=Paper=
{{Paper
|id=Vol-3736/paper8
|storemode=property
|title=Information technology for joint decision making in machine embroidery with means of augmented reality
|pdfUrl=https://ceur-ws.org/Vol-3736/paper8.pdf
|volume=Vol-3736
|authors=Iryna Zasornova,Tetiana Hovorushchenko,Mykola Fedula,Alina Hnatchuk
|dblpUrl=https://dblp.org/rec/conf/icyberphys/ZasornovaHFH24
}}
==Information technology for joint decision making in machine embroidery with means of augmented reality==
https://ceur-ws.org/Vol-3736/paper8.pdf
Information technology for joint decision making in
machine embroidery with means of augmented reality
⋆
Iryna Zasornova1,∗,†, Tetiana Hovorushchenko1,†, Mykola Fedula1,† and Alina
Hnatchuk1,2,†
1 Khmelnytskyi National University, Instytutska 11, Khmelnytskyi, 29016, Ukraine
2 Prague University of Economics and Business, Winstona Churchilla str., 1938/4, Prague, 13067, Czech Republic
Abstract
The described information technology for joint decision-making in machine embroidery leverages
augmented reality (AR) and artificial intelligence (AI) to significantly enhance the design and
production process. AR enables real-time visualization of embroidery patterns projected onto
garments, allowing designers and stakeholders to interactively see and adjust designs directly on the
clothing. This immersive visualization helps in identifying design flaws and making immediate
adjustments, ensuring the design aligns perfectly with the intended outcome.
The AI system processes a comprehensive set of input data, including the input video signal from
cameras, initial embroidery images, the geometry of the human body and clothes, general design
requirements, and expert recommendations. The AI system uses this data to iteratively refine the
embroidery design, continuously improving the quality and precision of the embroidery through
sophisticated neural network models. The AI-generated correction signal is fed back into the AR
system, updating the virtual embroidery projection in real-time. This dynamic feedback loop ensures
that the design is constantly optimized to meet both aesthetic and technical standards. The
integration of AR and AI fosters seamless collaboration among designers, production teams, and
clients, allowing efficient communication and consensus building. This results in higher quality, more
precise, and personalized embroidery designs, ultimately leading to more efficient production
processes and enhanced customer experience.
Keywords
cyber-physical system, information technology, digital embroidery, augmented reality, optimization
1. Introduction
The advent of Industry 4.0 has changed traditional manufacturing processes significantly,
including machine embroidery. Cyber-physical systems (CPS) integrate computational and
ICyberPhyS-2024: 1st International Workshop on Intelligent & CyberPhysical Systems, June 28, 2024, Khmelnytskyi,
Ukraine
∗ Corresponding author.
† These authors contributed equally.
zasornovair@khmnu.edu.ua (I. Zasornova); tat_yana@ukr.net (T. Hovorushchenko); mailfm2000@gmail.com (M.
Fedula); alinahnatchuk@ukr.net (A. Hnatchuk)
0000-0001-6655-5023 (I. Zasornova); 0000-0002-7942-1857 (T. Hovorushchenko); 0000-0002-3765-2016 (M. Fedula);
0000-0003-0155-9255 (A. Hnatchuk)
© 2023 Copyright for this paper by its authors. Use permitted under Creative Commons License Attribution 4.0 International (CC BY 4.0).
CEUR
ceur-ws.org
Workshop ISSN 1613-0073
Proceedings
physical processes, enabling seamless interaction between humans, machines, and data.
Information technologies play a crucial role in facilitating joint decision-making within these
systems, enhancing efficiency, quality, and innovation in machine embroidery. The real-time
data analytics [1] is pivotal in CPS for machine embroidery. Internet of Things (IoT) sensors
embedded in embroidery machines collect data on machine performance, thread tension, stitch
quality, and operational status. This data is transmitted to a central system where it is analyzed
to provide actionable insights. Decision-makers can monitor production in real-time, identify
issues, and make informed decisions to optimize operations.
The modern cloud computing technologies provide a robust platform for collaborative
decision-making. By centralizing data storage and processing, cloud platforms enable multiple
stakeholders, including designers, production managers, and quality control teams, to access
and share information simultaneously. This fosters a collaborative environment where
decisions can be made based on the latest data and insights, ensuring coherence and alignment
across the production process.
AI and machine learning (ML) algorithms are useful to optimizing decision-making in CPS.
Predictive maintenance algorithms analyze historical data to forecast machine failures, allowing
for proactive maintenance scheduling. ML models can also optimize embroidery patterns by
analyzing previous designs and production outcomes, suggesting improvements to enhance
quality and efficiency.
AR systems enhance joint decision-making by providing a visual and interactive platform
for design and production planning. Designers can use AR to visualize embroidery designs on
garments in a 3D space, making real-time adjustments collaboratively with other stakeholders.
AR tools enable virtual prototyping, reducing the need for physical samples and accelerating
the design approval process.
It should be noted that collaborative platforms, such as cloud-based project management
tools, facilitate communication and coordination among teams. These platforms enable
stakeholders to share updates, provide feedback, and track project progress in real-time.
Features like version control, task assignments, and milestone tracking ensure that everyone is
aligned and informed, leading to more effective joint decision-making.
Thus, information technologies can improve machine embroidery by enabling more efficient
and effective joint decision-making within cyber-physical systems. Real-time data analytics,
cloud computing, AI, AR and collaborative platforms collectively enhance the ability of
stakeholders to make informed decisions, optimize processes, and innovate continuously. As
Industry 4.0 continues to evolve, these technologies will play an increasingly important role in
driving the future of machine embroidery.
