Использование моделей глубокого обучения для оптимизации отслеживания грузов и повышения эффективности транспортировки в логистике

Жамият

ва

инновациялар

–

Общество

и

инновации

–

Society and innovations

Journal home page:

https://inscience.uz/index.php/socinov/index

Leveraging deep learning models for optimized cargo
tracking and transportation efficiency in Logistics

Azam NURULLAEV

1

University of Cumberlands in KY

ARTICLE INFO

ABSTRACT

Article history:

Received October 2024

Received in revised form

15 November 2024
Accepted 15 December 2024

Available online

25 January 2025

Amid the digital transformation of the logistics industry,

smart logistics algorithms have emerged as a crucial technology

to enhance efficiency and reduce costs. This paper reviews the

evolution of traditional logistics technologies and highlights the

pivotal roles played by advancements such as the Internet of
Things, big data analytics, artificial intelligence, and automation

in driving logistics innovation. It delves into the application of

intelligent logistics algorithms across areas like path

optimization, intelligent scheduling, data mining and prediction,
and smart warehousing. To address the challenge of

inconsistencies between training and testing objectives, the

paper introduces DRL4Route, a deep reinforcement learning-

based framework for path optimization, along with the
DRL4Route-GAE model. Extensive offline experiments and

online deployments validate that the model significantly

outperforms existing optimal benchmark methods on real

datasets, improving metrics like location deviation squared and
top-three location prediction accuracy. These research findings

provide essential support for advancing the intelligent

development of the logistics industry.

2181-

1415/©

2024 in Science LLC.

DOI:

https://doi.org/10.47689/2181-1415-vol5-

iss6-pp

181-192

This is an open access article under the Attribution 4.0 International
(CC BY 4.0) license (https://creativecommons.org/licenses/by/4.0/deed.ru)

Keywords:

smart logistics algorithms,

route optimization,
advanced reinforcement
learning,

data analysis, enhancement
of transportation efficiency.

1

Master’s Graduate, University of Cumberlands in KY, USA

. E-mail: nurullayevazam4@gmail.com

Жамият

ва

инновациялар

–

Общество

и

инновации

–

Society and innovations

Issue

–

5

№

6 (2024) / ISSN 2181-1415

182

Логистикада юкларни монитор қилиш ва транспорт

самарадорлигини оптималлаштириш учун чуқур
ўрганиш моделларидан фойдаланиш

АННОТАЦИЯ

Калит сўзлар:

ақлли логистика
алгоритмлари,

йўналишни
оптималлаштириш,

илғор мустаҳкамлашга
асосланган ўрганиш;

маълумотларни таҳлил
қилиш,

транспорт

самарадорлигини
ошириш.

Логистика

соҳасининг

рақамли

трансформацияси

давомида ақлли логистика алгоритмлари самарадорликни

ошириш ва харажатларни камайтириш учун муҳим

технологияга айланди. Ушбу мақолада анъанавий логистика
технологияларининг ривожланиши таҳлил қилиниб, Internet

of Things(IoT), катта маълумотлар таҳлили, сунъий интеллект

ва

автоматлаштириш

каби

ютуқларнинг

логистика

инновациясини

ривожлантиришдаги

асосий

роллари

таъкидланади. У ақлли логистика алгоритмларининг

йўналишни оптималлаштириш, ақлли режалаштириш,

маълумотларни қазиб олиш ва прогнозлаш, ҳамда ақлли

омборни бошқариш каби соҳалардаги қўлланилишига
тўхталади. Тренинг ва тест мақсадлари

ўртасидаги

номувофиқлик муаммосини ҳал қилиш учун мақолада

DRL4Route-GAE

–

йўналишни оптималлаштириш учун чуқур

мустаҳкамланишга асосланган ўқув тизими, шунингдек

DRL4Route-

GAE модели тақдим этилади. Кенг кўламли офлайн

тажрибалар ва онлайн жорий этишлар моделнинг ҳақиқий

маълумотлар тўпламида мавжуд оптимал мезон усулларидан

сезиларли даражада устун эканлигини тасдиқлайди, масалан,

жойлашувдаги четланиш квадрати ва энг яхши учта
жойлашув прогнози аниқлиги каби кўрсаткичларни

яхшилайди. Ушбу тадқиқот натижалари логистика соҳасининг

ақлли ривожланишини қўллаб

-

қувватлашда муҳим асос бўлиб

хизмат қилади.

