figure 10: A snapshot from navigation when storytelling is active

5. Discussion and Conclusion

In conclusion, this paper presents a novel approach to enhancing the driving experience
through the integration of AI technologies into the development of an automotive
navigation system. The proposed system, implemented on the Organic Maps navigation
application for the Android Automotive Operating System, leverages AI algorithms to
provide drivers with insightful information about historically significant landmarks
along their route. The integration of Gemini AI, a multimodal AI technology, along with
Text-to-Speech capabilities, enriches the journey by delivering historical context and
details about important locations.

The developed system successfully achieves its objectives, seamlessly integrating AI-
driven contextual information delivery into the navigation paradigm. By periodically
collecting and processing data about nearby historical locations, the system enhances
situational awareness and provides an engaging and informative driving experience. The
integration of the Android Automotive Operating System, NXP i.MX 8QuadMax CPU,

Online ISSN: 2822-2296 journals.orclever.com/ejrnd 127

The European Journal of Research and
Development, 4(2), 2024 https://doi.org/10.56038/ejrmd.v4i2.433

and Organic Maps application forms a robust foundation for the implementation of this
innovative navigation system. The Android Automotive Operating System provides a
flexible and customizable platform for in-car infotainment and navigation systems,
allowing seamless integration of diverse functionalities. Organic Maps, chosen for its
open-source development and privacy-focused approach, serves as a suitable base for
incorporating AI-driven features.

Moreover, the integration of Gemini AI introduces a multimodal aspect, enabling the

system to process information across text, images, and audio simultaneously. This not
only enhances the richness of information provided to the driver but also sets the stage
for potential future advancements in other multimodal applications.

The system's functionality, as outlined in the methodology section, successfully captures

the essence of providing historical information during the driving experience. The user
interface modifications, including the addition of control buttons for narration, contribute
to user interaction and control. The periodic updates of the current vehicle coordinates,
coupled with the AI-driven storytelling mechanism, create a dynamic and engaging
experience for the driver. The ability to control the narration and obtain information
about historical landmarks adds a layer of personalization to the navigation system.

Despite the achievements, there are certain limitations to the current implementation. The
system relies on the periodic retrieval of historical information, and the level of detail
provided by AI may vary. Future iterations could explore real-time data streaming for a
more dynamic experience. Additionally, the system's reliance on GPS coordinates may
face challenges in dense urban environments or areas with limited historical data. Future
research could investigate alternative methods, such as incorporating sensor fusion
technologies, to enhance location accuracy. The user interface and control buttons
introduced in this development are designed for practicality, but further user testing and
feedback could refine the user experience. Incorporating user preferences and expanding
language support could contribute to a more inclusive and personalized system.

6. Acknowledge
We want to express our appreciation to Huseyin Karacali, the Software Architect, for his
adept guidance and inspiring leadership. Furthermore, we recognize the precious
support offered by TTTech Auto Turkey during the project's development phases.

References

Online ISSN: 2822-2296 journals.orclever.com/ejrnd 128

The European Journal of Research and
Development, 4(2), 2024 https://doi.org/10.56038/ejrmd.v4i2.433

[1] Vékony, A. (2016). Speech Recognition Challenges in the Car Navigation Industry. In: Ronzhin,
A., Potapova, R., Németh, G. (eds) Speech and Computer. SPECOM 2016. Lecture Notes in Computer
Science(), vol 9811. Springer, Cham. https://doi.org/10.1007/978-3-319-43958-7_3

[2] Piriguide, “Piri Guide – Travel planner,” App Store, https://apps.apple.com/tr/app/piri-guide-

travel-planner/id1095698831.

[3] “Central Processing Unit,” Central Processing Unit - an overview | ScienceDirect Topics,
https://www.sciencedirect.com/topics/engineering/central-processing-unit.

[4] “I.MX 8 Family applications processor: ARM cortex-A53/A72/M4,” i.MX 8 Family Applications
Processor | Arm Cortex-A53/A72/M4 | NXP Semiconductors, https://www.nxp.com/products/processors-
and-microcontrollers/arm-processors/i-mx-applications-processors/i-mx-8-applications-processors/i-mx-
8-family-arm-cortex-a53-cortex-a72-virtualization-vision-3d-graphics-4k-video:i.MX8.

[5] “What is Android Automotive?  :   Android Open Source Project,” Android Open
Source Project. [Online]. Available:
https://source.android.com/docs/devices/automotive/start/what_automotive. [Accessed: 01-Feb-2023].

[6] “Design for driving  |  google developers,” Google. [Online]. Available:
https://developers.google.com/cars/design/automotive-os?hl=tr. [Accessed: 01-Feb-2023].

[7] H. Hofmann et al., “Android Automotive OS Whitepaper: Android Automotive OS Book,”

Android Automotive OS Book | Build your infotainment system based on Android Automotive OS,
https://www.androidautomotivebook.com/android-automotive-embedded-os-whitepaper/.

[8] Organic Maps. (2023, Jan 1). Organic Maps: Open-source, offline maps: https://organicmaps.app/.

[9] Organic Maps: Open-source, offline maps on GitHub:

https://github.com/organicmaps/organicmaps.

[10] OpenStreetMap. (n.d.). OpenStreetMap. Retrieved from https://www.openstreetmap.org/.

[11] LaMDA, et al. (2023). Gemini - Google DeepMind. https://deepmind.google/technologies/gemini/.

[12] Baptista, J., et al. (2023). Introducing Gemini: Google's most capable AI model yet. Google AI
Blog. https://developers.googleblog.com/2023/12/how-its-made-gemini-multimodal-prompting.html.

[13] Android Developers Documentation:

https://developer.android.com/reference/android/speech/tts/TextToSpeech.

[14] Implementing Text-to-Speech in Android:

https://www.tutorialspoint.com/android/android_text_to_speech.htm.

[15] “Android Developers: Android Mobile App Developer Tools.” https://developer.android.com/.

Online ISSN: 2822-2296 journals.orclever.com/ejrnd 129