2. The known joint decision making technologies in machine
embroidery
In the context of Industry 4.0, machine embroidery has been revolutionized through various
information technologies that enhance joint decision-making. These technologies include
real-time data analytics, cloud computing, AI, ML, AR, blockchain, and digital twins, each
playing a critical role in the decision-making process.
2.1. Real-time data analytics
Real-time data analytics techniques play a crucial role in information technologies for joint
decision-making in machine embroidery, leveraging various methodologies to optimize
production processes, enhance design quality, and ensure efficient operation. These
techniques utilize a range of data sources, including IoT sensors, ML algorithms, and
advanced data processing tools, to provide actionable insights. One of the primary
techniques used is sensor data collection and monitoring. IoT sensors embedded in
embroidery machines collect real-time data on various parameters such as thread tension,
stitch density, machine speed, and operational status. This data is transmitted to a central
system where it is processed and analyzed. For instance, sensors can detect anomalies in
thread tension or stitch patterns, prompting immediate adjustments to prevent defects and
ensure consistent quality. Real-time monitoring allows for proactive maintenance, reducing
downtime and extending machine lifespan. Studies have shown that integrating IoT sensors
in manufacturing can significantly enhance process efficiency and product quality [1].
Data visualization tools are also useful in real-time data analytics. These tools transform
complex data sets into easily interpretable visual formats such as graphs, charts, and
dashboards. Real-time dashboards provide a comprehensive view of the production
process, highlighting key performance indicators and alerting stakeholders to any
deviations from the norm. This immediate visibility enables quick decision-making and
fosters collaboration among team members. Visualization tools are particularly valuable in
joint decision-making as they ensure that all stakeholders, regardless of technical expertise,
can understand and act on the data [2]. Advanced statistical methods and algorithms are
employed to analyze the data and derive insights. Techniques such as regression analysis,
anomaly detection, and time-series analysis are used to understand trends and make
predictions. For instance, regression analysis can identify the relationship between machine
speed and stitch quality, enabling optimization of machine settings. Anomaly detection
algorithms can identify outliers in the data that may indicate potential issues, allowing for
quick intervention. Time-series analysis helps in understanding how production metrics
evolve over time, providing insights into seasonal trends and helping in capacity planning
[3].
In joint decision-making, these real-time data analytics techniques facilitate informed
and timely decisions. By providing a continuous stream of actionable insights, stakeholders
can collaborate effectively, address issues promptly, and optimize production processes.
The integration of these techniques into machine embroidery systems represents a
significant advancement, driving the industry towards greater efficiency, quality, and
innovation.
2.2. Cloud computing
Cloud computing techniques impact significantly the joint decision-making in machine
embroidery, leveraging advanced data storage, processing, and collaboration capabilities to
enhance efficiency and innovation. One of the primary techniques is data centralization,
where cloud platforms store vast amounts of data collected from embroidery machines and
other sources. This centralization ensures that all stakeholders have access to the same
data, facilitating real-time collaboration and decision-making. Cloud platforms such as
Amazon Web Services (AWS), Microsoft Azure, and Google Cloud provide scalable storage
solutions that can handle the large volumes of data generated in modern embroidery
processes. This scalability is crucial for accommodating fluctuating workloads and ensuring
that data is readily available when needed [4]. Another key cloud computing technique is
the use of cloud-based analytics tools. These tools enable the processing and analysis of data
in the cloud, eliminating the need for significant on-premises infrastructure. For example,
cloud-based ML services can analyze historical and real-time data to identify patterns and
predict future trends. These insights help in optimizing embroidery designs, improving
machine performance, and reducing production costs. The ability to perform complex
analytics in the cloud also allows for more sophisticated decision-making processes, as
stakeholders can leverage advanced algorithms and models without needing specialized
hardware or software on-site [5].
Cloud computing also supports the integration of IoT devices in machine embroidery.
IoT sensors embedded in embroidery machines collect real-time data on various
parameters, such as thread tension, stitch quality, and machine status. This data is
transmitted to the cloud, where it can be processed and analyzed. The integration of IoT
with cloud computing enables real-time monitoring and control of embroidery machines,
allowing for immediate adjustments to improve quality and efficiency. Furthermore, cloud
platforms can aggregate data from multiple machines and locations, providing a
comprehensive view of the entire production process [6].
Collaboration and communication are further enhanced by cloud computing through the
use of cloud-based project management and collaboration tools. These tools, such as Slack,
Trello, and Asana, enable teams to share updates, provide feedback, and track project
progress in real-time. This fosters a collaborative environment where all stakeholders can
contribute to decision-making processes, regardless of their physical location. The use of
these tools ensures that everyone is aligned and informed, leading to more effective and
efficient decision-making [7]. Cloud computing also facilitates the deployment of digital
twins in machine embroidery. Digital twins are virtual replicas of physical machines and
processes, allowing stakeholders to simulate and optimize production scenarios without
disrupting actual operations. By leveraging cloud computing, digital twins can be updated
in real-time with data from IoT sensors, providing an accurate and up-to-date
representation of the physical system. This enables stakeholders to experiment with
different configurations and settings, identify potential issues, and implement
improvements before making changes to the actual production environment [8].
2.3. Artificial intelligence and machine learning
AI and ML algorithms are pivotal in the modern information technologies for joint decision-
making in machine embroidery, offering advanced methods to optimize design, production,
and maintenance processes. These technologies use large datasets and sophisticated
computational models to provide insights and predictions that enhance decision-making
capabilities [9]. One key AI technique is predictive maintenance, which uses historical data
and real-time monitoring to predict equipment failures before they occur. ML algorithms
such as support vector machines, neural networks, and decision trees analyze data from IoT
sensors embedded in embroidery machines to identify patterns indicative of potential
issues. By predicting when a machine is likely to fail, maintenance can be scheduled
proactively, thus preventing unexpected breakdowns and minimizing downtime [10]. This
predictive capability is vital for maintaining continuous production and high-quality output.