Использование моделей глубокого обучения для
оптимизации отслеживания грузов и повышения
эффективности транспортировки в логистике

АННОТАЦИЯ

Ключевые слова:

умные алгоритмы
логистики,

оптимизация маршрутов,
продвинутое обучение

с подкреплением,

анализ данных,
повышение
эффективности
транспортировки.

В условиях цифровой трансформации логистической

отрасли интеллектуальные алгоритмы логистики стали

ключевой технологией для повышения эффективности и

снижения издержек. В данной статье рассматривается
эволюция традиционных логистических технологий, а также

подчеркивается важная роль таких достижений, как

Интернет вещей (Internet of Things), аналитика больших

данных, искусственный интеллект и автоматизация, в
стимулировании инноваций в логистике. В статье

исследуются

области

применения

интеллектуальных

логистических

алгоритмов,

включая

оптимизацию

Жамият

ва

инновациялар

–

Общество

и

инновации

–

Society and innovations

Issue

–

5

№

6 (2024) / ISSN 2181-1415

183

маршрутов, интеллектуальное планирование, добычу

данных и прогнозирование, а также умное управление
складскими

процессами.

Для

решения

проблемы

несоответствий между целями обучения и тестирования

представлена система DRL4Route, основанная на глубоком

обучении с подкреплением, для оптимизации маршрутов,

а также модель DRL4Route

-

GAE. Широкие оффлайн

-

эксперименты и онлайн

-

развертывания подтверждают, что

предложенная

модель

значительно

превосходит

существующие оптимальные эталонные методы на реальных
наборах данных, улучшая такие показатели, как квадрат

отклонения от местоположения и точность прогнозирования

местоположения в топ

-

3. Результаты исследования

обеспечивают важную поддержку для дальнейшего
интеллектуального развития логистической отрасли.

Introduction

With the digital transformation of the logistics industry, intelligent logistics

algorithms have become essential for boosting efficiency and cutting costs. By harnessing
advanced technologies like Big Data and Artificial Intelligence, logistics companies can
optimize data for smart decision-making and streamlined operations. These algorithms
are crucial in several areas:

1.

Route Planning (Alternative path for Pick up/Delivery):

They significantly

reduce transport time and costs, improving overall transport efficiency.

2.

Demand Forecasting and Inventory Management:

They optimize supply

chain management, lowering inventory costs and minimizing the risk of stock-outs.

3.

Real-Time Monitoring and Quick Response:

Intelligent algorithms enhance

the flexibility and resilience of logistics systems by enabling immediate reactions to
abnormalities, resulting in more efficient resource allocation.

4.

Logistics Demand Prediction:

Accurate forecasting improves customer

experience and satisfaction.

Overall, intelligent logistics algorithms improve efficiency, lower costs, and

enhance customer service. This paper explores strategies and methods for optimizing
logistics cargo tracking and transport efficiency using data science and deep learning
models, supporting the ongoing development of intelligent logistics systems.

Related work

2.1 Traditional Logistics Technologies (TLT):

The rapid acceleration of digital transformation has ushered in a new era of smart

logistics, driving technological innovation within the industry. By leveraging digital
technologies, companies can fundamentally transform their business models,
significantly improving the efficiency and accuracy of logistics processes. Central to these
advancements are real-time data availability and seamless collaboration across supply
chain segments. Digital platforms, for instance, enable real-time collaboration, enhancing
the flow of supply chain information and boosting overall operational efficiency.

Historically, logistics technology relied on manual records and basic computer

systems, which were prone to errors and inefficiency. In the late 20th century, the
adoption of barcode and radio frequency identification (RFID) technologies greatly

Жамият

ва

инновациялар

–

Общество

и

инновации

–

Society and innovations

Issue

–

5

№

6 (2024) / ISSN 2181-1415

184

improved data collection and processing. The introduction of Electronic Data Interchange
(EDI) further modernized logistics by enabling faster, more accurate information
transmission across supply chain segments.

Entering the 21st century, the intelligent logistics era emerged, emphasizing

automation and data-driven decision-making. Technologies like artificial intelligence (AI)
and the Internet of Things (IoT) became integral to logistics innovation. Companies like
Amazon, for example, use intelligent path planning and inventory management systems
to optimize operations, cut costs, and boost transport efficiency [1]. Emerging
technologies such as blockchain and edge computing have also created new
opportunities. Blockchain enhances data security and transparency, while edge
computing improves real-time monitoring and responsiveness.

