Evaluation of Detection Approaches for Road Anomalies Based on

Accelerometer Data

Edvinas Pocevicius

, Donatas Paulauskas

, Tomas Eglynas

Valdas Jankunas

, Sergej Jakovlev

Mindaugas Jusis

and Dovydas Lizdenis

Department of Marine Engineering, Klaipeda University, H. Manto g. 84, Klaipėda, Lithuania

Marine Research Institute, Klaipeda University, H. Manto g. 84, Klaipėda, Lithuania

Department of Informatics and Statistics, Klaipeda University, H. Manto g. 84, Klaipėda, Lithuania

Department of R&D, Ltd Inotecha, Danes Street 47, Klaipeda, Lithuania

Keywords: Accelerometers, IoT, Road Anomalies, Containers, Green Transportation Infrastructure, High-Frequency

Filtering, Data Analysis, Shock Detection, Transportation Systems, Smart Algorithms.

Abstract: Current container security systems record vibrations and shocks, but their potential for creating smart

transportation systems remains underutilized. This study analyzes data collected from a truck and discusses a

concept for generating road condition maps from accelerometer data. An experiment was conducted by

mounting an accelerometer on a container door to gather acceleration data in various transport conditions.

The study focuses on analyzing vertical (Z-axis) accelerations as a primary indicator of road anomalies. The

developed concept can be integrated into logistics platforms, enabling vehicle drivers and infrastructure

managers to respond to road defects in a timely manner.

1 INTRODUCTION

Transport infrastructure is a fundamental element of

the economy and societal well-being; however,

monitoring and maintaining its condition is a complex

and resource-intensive process (Yarmukhamedov et

al., 2020). Flaws such as potholes, sunken manhole

covers, or uneven railway crossings can cause vehicle

breakdowns, cargo damage, and an increased risk of

accidents. Therefore, it is crucial to develop

innovative, cost-effective solutions that enable real-

time monitoring and assessment of road conditions.

In this context, next-generation IoT (Internet of

Things) technologies offer significant potential,

enabling data collection and analysis using existing

infrastructure (Ye et al., 2024).

Most modern containers designed for cargo

transportation are equipped with integrated IoT

systems that collect information about the container's

location, vibrations, and impacts. These systems,

based on low-power accelerometers, are primarily

used for ensuring security; however, their potential

https://orcid.org/0000-0002-9973-5896

for secondary data utilization remains underexplored

(Barlogis et al., 2025). The simplified process

illustrated in Figure 1 demonstrates how integrated

IoT systems in containers can be adapted to monitor

secondary events, such as detecting and recording

road anomalies. These systems, utilizing existing

accelerometer data, could identify events exceeding

threshold values, such as impacts caused by potholes,

railway crossings, or other uneven surfaces.

Figure 1: Concept of secondary event.

When such an event is detected, the recording

device can analyze the nature of the event and, along

with geolocation data, transmit this information not

only to the owners of the recording devices but also

to centralized data processing systems. Third-party

systems (platforms), using algorithms and artificial

intelligence, could identify recurring events and

determine road anomalies in specific locations. The

588

Pocevicius, E., Paulauskas, D., Eglynas, T., Jankunas, V., Jakovlev, S., Jusis, M. and Lizdenis, D.

Evaluation of Detection Approaches for Road Anomalies Based on Accelerometer Data.

DOI: 10.5220/0013426300003941

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

In Proceedings of the 11th International Conference on Vehicle Technology and Intelligent Transport Systems (VEHITS 2025), pages 588-595

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

conclusions obtained, integrated into GIS

(Geographic Information Systems) platforms, could

be forwarded to the responsible authorities tasked

with road maintenance and repair. The figure below

illustrates how accelerometer data can be used for

secondary events (see

Figure 2

Figure 2: Secondary data usage model.

This process would ensure not only accurate and

timely detection of road defects but also more

efficient resource utilization for road maintenance.