Another crucial application is design optimization. ML models can analyze vast amounts of
historical design data to identify trends and preferences. Algorithms such as k-means
clustering and principal component analysis can segment designs based on various
features, helping designers understand which attributes are most popular or effective. This
analysis can guide the creation of new designs that are more likely to succeed in the market.
Additionally, generative adversarial networks are used to generate new embroidery
patterns by learning from existing designs, fostering innovation and creativity [11].
Real-time quality control is also enhanced through AI and ML. Convolutional neural
networks (CNNs), a type of deep learning model, are employed to inspect the quality of
embroidery in real-time. These models can detect defects such as misaligned stitches, color
inconsistencies, or thread breakages by analyzing images captured during the embroidery
process. When a defect is detected, the system can automatically adjust machine settings to
correct the issue or alert human operators for intervention [12]. This immediate feedback
loop ensures high-quality production and reduces waste.
AI and ML also improve supply chain management in machine embroidery. By analyzing
data from various stages of the supply chain, algorithms can optimize inventory levels,
forecast demand, and manage logistics more efficiently. Techniques such as linear
regression, time-series forecasting, and reinforcement learning are used to predict demand
trends and optimize stock levels, ensuring that materials are available when needed
without overstocking [13]. This optimization leads to cost savings and improved
operational efficiency. Moreover, collaborative AI platforms facilitate joint decision-making
by integrating insights from various data sources and stakeholders. These platforms use
natural language processing (NLP) to understand and process human language, enabling
seamless communication between machines and human operators. NLP algorithms can
analyze feedback from designers, operators, and customers to provide comprehensive
insights that inform decision-making processes [14]. This integration ensures that decisions
are based on a holistic view of the entire production ecosystem.
Thus, AI and ML algorithms significantly enhance joint decision-making in machine
embroidery by providing predictive maintenance, design optimization, real-time quality
control, supply chain management, and collaborative decision-making platforms. These
technologies leverage sophisticated computational models and large datasets to offer
insights and predictions that drive efficiency, quality, and innovation in the industry [15].
2.4. Augmented reality systems
AR systems improve joint decision-making in machine embroidery by providing immersive,
interactive environments for design, visualization, and collaboration. These systems
overlay digital information onto the physical world, enabling users to see and interact with
embroidery designs in real-time. One of the primary applications of AR in this field is design
visualization. Designers can use AR to project embroidery patterns onto fabrics, allowing
them to see how the final product will look and make adjustments on the fly. This capability
is particularly useful for customizing designs to meet specific customer requirements, as it
provides a tangible preview before production begins [16]. AR also enhances collaborative
decision-making by enabling multiple stakeholders to view and interact with the same
design simultaneously, regardless of their location. Using AR headsets or mobile devices,
team members can discuss and modify designs in real-time, ensuring that all feedback is
incorporated before the final approval. This level of collaboration is facilitated by platforms
like Microsoft HoloLens and Google Glass, which support AR applications tailored for
industrial use [17].
Besides, AR systems provide interactive tutorials and simulations that help new
designers and machine operators learn complex embroidery techniques more effectively.
For instance, AR can guide users through the steps of setting up an embroidery machine or
troubleshooting common issues, overlaying instructions directly onto the physical
equipment. This hands-on approach to training accelerates learning and improves
retention, ultimately enhancing overall productivity [18].
It should be noted that AR is also used for quality control in machine embroidery. By
overlaying design templates onto finished products, AR systems can quickly identify
deviations from the intended pattern, such as misaligned stitches or incorrect colors. This
real-time inspection capability allows for immediate corrections, reducing waste and
ensuring that the final products meet high-quality standards. Advanced AR systems can
even integrate with machine vision technology to automate this inspection process, further
enhancing efficiency [19].
In addition to these applications, AR supports marketing and customer engagement by
providing interactive experiences. For example, customers can use AR applications on their
smartphones to visualize custom embroidery designs on their clothing before purchasing.
This personalized shopping experience enhances customer satisfaction and can drive sales
by showcasing the potential of custom embroidery in a compelling way [20].
2.5. Collaborative platforms
Collaborative platforms are essential in the joint decision-making processes within machine
embroidery, particularly through their ability to facilitate real-time communication, project
management, and data sharing. These platforms integrate various functionalities that allow
designers, production managers, and clients to collaborate seamlessly, ensuring that all
parties are aligned and informed throughout the production cycle.
Slack is one of the most widely used collaborative platforms, offering real-time
messaging, file sharing, and integration with other tools. Its channel-based communication
system allows teams to organize discussions by projects or topics, ensuring that relevant
information is easily accessible. Slack's integration capabilities enable users to connect with
other software tools commonly used in machine embroidery, such as design software and
project management applications, enhancing workflow efficiency [21]. Trello and Asana are
popular project management tools that help teams track progress, assign tasks, and manage
deadlines. Trello uses a card-based system to represent tasks, which can be moved across
different stages of a project pipeline. This visual approach to task management helps teams
understand the status of ongoing projects at a glance. Asana, on the other hand, offers more
detailed task management features, including dependencies, milestones, and custom
workflows. Both platforms facilitate collaboration by allowing team members to comment
on tasks, upload files, and set reminders, ensuring that everyone stays on the same page
[22]. Microsoft Teams combines the communication features of Slack with the project
management capabilities of Trello and Asana. It offers chat, video conferencing, and
integration with Microsoft Office applications, making it a comprehensive tool for
collaboration. Teams' ability to host virtual meetings is particularly valuable for
geographically dispersed teams, enabling real-time discussions and decision-making
without the need for physical presence.