Overall, logistics technology has evolved from manual systems to intelligent,

automated solutions. Technologies like barcodes, RFID, and EDI laid a crucial foundation,
while advancements in AI, IoT, blockchain, and edge computing continue to drive
innovation and ensure the industry's sustainable growth. These developments highlight
the vast potential and transformative possibilities for logistics in the digital age.

2.2 AI-Driven Logistics Technology Innovation

Logistics technology innovation is driven by key advancements, including the

Internet of Things (IoT), big data analytics, artificial intelligence (AI), and automation. IoT
enables real-time monitoring of logistics by connecting devices and sensors, providing
essential data for intelligent decision-making. For instance, IoT temperature monitoring
systems track perishable goods in transit, ensuring their safety. According to McKinsey,
applying IoT technology can boost logistics efficiency by 10-15% [2].

Big data analytics is also crucial, as it allows logistics companies to process vast

amounts of information for predictive analysis and real-time monitoring. By analyzing
historical data, companies can forecast market demand, optimize transport plans, and
quickly address issues in the supply chain. A DHL study shows that big data-driven
forecasting systems can improve accuracy by 20-30%, significantly lowering inventory
and transport costs.

Artificial intelligence enhances logistics through intelligent path planning and

automated decision-making. Machine learning optimizes transport routes, boosting
efficiency, while deep learning enables more effective automated operations. For
example, machine learning-based cargo loading systems increase vehicle loading
efficiency, reducing transport costs. Studies indicate that AI-driven route planning can
enhance transport efficiency by 25%.

Automation technology reduces dependency on manual labor and boosts

operational efficiency. Automated systems enable rapid storage and retrieval in
warehouses, while autonomous vehicles improve transport safety and efficiency.
Amazon's automated warehouse robots, for example, have cut costs and increased
picking efficiency, with The Economist reporting a 30-40% improvement in operational
efficiency through automation [3].

The integration of these technologies has revolutionized logistics, enhancing

efficiency and accelerating digital transformation. By leveraging IoT, big data, AI, and
automation, logistics companies can achieve comprehensive, intelligent, and automated
operations, staying ahead in a competitive market. As these technologies continue to
evolve, the logistics industry is set for a more efficient, secure, and intelligent future.

Жамият

ва

инновациялар

–

Общество

и

инновации

–

Society and innovations

Issue

–

5

№

6 (2024) / ISSN 2181-1415

185

2.3 Alternative Logistics Algorithms

Intelligent logistics algorithms rely on sensing technology for real-time data

collection and advanced data processing to transform raw data into actionable insights.

Sensing technology

enables continuous monitoring, tracking the location, temperature,

and status of goods. For example, temperature sensors are crucial in cold chain logistics
to maintain product quality, and GPS systems optimize route planning by tracking vehicle
locations.

Data processing

uses big data analysis to extract meaningful patterns, support

decision-making, and predict future logistics needs. Real-time decision-making ensures
quick responses to changing conditions, while predictive analytics combines historical
and current data to anticipate demand and plan efficiently. Together, sensing technology
and data processing form the backbone of intelligent logistics algorithms, driving
optimized and efficient logistics operations.

Key Intelligent Logistics Algorithms

1. Path Optimization Algorithms

These algorithms, grounded in graph theory, find the shortest or most efficient

paths within a network. Popular methods include Dijkstra’s algorithm, which employs a

greedy approach for weighted graphs, and the A* algorithm, which enhances search
efficiency using heuristics. Genetic algorithms introduce biological evolution concepts to
solve complex network problems, making them ideal for large-scale urban planning [4].

2. Intelligent Scheduling Algorithms

These optimize resource allocation and task scheduling using genetic algorithms,

simulated annealing, and ant colony algorithms. Genetic algorithms improve vehicle
routing and cargo loading by simulating natural selection processes. Simulated annealing
avoids local optima by accepting suboptimal solutions with a certain probability, aiding
in global optimization. Ant colony algorithms simulate collective behavior, optimizing
routes and enabling vehicle cooperation.

3. Data Mining and Predictive Algorithms

Data mining techniques uncover relationships within logistics data to enhance

operations. For instance, association rule mining identifies item correlations for
improved inventory layout and combined shipments. Time series analysis forecasts
demand and transit times, revealing trends to refine planning. Machine learning predicts
future demand based on historical sales, optimizes routes considering traffic and weather
and identifies risks within the supply chain.

4. Intelligent Warehousing Algorithms

These improve warehouse efficiency using automation and intelligent systems.

Cargo distribution algorithms streamline goods movement and picking, while machine
learning-driven picking systems optimize sequences to reduce time and errors. Dynamic
scheduling allocates tasks to automated equipment in real-time, maximizing efficiency.
Automated storage systems use robots to manage inventory, and real-time monitoring
ensures accuracy and rapid response to issues.

5. Methodology

Recent advances in route prediction leverage learning-based approaches, such as

deep neural networks, to model courier patterns from historical data. Techniques like
oSquare and DeepRoute transform delivery predictions into next-location forecasting,
employing recurrent neural networks, Transformer-based encoders, and dynamic graph

Жамият

ва

инновациялар

–

Общество

и

инновации

–

Society and innovations

Issue

–

5

№

6 (2024) / ISSN 2181-1415

186

networks. However, existing methods face performance challenges due to discrepancies
between training and testing objectives, often using mismatched metrics like cross-
entropy loss and position squared deviation. Addressing this inconsistency is crucial for
enhancing model performance in real-world applications.

How methodology works?

An intuitive solution to the problem would be to directly transform the test targets

into loss functions for updating the model parameters. However, this approach is not
feasible because the test targets in this task, such as position squared deviation and
Kendall rank correlation coefficient, are non-differentiable. These test targets measure
the similarity between the real and predicted routes, with the predicted routes being
generated step by step based on the maximum values of the probability distributions
from the model outputs. Due to the non-differentiability of these test targets, it becomes
challenging to use them directly for model training.

To address this issue, a promising solution lies in utilizing

reinforcement

learning

. Reinforcement learning has proven effective in optimizing non-differentiable

objectives across various tasks such as machine translation, text summarization, and
image captioning. In these domains, reinforcement learning has shown superior
performance over traditional supervised deep learning methods, making it a viable
approach for improving model performance on non-differentiable test metrics [6].

Figure 1. An illustration highlighting the discrepancy between training and testing

objectives. The vector at each location represents the transition probabilities

associated with A, B, C, and D.

I also want to highlight a novel approach to modeling the route prediction problem

in dispatching through a reinforcement learning perspective. It introduces the
DRL4Route framework, which employs a policy-based reinforcement learning
methodology. By leveraging rewards derived from non-differentiable test objectives, the
framework optimizes a deep neural network using a policy gradient approach to address

Жамият

ва

инновациялар

–

Общество

и

инновации

–

Society and innovations

Issue

–

5

№

6 (2024) / ISSN 2181-1415

187

the inconsistency between training and testing objectives. Within this framework, the
DRL4Route-GAE model is developed specifically for route prediction in logistics
collection services.

The model employs a strategy gradient algorithm to reduce variance in gradient

estimation by utilizing an actor-critic framework. While this framework introduces some

bias due to the critic and actor models’ limitations in accurately estimating the value

function, a generalized advantage estimation method is used to strike a balance between
bias and variance during gradient estimation. The dominance value approximation is
applied when updating the loss function to refine the process.

Key contributions of this paper include:
1.

Reinforcement Learning Framework for Route Prediction

: This is the first

study to approach collection and delivery route prediction from a reinforcement learning

perspective, proposing the DRL4Route framework. It combines reinforcement learning’s

capability to optimize non-differentiable objec

tive functions with deep neural networks’

ability to learn historical behavior patterns, offering an improvement over traditional
supervised learning methods.

2.

Development of DRL4Route-GAE

: The DRL4Route-GAE model applies an

actor-critic framework to compute rewards at each decoding step based on test
objectives. It introduces generalized advantage estimation to balance bias and variance
during gradient estimation, enhancing the training process.

3.

Validation and Performance

: Extensive offline experiments using real datasets

and online deployments confirm the effectiveness of the proposed method. Compared to
the optimal baseline, DRL4Route-GAE demonstrates improvements in location bias
squared metrics by 0.9%-2.7% and top-three location prediction accuracy metrics by
2.4%-3.2%.

3.2. DRL4Route

The DRL4Route (Deep Reinforcement Learning for Route Optimization) model is a

cutting-edge algorithm designed to tackle complex logistics route planning challenges
through deep reinforcement learning. By leveraging reinforcement learning strategies, the
model autonomously learns and optimizes transportation routes, effectively reducing both
transit time and costs. Its advantages in logistics forecasting include the following [7]:

1.