Additionally, Teams supports integration with various third-party apps, allowing users
to customize their workspace to fit their specific needs [23]. Google Workspace, formerly
known as G Suite, provides a suite of cloud-based productivity tools, including Google Docs,
Sheets, and Drive. These tools allow multiple users to work on the same document
simultaneously, making real-time collaboration straightforward. The commenting and
suggestion features in Google Docs are particularly useful for reviewing and refining design
documents, while Google Sheets can be used for project tracking and data analysis. Google
Drive offers secure cloud storage, ensuring that all project files are accessible to authorized
team members from any location [24].
Collaborative platforms also incorporate advanced data sharing and security features to
protect sensitive information. Platforms like Slack and Microsoft Teams offer end-to-end
encryption and compliance with industry standards such as GDPR and HIPAA, ensuring that
data privacy is maintained.
Furthermore, these platforms provide detailed access controls, allowing administrators
to define who can view, edit, or share specific information [25].
Comparing the described information technologies in terms of overall effectiveness and
practical application, real-time data analytics and cloud computing stand out for their broad
applicability and immediate impact on decision-making processes. AI and ML offer
substantial long-term benefits through continuous learning and optimization. AR provide
powerful visualization and simulation capabilities, crucial for design and planning stages.
Each technology has its unique properties, and their combined use can significantly enhance
the joint decision-making process in machine embroidery, driving the industry towards
greater efficiency and innovation.
3. The information technology for joint decision making in
machine embroidery with augmented reality
The use of AR systems in machine embroidery technologies offers several significant
advantages. Firstly, AR enhances design visualization by overlaying digital embroidery
patterns onto physical garments in real-time. This allows designers and stakeholders to see
how the final product will look and make immediate adjustments, ensuring the design
meets aesthetic expectations. AR also facilitates collaborative decision-making by enabling
multiple users to view and interact with the same virtual design simultaneously, regardless
of their physical location. This fosters better communication and quicker consensus,
reducing the time needed for approvals.
Moreover, AR improves training and education for new designers and machine
operators. Interactive AR tutorials can guide users through complex tasks, such as setting
up embroidery machines or troubleshooting common issues, providing hands-on learning
experiences that are more effective than traditional methods. In quality control, AR systems
can overlay design templates onto finished products to quickly identify and correct
deviations, ensuring high-quality output. This real-time inspection capability reduces waste
and enhances production efficiency.
3.1. The joint decision-making information technology for machine
embroidery
The information technology for joint decision-making in machine embroidery, integrating AR
and AI systems, involves a complex, interactive process designed to optimize embroidery design
and production. The system starts with the collection of multiple input data streams: video data
from a camera, initial embroidery data, geometry of the human body, geometry of the clothes,
general design requirements, and expert recommendations. These inputs are important for
creating a personalized and precise embroidery design.
The diagram of the proposed information technology for digital embroidery with means of AR
is presented in Fig. 1.
Figure 1: The diagram of proposed information technology for digital embroidery with AR.
The AR device receives video data and initial embroidery data to generate AR visuals. This
allows designers, experts and other participants to see how the embroidery will appear on the
garment in real time, facilitating immediate feedback and adjustments. The AI system
simultaneously processes additional inputs such as the body and clothes geometries, design
requirements, and expert recommendations to optimize the embroidery parameters. Using
advanced algorithms, the AI system analyzes these data to refine the embroidery technique,
coordinates, and colors, ensuring that the design meets both aesthetic and technical criteria.
After optimization, the AI system sends a correction signal back to the AR device, updating
the embroidery technique, coordinates, and colors. This ensures that the AR representation
remains accurate and synchronous with the real time. The output is a dynamic, real-time
visualization of the optimized embroidery on the clothes, enabling effective collaboration
among designers, production teams, and clients [26].
This integration of AR and AI facilitates a seamless and iterative design process, where
continuous feedback loops between the AI and AR systems allow for constant refinement of the
designs. The technology enhances joint decision-making by providing a visual, interactive
platform for evaluating and optimizing embroidery designs, leading to higher quality outcomes
and more efficient production processes.
3.2. The design requirements and conditions
For the efficient realization of the information technology for joint decision making in machine
embroidery with AR, the AR device and AI system must meet specific technical requirements.
The AR device requires high-performance graphics processing units (GPUs) to handle the
rendering of complex embroidery designs in real-time. These GPUs must support advanced
shading techniques and high frame rates to ensure smooth and realistic visualizations.
Additionally, the device should incorporate precise tracking systems, such as optical trackers
or inertial measurement units (IMUs), to maintain accurate alignment of virtual elements with
the physical garment, even during rapid movements. The AR device must also include high-
fidelity audio components to provide auditory feedback and enhance the immersive experience.
This involves integrating spatial audio technology, which allows sounds to be placed and moved
in 3D space, aligning with the visual elements. The device should support multiple input
modalities, including voice commands, hand gestures, and touch interfaces, to facilitate intuitive
interaction with the virtual environment and design elements. Connectivity is a critical aspect,
requiring the AR device to support low-latency wireless communication standards like Wi-Fi 6
or 5G. This ensures seamless data transfer between the device and cloud-based AI systems,
enabling real-time processing and feedback. The device must also feature robust security
protocols to protect sensitive design data and user information, employing encryption
standards and secure authentication methods.