Dynamic Adaptability

: The DRL4Route model adjusts in real-time to

fluctuating logistics demands and traffic conditions, delivering optimal solutions
dynamically.

2.

Efficiency in Handling Large-Scale Data

: It processes extensive datasets

efficiently and identifies optimal paths within complex networks swiftly.

3.

Continuous Improvement

: By consistently learning and refining its approach,

the model enhances the accuracy and efficiency of route planning over time, significantly
boosting the overall performance and responsiveness of logistics systems.

The framework optimizes parameters based on rewards derived from test

objectives using a policy gradient method. This enables it to address non-differentiable
objective functions for more precise route predictions. Actions that align with improving
test metrics receive higher reward values, incentivizing the model to favor such
decisions. Conversely, actions that negatively impact test metrics result in lower rewards,
prompting parameter updates to avoid such choices. This iterative learning process
ensures increasingly accurate and efficient route optimization.

Жамият

ва

инновациялар

–

Общество

и

инновации

–

Society and innovations

Issue

–

5

№

6 (2024) / ISSN 2181-1415

188

Figure 2. DRL4Route Framework

In this paper, based on the DRL4Route framework, a model named DRL4Route-

GAE is proposed for parcel collection services in logistics scenarios to illustrate the
effectiveness of the proposed framework. DRL4Route-GAE generates the spatio-temporal
representation of the unfinished task using a transformer-based encoder and models the
decision-making process of the courier by using an attention mechanism and recurrent
neural network-

based decoder to model the courier’s decision

-making process [8].

The model training is guided by the strategy gradient so we can optimize the non-

differentiable test objective to solve the problem of inconsistency between the training
and test objectives, furthermore, we use a generalized dominance estimation method to
compute an approximation of the dominance to balance the bias and variance during
gradient estimation, so we can get better strategies as well as better results.

3.3. Experimental Data Sets and Methods

We conduct offline experiments on the logistics parcel-acquisition dataset

provided by Cainiao, and the sample ratio of the training set, validation set, and test set is
about 6:2:2, and the data statistical information is shown in Table 1.

Table 1

Type

Time Range

City

ANUT

#Workers #Samples

Logistics-I1Z

07/10/2021-

10/10/2021

Ilangzhou

1,117

373,072

Logistics-SI1

03/29/2021-

05/27/2021

Shanghai

2,344

208,202

Baseline Methods

This study incorporates several baseline methods alongside state-of-the-art deep

learning models across different scenarios, such as food distribution and last-mile
logistics, to facilitate comparisons [9]:

1.

Time Greedy

: Routes are generated by prioritizing locations based on the

remaining time until task timeout, ensuring tasks with imminent deadlines are addressed
first.

Жамият

ва

инновациялар

–

Общество

и

инновации

–

Society and innovations

Issue

–

5

№

6 (2024) / ISSN 2181-1415

189

2.

Distance Greedy

: At each step, the courier selects the nearest location to visit,

constructing the route incrementally by continuously choosing the closest subsequent
destination.

Baseline Methods for Comparison

1.

OR-Tools:

A heuristic algorithm designed to find the shortest path, optimized

for minimizing travel distance.

2.

OSquare:

An XGBoost-based approach that sequentially generates routes one

node at a time.

3.

FDNET:

A model tailored for takeaway scenarios that combines LSTM and

attention mechanisms to predict paths and times.

4.

DeepRoute:

A deep learning-based route prediction model employing a

Transformer encoder and attention decoder.

5.

DeepRoute+:

An enhanced version of DeepRoute that incorporates a courier

decision preference modeling module for more personalized route predictions.

6.

Graph2Route:

A novel approach using graph structures to represent locations.

It employs a GCN-based encoder with an attention mechanism to better capture spatial
and temporal relationships and improve path predictions.

Experimental Results Summary

The comparative analysis, summarized in Table 2, reveals key insights from testing

on two datasets:

▪

Limitations of Greedy Methods

: Distance-based and time-based greedy

approaches (as well as OR-Tools) focus solely on optimizing a single aspect, such as
distance or time. Consequently, they fail to account for complex spatiotemporal
constraints, offering suboptimal solutions in real-world logistics scenarios.

▪

OSquare’s Weaknesses

: The tree-based OSquare model struggles to model

spatiotemporal correlations effectively. Furthermore, its objective is limited to
maximizing the probability of predicting the next position, rather than optimizing the
entire route.

▪

Challenges with Sequence-Based Models

: Models like FDNET and DeepRoute

face difficulties in capturing neighborhood relationships among locations, often resulting
in unreasonable route outputs.

▪

Advantages of Graph2Route

: By incorporating a graph-based encoder,

Graph2Route excels at modeling decision context information and spatiotemporal
relationships, addressing key limitations of sequence-based models and producing more
coherent and realistic routes [10].

Table 2.

Method

Logistics-HZ n

E(0.25)

Logistics-HZ n

(0.11)

Logistics-SH

n E(0.25)

LogisticsSH n

(0.11)

HR@I

ACC@3

KRC

LMD

Time-Greedy

33.15

20.32

41.92

1.7

Distance-Greedy

33.13

51.82

136

5.73

OR-Tools

53.93

1.23

4.68

1.46

OSquare

54

33.1

58.5

1.16

FDNET

52.76

33.22

55.47

1.18

DeepRoute

54.76

34.64

58.61

1.1

Жамият

ва

инновациялар

–

Общество

и

инновации

–

Society and innovations

Issue

–

5

№

6 (2024) / ISSN 2181-1415

190

DeepRoute+

55.42

35.63

59.32

1.08

Graph2Route

56.45

36.12

60.63

1.05

DRLAROute-

REINFORCE

55.88

29.02

59.97

1.38

DRL4Route-AC

56.36

36.16

60.86

1.05

DRL4Route-GAE

57.72

37.23

61.47

1.03

Improvement

2.20%

3.10%

1.40%

1.90%

Building on DeepRoute,

DRL4Route-REINFORCE

delivers superior performance,

particularly in reducing the squared deviation of position. This improvement stems from
its ability to directly optimize evaluation metrics, effectively resolving the inconsistency
between training and testing objectives.

DRL4Route-AC

surpasses DRL4Route-REINFORCE by incorporating dominance

values computed from rewards at each time step to update model parameters. This
approach mitigates the issue of error accumulation, leading to more robust results.

Finally,

DRL4Route-GAE

outperforms DRL4Route-AC by employing a generalized

dominance estimation method. This technique balances bias and variance during
gradient estimation, enabling even more precise and efficient route optimization.

Figure 3: Performance comparison curves for the algorithms DRL4Route-GAE and

DRL4Route-AC, highlighting their distinct trends and effectiveness in addressing

the route optimization problem.

In

Figure 3

, the cumulative expected rewards for DRL4Route-AC and DRL4Route-

GAE are compared throughout the training process. Both methods show increasing
reward values with the progression of training rounds, demonstrating the effectiveness
of the proposed framework. Notably, DRL4Route-GAE outperforms DRL4Route-AC,
achieving higher reward values. This improvement highlights the efficacy of the
generalized dominance estimation method in balancing bias and variance during
gradient estimation, ultimately resulting in a superior strategy.

Жамият

ва

инновациялар

–

Общество

и

инновации

–

Society and innovations

Issue

–

5

№

6 (2024) / ISSN 2181-1415

191

Addressing Mismatched Objectives

Deep models applied to range and delivery path prediction traditionally face

challenges due to the non-differentiability of test objectives. Under a supervised training

paradigm, these models cannot incorporate test criteria into the training process, causing a

mismatch between training and test objectives and limiting their real-world performance.

To overcome these limitations, this paper introduces

DRL4Route

, a novel

framework that integrates the behavioral pattern learning capabilities of deep neural

networks with reinforcement learning’s strength in optimizing non

-micro objectives.

This plug-and-play framework enhances the performance of existing deep models and

introduces

DRL4Route-GAE

, an actor-critic-based model utilizing a generalized

dominance estimation method to balance bias and variance in strategy gradient

estimation [11].

Experimental Validation

Through extensive offline experiments and online deployments on real datasets,

DRL4Route demonstrates significant improvements over competitive baseline models.

The DRL4Route-GAE model achieves superior results, particularly in positional bias

squared metrics and top-three positional prediction accuracy, solidifying its effectiveness

in logistics route prediction tasks.

CONCLUSION

This research underscores the transformative role of intelligent logistics

algorithms in modern systems. By examining the evolution of logistics technologies and

the integration of IoT, big data analytics, AI, and automation, the study highlights how

innovations like DRL4Route drive advancements in logistics path optimization,

intelligent scheduling, and prediction accuracy. The results offer valuable technical

support and direction for the intelligent development of the logistics industry, improving

the efficiency and accuracy of logistics transportation significantly.