For the AI system, the requirements include powerful computational resources capable of
handling large-scale data processing and ML tasks. This necessitates the use of multi-core
processors and high-memory bandwidth to support deep learning frameworks. The AI system
should be equipped with extensive storage capacity to manage vast datasets, including historical
design data, user preferences, and sensor readings. The AI system must incorporate advanced
neural network architectures tailored for image recognition, pattern detection, and optimization
tasks. These models should be trained on diverse datasets to ensure high accuracy and
generalization across different embroidery designs and fabric types. Additionally, the system
should employ real-time inference engines to deliver instantaneous feedback and adjustments
to the AR device. Integration with cloud services is essential for the AI system to leverage
distributed computing resources and facilitate collaborative design processes. The system
should support scalable cloud infrastructure, allowing for dynamic allocation of resources based
on computational demands. It should also include robust APIs and software development kits
to enable seamless integration with various AR devices and third-party applications. The AI
system must feature comprehensive data analytics capabilities, utilizing ML algorithms to
derive insights from user interactions and design outcomes. This involves implementing
predictive analytics to anticipate design trends and recommend optimizations. Furthermore, the
system should support continuous learning, updating its models based on new data and user
feedback to improve performance over time.
User interface design is another critical component, requiring the AI system to present data
and recommendations in a clear, actionable format. This involves developing intuitive
dashboards and visualization tools that allow users to monitor design progress, review
suggestions, and make informed decisions. The system should also provide collaborative
features, enabling multiple users to contribute to the design process simultaneously, regardless
of their physical location (Table 1).
3.3. The proposed mathematical model
The information technology for joint decision-making in machine embroidery, as shown in the
structure diagram (Fig. 1), integrates AR and AI systems to optimize embroidery designs and
processes. The input data includes the video signal V = {v R ,v G ,v B } from the camera, initial
=
embroidery image =
X { x R , x G ,… , x B } , geometry of the human body H {h1 , h2 ,… , hM } ,
=
geometry of the clothes G { g 1 , g 2 ,… , g L } , general embroidery design requirements
=R {r1 , r2 ,… , rN } , and expert recommendations
= D {d 1 ,d 2 ,… ,d K } . The output data is the
AR video image Y = { y R , y G , y B } , which displays the virtual embroidery projected onto the
real clothes.
The AR system functions by projecting the embroidery image X onto the cloth surface
depicted in the video signal V . This process is described by the model Y = F ( X ,V ) , where Y
represents the AR output. The AR system ensures that the embroidery design is correctly
aligned and proportioned on the garment as seen in the video feed.
The AI system plays a critical role by analyzing both the input data (including X , H , G , R
, D ,) and the AR output Y . Using a neural network model C =W ( X ,Y , H ,G , R , D ) , the AI
system optimizes the embroidery technique, coordinates, and colors. The correction signal
C = {c R ,cG ,c B } generated by the AI system is then transmitted back to the AR system. This
signal adjusts the embroidery parameters, ensuring that the virtual projection is refined and
improved continuously. The neural network model within the AI system processes the multi-
dimensional input data to identify optimal settings for the embroidery process. It uses learning
algorithms to evaluate various configurations and predict the best outcomes based on historical
data and expert inputs. This iterative optimization ensures that the final embroidery design not
only meets aesthetic requirements but also adheres to technical constraints, such as thread
tension and stitch density.
Table 1
Technical requirements for the information technology realization
Component Technical AR Device AI System
Requirement
Processing High- Advanced shading Multi-core processors,
Power performance techniques, high frame rates high-memory
CPUs and GPUs bandwidth
Tracking Precise tracking Optical trackers, IMUs N/A
Systems systems
Input Multiple input Voice commands, hand N/A
Modalities methods gestures, touch interfaces
Connectivity Low-latency Wi-Fi, 5G recommended N/A
wireless
Security Robust security Encryption standards, N/A
Protocols secure authentication
Neural Advanced neural N/A Tailored for image
Network network recognition, pattern
Models architectures detection, optimization
Real-time Real-time N/A Real-time inference
Inference processing and engines
feedback
Cloud Scalable cloud Low-latency wireless Distributed, scalable
Integration infrastructure communication with AI computing resources
Data Analytics Comprehensive N/A Predictive analytics,
data analytics continuous learning,
deriving insights from
user interactions
User Interface Intuitive and High-resolution display, Intuitive dashboards,
Design actionable data intuitive interaction visualization tools,
presentation interfaces collaborative features
The objective is to minimize the error E between the desired AR output Yopt and the actual
AR output Y . The error function E can be defined as the mean squared error (MSE) between
the two outputs:
1 N
=
E (Yopt ,Y ) ∑ (Yopt ,i −Yi )2 . (1)
N i =1
The neural network aims to minimize this error by finding the optimal correction signal C that
adjusts the AR projection. Thus, the optimization problem can be expressed as:
minC E (Yopt , F ( X ,V ,C )) . (2)
Subject to the constraint that the correction signal C is generated by the neural network model:
C = W ( X , Y , H , G , R, D ) . (3)
The neural network model W processes the input data and current AR output through multiple
layers, each with its own set of weights and activation functions. The goal is to iteratively adjust
these weights to minimize the error function E .
The neural network (3) is a complex multilayer model designed to determine the vector of
embroidery correction parameters C based on diverse input data: initial embroidery data X ,
AR data Y , human body geometry H , garment geometry G , design requirements R , and
expert recommendations D . This network leverages a hierarchical structure where each layer
is tailored to process different aspects of the input data, integrating them to produce optimized
outputs. The primary learning method employed here is the minimization of an error function,
augmented by attention mechanisms that direct the network's focus to the most critical
parameters of the embroidery.
The first layer of the neural network consists of input units for each data type. These units
are followed by specialized sub-networks that process the respective data types individually.
For instance, CNNs handle the image data from X and Y , extracting features such as edges,
textures, and patterns. These CNNs can be mathematically represented by a series of
convolution operations:
N
=f ( X ) σ ∑ Wi * X i + bi , (4)
i =1
where σ denotes the activation function, Wi are the convolution filters, * represents the
convolution operation, X i are the input image patches, and bi are the biases.
Parallel to the image-processing sub-networks, recurrent neural networks (RNNs) or
transformers process sequential data from R and D , capturing temporal dependencies and
semantic context. The operations in an RNN layer can be represented as:
ht σ (Wh ht −1 + Wx xt + b) ,
= (5)
where ht is the hidden state at time t , Wh and Wx are weight matrices, xt is the input at time
t , and b is the bias.
The geometric data H and G are processed through fully connected layers, transforming
these spatial coordinates into a feature space. This transformation can be represented as:
g ( H ) σ (Wh H + bh ) ,
= (6)
g (G ) σ (Wg G + bg ) ,
= (7)
where Wh and Wg are the weight matrices, and bh and bg are the biases.
The outputs from these specialized sub-networks are concatenated to form a comprehensive
feature vector. This vector serves as the input to subsequent fully connected layers that
integrate the information and perform higher-level feature extraction. The integrated feature
vector F is then passed through attention mechanisms, which enhance the network's focus on
the most relevant parts of the input data.
The attention mechanism can be mathematically described using the concept of attention
weights. Let P be the query, K the key, and V the value, which are derived from the
integrated feature vector F . The attention output A is given by:
PK T
A = softmax V , (8)
d
k
PK T
where d k is the dimensionality of the key vectors. The attention weights softmax
d
k
determine the importance of each value V based on the query P and the key K .
Finally, the output of the attention mechanism is fed into the final fully connected layers,
which compute the embroidery correction parameters C . The overall transformation can be
represented as:
C = φ ( A) , (9)
where φ denotes the non-linear transformation applied by the fully connected layers.
The learning process involves minimizing an error function E (Cp , Ctr ) , where Cp is the
predicted correction vector and Ctr is the true correction vector. The error function is typically
the mean squared error (MSE):
1 N
(Cp , Ctr )
E= ∑ (Cp,i − Ctr,i )2 .
N i =1
(10)
The network parameters are optimized using gradient descent techniques. The gradient of the
error function with respect to the network parameters u is computed as:
1 N ∂Cp,i
∇u E
= ∑
N i =1
(Cp,i − Ctr,i )
∂u
. (11)
It should be noted that the optimization (11) can be improved by using of fractional calculus
methodologies with accounting of relaxation-like processes [27,28]. But, such calculations add
a significant computational burden with the known algorithms which require the further
development of fast discrete fractional transforms.
The parameters are then updated using the Adam optimization algorithm [12-14,16], which
adjusts the learning rate adaptively:
mt
ut +=
1 ut − η . (12)
vt + s
Here, ut represents the parameters at iteration t , η is the learning rate, mt is the first moment
estimate (mean of the gradients), vt is the second moment estimate (uncentered variance of the
gradients), and s is a small constant to prevent division by zero.
The Adam algorithm initializes the first moment vector m0 and the second moment vector
v0 as zeros. It also initializes a timestep t = 0 . The hyperparameters β1 and β 2 are set to
control the exponential decay rates for the moment estimates, typically with values β1 = 0.9
and β 2 = 0.999 . At each iteration t , the gradient of the objective function E (u ) with respect
to the parameters ut is computed:
gt = ∇u E (ut ) . (13)
Next, the first moment estimate mt is updated as an exponential moving average of the
gradients:
mt β1mt −1 + (1 − β1 ) gt .
= (14)
The second moment estimate vt is updated as an exponential moving average of the squared
gradients:
vt β 2 vt −1 + (1 − β 2 ) gt2 .
= (15)
Because mt and vt are biased towards zero initially, the algorithm computes bias-corrected
first and second moment estimates:
mt
mˆ t = , (16)
1 − β1t
vt
vˆt = . (17)
1 − β 2t
ˆ t is an estimate of the mean of the gradients, while
The bias-corrected first moment estimate m
the bias-corrected second moment estimate vˆt is an estimate of the uncentered variance of the
gradients. These bias corrections are crucial for maintaining accurate estimates during the
initial stages of training.
The Adam optimization algorithm's strength lies in its ability to adapt the learning rates of
each parameter individually based on the first and second moment estimates of the gradients.
This adaptability makes Adam particularly effective for training deep neural networks with
large and noisy datasets. The algorithm combines the benefits of AdaGrad's ability to handle
sparse gradients and RMSProp's ability to handle non-stationary objectives, providing a robust
and efficient optimization method.
Incorporating attention mechanisms ensures that the network focuses on the most
significant parameters, dynamically weighting different parts of the input data based on their
relevance. This allows the network to make more accurate predictions and optimizations,
leading to improved performance in joint decision-making for machine embroidery with AR.
The integration of advanced learning methods and attention mechanisms within this multilayer
neural network facilitates the efficient and effective realization of the described information
technology.
4. Results and Discussion
In this experiment setup, we aim to demonstrate the functionality and efficiency of the
information technology for joint decision making in machine embroidery with AR. The setup
involves a smartphone with an integrated high-resolution camera acting as the Virtual Reality
(VR) device [29]. This smartphone captures real-time video data of the garment on which the
embroidery is to be performed [30]. The AI system, deployed on a personal computer, processes
this data to optimize the embroidery design. The smartphone and the personal computer are
connected via WiFi, enabling seamless data transmission between the two devices.
The AI system uses an external GPT model to analyze the recommendations from experts,
creating an array D that encapsulates the expert advice. This array, along with other input
data such as the initial embroidery design X , the video data Y , the human body geometry H
, the garment geometry G , and the design requirements R , is processed by the neural network
to produce optimized embroidery data C .
The personal computer, equipped with sufficient computational power and storage, ensures
that the AI system can handle these complex data inputs and perform real-time optimization.