Major Trends in Future Logistics Technology Innovation

❖

Advancement of Intelligent Logistics Systems

➢

Integration of Emerging Technologies

: Automation, artificial intelligence (AI),

and big data analytics will become more deeply embedded in logistics processes.

➢

Enhanced Decision-Making

: Deep learning algorithms will improve data

processing and decision-making capabilities, enabling smarter operations.

➢

Real-Time Response

: The adoption of edge computing will facilitate faster data

processing and quicker responses to real-time events.

➢

Human-Machine Collaboration

: Technologies that promote cooperation

between humans and intelligent systems will boost logistics efficiency.

❖

Widespread Application of Emerging Technologies

➢

5G Technology

: High-speed data transmission will improve real-time

monitoring and enable seamless remote operations.

➢

Internet of Things (IoT)

: The expansion of IoT will connect a growing number

of devices and sensors, creating more interconnected logistics networks.

➢

Edge Computing

: By processing data at the edge of the logistics network, edge

computing will reduce latency and improve operational efficiency.

❖

Emergence of Green Logistics Technologies

➢

Sustainable Transport

: The use of electric and self-driving vehicles will

contribute to reduced carbon emissions.

Жамият

ва

инновациялар

–

Общество

и

инновации

–

Society and innovations

Issue

–

5

№

6 (2024) / ISSN 2181-1415

192

➢

Circular Economy

: Initiatives like the reuse of packaging materials will

minimize waste.

➢

Smart Energy Management

: Optimizing energy usage will help reduce

resource consumption and environmental impact.

Future Vision

In the coming years, intelligent green port management systems and circular

economy logistics networks will become integral to the industry. These innovations will
foster smarter, more efficient, and environmentally sustainable logistics systems, aligning
the industry with global sustainability goals.

REFERENCES:

1.

Liang, P., Song, B., Zhan, X., Chen, Z., & Yuan, J. (2024). Automating the training

and deployment of models in MLOps by integrating systems with machine learning.
Applied and Computational Engineering, 67, 1- 7

2.

Li, A., Yang, T., Zhan, X., Shi, Y., & Li, H. (2024). Utilizing Data Science and AI for

Customer Churn Prediction in Marketing. Journal of Theory and Practice of Engineering
Science, 4(05), 72-79.

3.

Xu, J., Wu, B., Huang, J., Gong, Y., Zhang, Y., & Liu, B. (2024). Practical applications

of advanced cloud services and generative AI systems in medical image analysis. Applied
and Computational Engineering, 64, 82-87.

4.

Zhang, Y., Liu, B., Gong, Y., Huang, J., Xu, J., & Wan, W. (2024). Application of

machine learning optimization in cloud computing resource scheduling and management.
Applied and Computational Engineering, 64, 9-14.

5.

Huang, J., Zhang, Y., Xu, J., Wu, B., Liu, B., & Gong, Y. Implementation of Seamless

Assistance with Google Assistant Leveraging Cloud Computing.

6.

Lin, Y., Li, A., Li, H., Shi, Y., & Zhan, X. (2024). GPU-Optimized Image Processing

and Generation Based on Deep Learning and Computer Vision. Journal of Artificial
Intelligence General science (JAIGS) ISSN: 3006-4023, 5(1), 39-49.

7.

Chen, Zhou, et al. “Application of Cloud

-Driven Intelligent Medical Imaging

Analysis in Disease Detection.” Journal of Theory and Practice of Engineering Science

4.05 (2024): 64-71.

8.

Wang, B., Lei, H., Shui, Z., Chen, Z., & Yang, P. (2024). Current State of

Autonomous Driving Applications Based on Distributed Perception and Decision-Making.

9.

Li, Zihan, et al. “Robot Navigation and Map Construction Based on SLAM

Technology.” (2024).

10.

Fan, C., Ding, W., Qian, K., Tan, H., & Li, Z. (2024). Cueing Flight Object

Trajectory and Safety Prediction Based on SLAM Technology. Journal of Theory and
Practice of Engineering Science, 4(05), 1-8.

11.

Sarkis, R. A., Goksen, Y., Mu, Y., Rosner, B., & Lee, J. W. (2018). Cognitive and

fatigue side effects of antiepileptic drugs: an analysis of phase III add-on trials. Journal of
neurology, 265(9), 2137-2142.