The optimized embroidery data C and the AR output Y are crucial outcomes of this process.
The Y data allows for real-time visualization of the embroidery on the garment through the
VR interface on the smartphone.
This visualization helps in making immediate adjustments and ensures that the final design
aligns with the intended specifications.
The C data is used to generate the sequence of commands required by the digital
embroidery machine to execute the design.
The Janome Memory Craft 350E digital embroidery machine was used to test the proposed
information technology.
This machine is known for its precision and advanced features, making it ideal for complex
embroidery tasks. The technical characteristics of the Janome Memory Craft 350E include a
maximum embroidery speed of 650 stitches per minute, an embroidery area of 5.5 x 7.9 inches,
and a variety of built-in embroidery designs and fonts. It supports USB connectivity, allowing
it to receive embroidery designs directly from a computer.
The Janome Memory Craft 350E supports a variety of file formats, including JEF, which is
specific to Janome machines. This compatibility allows users to import a wide range of designs
from different sources. The machine's backlit LCD screen provides a clear and intuitive interface
for selecting and customizing designs, adjusting settings, and monitoring the embroidery
process.
The screen displays essential information such as stitch count, design size, and estimated
time to completion, helping users manage their projects effectively. Furthermore, the machine
includes features like an auto-declutch bobbin winder, which automatically stops winding when
the bobbin is full, ensuring consistent thread tension and reducing waste.
The Janome Memory Craft 350E also offers multiple hoop sizes, allowing users to switch
between different hoop configurations based on their project requirements. The machine's
sturdy construction and durable components contribute to its reliability and longevity, making
it a dependable choice for continuous use.
Once the optimized embroidery data C is ready, it is transformed into a sequence of
commands compatible with the Janome Memory Craft 350E. This transformation involves
converting the design data into a format that the machine can understand, such as JEF files,
which are specific to Janome embroidery machines.
The personal computer sends these commands to the Janome Memory Craft 350E via USB
or direct computer connectivity.
The embroidery machine then executes the design, stitching the optimized embroidery
pattern onto the garment with high precision.
Figure 2: The machine embroidery process (Janome Memory Craft 350E embroidery machine).
Throughout the embroidery process, the smartphone VR device continues to provide real-
time visual feedback, allowing for any necessary adjustments to be made promptly.
This integrated system ensures that the final embroidered garment meets the desired quality
and design specifications. The combination of advanced AI optimization, AR visualization, and
precise embroidery execution demonstrates the effectiveness of this information technology for
joint decision making in machine embroidery.
The integration of AR and AI in machine embroidery for joint decision-making offers several
advantages and disadvantages.
One of the main advantages is enhanced design visualization. AR enables designers to
overlay digital embroidery patterns onto physical garments in real-time, allowing stakeholders
to see how the final product will look and make immediate adjustments. This interactive
visualization ensures that the design aligns with aesthetic expectations and reduces the
likelihood of errors. Additionally, the use of AI for optimizing embroidery techniques,
coordinates, and colors ensures that the final product meets high-quality standards. The AI
system continuously analyzes input data, including the geometry of the human body and
clothes, to refine the embroidery parameters, leading to a more precise and tailored outcome.
Another significant advantage is improved collaboration. By utilizing AR, multiple users can
view and interact with the same virtual design simultaneously, regardless of their physical
location. This facilitates better communication and faster consensus among designers,
production teams, and clients. The iterative optimization process enabled by AI allows for
continuous improvement of the design based on real-time feedback, ensuring that all
stakeholders are satisfied with the final product before production begins.
Training and education also benefit from this technology. AR provides interactive tutorials
that guide new designers and machine operators through complex tasks, enhancing their
learning experience and improving retention. This hands-on approach to training can accelerate
the onboarding process and increase overall productivity.
However, there are also disadvantages to consider. The implementation of AR and AI
systems requires significant investment in advanced hardware and software. High-quality AR
devices and powerful AI computational resources can be expensive, potentially limiting
accessibility for smaller companies.
Additionally, the technology necessitates specialized skills for development and
maintenance, which may require additional training for existing staff or hiring new personnel
with the necessary expertise.
Data privacy and security are also concerns, especially when dealing with sensitive design
information and customer data. Ensuring that the AI system complies with data protection
regulations and implementing robust security measures can be challenging and costly.
Furthermore, the reliance on continuous data input and feedback loops means that any
disruption in data flow can impact the accuracy and effectiveness of the system.
Therefore, the described information technology for joint decision-making in machine
embroidery with AR and AI offers significant advantages in terms of design visualization,
collaboration, and training. However, it also presents challenges related to cost, technical
complexity, and data security. Balancing these factors is crucial for successful implementation
and maximizing the benefits of this advanced technology.
5. Conclusions
The integration of AR and AI in machine embroidery for joint decision-making presents
significant advantages and promising prospects. The primary advantages include enhanced
design visualization, where AR allows stakeholders to see and interact with embroidery designs
in real-time, ensuring alignment with aesthetic and technical requirements. This technology
also facilitates improved collaboration, enabling simultaneous input from geographically
dispersed teams, and providing interactive training modules that accelerate skill acquisition and
productivity. AI’s role in optimizing embroidery techniques, coordinates, and colors ensures
high-quality outputs tailored to individual garments.
The prospects of this technology are promising as it drives innovation, efficiency, and
quality in the embroidery industry. It allows for continuous improvement through real-time
data analysis and iterative optimization, leading to more precise and customized designs. As the
technology matures and becomes more accessible, it can revolutionize the industry by reducing
errors, speeding up the design-to-production cycle, and enhancing customer satisfaction
through personalized, high-quality products. The future of machine embroidery with AR and
AI looks bright, promising significant advancements and broader adoption across the industry.
References
[1] A. Kusiak, Smart manufacturing must embrace big data. Nature 544.7648 (2017) 23-25.
[2] P. C. Evans, M. Annunziata, Industrial internet: Pushing the boundaries of minds and
machines. General Electric Reports (2012).
[3] S. Wang, J. Wan, D. Li, C. Zhang, Implementing smart factory of Industrie 4.0: An outlook.
International Journal of Distributed Sensor Networks 2016 (2016).
[4] A. Botta, W. De Donato, V. Persico, A. Pescapé, Integration of cloud computing and internet
of things: a survey. Future Generation Computer Systems 56 (2016) 684-700.
[5] L. D. Xu, E. L. Xu, L. Li, Industry 4.0: state of the art and future trends. International Journal
of Production Research 56.8 (2018) 2941-2962.
[6] J. Lee, B. Bagheri, H. A. Kao, A Cyber-Physical Systems architecture for Industry 4.0-based
manufacturing systems. Manufacturing Letters 3 (2015) 18-23.
[7] J. Chen, J. Hu, & L. Zhang, Cloud computing for manufacturing: the case of a smart
production line. IEEE Software 33.3 (2016) 54-60.
[8] M. Grieves, & J. Vickers, Digital twin: Mitigating unpredictable, undesirable emergent
behavior in complex systems. Transdisciplinary Perspectives on Complex Systems (2017)
85-113.
[9] S. Lysenko, O. Bondaruk, Advanced methods for maintaining and managing the life cycle
of cloud environments: survey. Computer Systems and Information Technologies, (2024)
39-45.
[10] N. Mohamed, J. Al-Jaroodi, Real-time big data analytics: applications and challenges.
Proceedings of the First International Conference on Data Science, E-Learning and
Information Systems (2018) 1-6.
[11] F. Tao, et al. Digital twin-driven product design, manufacturing and service with big data.
The International Journal of Advanced Manufacturing Technology 94 (2018) 3563-3576.
[12] I. Goodfellow, J. Pouget-Abadie, M. Mirza, B. Xu, D. Warde-Farley, S. Ozair, A. Courville,
Y. Bengio, Generative adversarial nets. Advances in Neural Information Processing
Systems (2014) 2672-2680.
[13] Y. LeCun, Y. Bengio, & G. Hinton, Deep learning. Nature 521.7553 (2015) 436-444.
[14] D. Silver, et al. Mastering the game of Go with deep neural networks and tree search.
Nature 529.7587 (2016) 484-489.
[15] I. Zasornova, T. Hovorushchenko, M. Fedula, V. Buzyl, Intelligent Information Technology
for Identification of Remaining Defects in Software Products. Proceedings of the 2023 IEEE
12-th International Conference on Intelligent Data Acquisition and Advanced Computing
Systems: Technology and Applications (IDAACS-2023) (2023) 33-37.
[16] D. Jurafsky, J. H. Martin, Speech and Language Processing. 3rd ed., Prentice Hall, 2021.
[17] R. T. Azuma, A survey of augmented reality. Presence: Teleoperators & Virtual
Environments 6.4 (1997) 355-385.
[18] M. Billinghurst, A. Clark, G. Lee, A survey of augmented reality. Foundations and Trends®
in Human–Computer Interaction 8.2-3 (2014) 73-272.
[19] M. Mekni, & A. Lemieux, Augmented reality: Applications, challenges and future trends.
Applied Computational Science–Proceedings of the 13th International Conference on
Applied Computer and Applied Computational Science (ACACOS '14) (2014) 205-214.
[20] R. Palmarini, J. A. Erkoyuncu, R. Roy, H. Torabmostaedi, A systematic review of augmented
reality applications in maintenance. Robotics and Computer-Integrated Manufacturing 49
(2018) 215-228.
[21] A. Javornik, Augmented reality: Research agenda for studying the impact of its media
characteristics on consumer behaviour. Journal of Retailing and Consumer Services 30
(2016) 252-261.
[22] P. Loftus, Slack: The ultimate guide to remote work collaboration. Remote Work Report
(2020).
[23] J. Rodriguez, Project management with Trello and Asana: A comparative study. Journal of
Project Management 35.2 (2019) 88-97.
[24] J. Wilson, Microsoft Teams: A comprehensive overview of features and integrations.
Collaboration Tech Journal 12.3 (2021) 45-60.
[25] A. Cooper, Google Workspace for business: Enhancing productivity and collaboration.
Cloud Computing Insights 10.1 (2020) 55-70.
[26] P. Hryhoruk, S. Grygoruk, N. Khrushch, T. Hovorushchenko. Using non-metric
multidimensional scaling for assessment of regions' economy in the context of their
sustainable development. CEUR-WS 2713 (2020) 315-333.
[27] V. Martynyuk, M. Ortigueira, M. Fedula, O. Savenko, Fractional model of the
electrochemical capacitor relaxation phenomenon. Bulletin of the Polish Academy of
Sciences: Technical Sciences 66 (2018) 441–448.
[28] C. Shuang, Z. Changlun, M. Haibing, An Adaptive Learning Rate Deep Learning Optimizer
Using Long and Short-Term Gradients Based on G-L Fractional-order Derivative. Neural
Processing Letters, March 2024.
[29] T. Hovorushchenko, O. Voichur, O. Hovorushchenko, A. Boyarchuk, & I. Zasornova, The
Concept of Information Technology for Ensuring Accessibility to Art Objects for the
Visually Impaired Persons. CEUR-WS 3675 (2024) 208-222.
[30] T. Zohud, S. Zein, Cross-Platform Mobile App Development in Industry: A Multiple Case-
Study. International Journal of Computing, 20(1), 2021 46-54. doi: 10.47839/ijc.20.1.2091.