Additionally, this data could be used to create real-

time road condition maps accessible not only to

institutions but also to the transport sector and the

public. Such a solution would allow logistics

companies and drivers to plan their routes in advance,

avoiding hazardous or poorly maintained road

sections. Based on edge computing methodologies,

these systems could be deployed on a larger scale, for

example, for road condition monitoring [citation

needed]. This approach would not only contribute to

improving infrastructure maintenance but also

support the development of green transportation

infrastructure by reducing the costs and

environmental impact of traditional solutions.

Road condition monitoring typically requires

specialized equipment, such as LiDAR technology or

cameras which are expensive and demand significant

computational power. In this paper, we discuss

alternatives—using accelerometers installed in

containers as an additional data source.

Accelerometers, which detect changes in acceleration

during transport, enable the identification of road

irregularities based on the vibration patterns of the

vehicle. By integrating this data with GPS

information, real-time road condition maps can be

generated, which could be beneficial for both drivers

and infrastructure managers.

2 RELATED WORKS

2.1 Road Damage

Various methodologies have been proposed in recent

studies to assess and predict road damage, each

employing diverse approaches.

For instance, Yang, X. et al. (Yang et al., 2024) in

a recent study, sought to review and analyze

advancements in automated systems for road

maintenance, focusing on detecting road distress. The

study explores the integration of computer vision,

artificial intelligence, and robotics in the context of

road infrastructure, emphasizing developments in

detecting both surface and internal road damages. For

surface distress detection, sensor technology, image

processing, machine learning, and deep learning

techniques like convolutional neural networks

(CNNs) were integrated to enhance accuracy in

identifying cracks and other damages.

Similarly, Deepa, D. (Deepa & Sivasangari,

2023) aimed to create an efficient method for

detecting and classifying road damages using a hybrid

deep learning framework. Their approach utilized a

combination of techniques: pre-processing images

with adaptive histogram equalization, segmenting

damage using fuzzy c-means clustering, and

extracting features with Laplacian edge detection and

wavelet-Walsh transforms. These optimized features

were classified using a hybrid DeepCapsule

Autoencoder-based Convolutional Neural Network

(DCACN) enhanced by Whale Optimization. The

study concluded that "The proposed work attains

98.815% accuracy, and the obtained results

outperform the existing approaches.”

Furthermore, a study by Mattes, P. (Mattes et al.,

2023) introduced Competitive Reconstruction

Networks (CRNs) for detecting road damage in

mobile mapping data, specifically focusing on

anomaly detection within 3D LiDAR-derived point

clouds. The authors set out to develop a semi-

supervised machine learning approach capable of

identifying road anomalies with minimal reliance on

manually labeled data, addressing a major challenge

in road quality evaluation. The study concludes that

CRNs can effectively enhance road quality

assessments by providing accurate, automated

anomaly detection in road imagery derived from

LiDAR point clouds.

Lastly, in a recent study, Boyarchikov Y., &

Martinec T. (Boyarchikov & Martinec, 2024) aimed

to develop a cost-effective, sensor-based solution for

real-time road pavement monitoring, seeking an

alternative to costly methods like LiDAR and labor-

Evaluation of Detection Approaches for Road Anomalies Based on Accelerometer Data

589

intensive manual inspections. Using accelerometers

and gyroscopes mounted on vehicles, the system

captures vibration data, which is then processed

through neural networks to classify road sections as

"damaged" or "undamaged." Showing the potential of

replacing LiDAR based technologies with

accelerometer based devices.

2.2 Accelerometers for Road Anomaly

Detection

Other studies also underlined the affordability and

accuracy of using accelerometer based technologies

to monitor and detect road and bridge damage.

For instance, a recent study by Gupta J. et al.

(Gupta et al., 2024) set out to explore the integration

of IoT-based sensor technology for sustainable road

safety. It examined the role of sensors such as

accelerometers, gyroscopes, and ultrasonic sensors in

enhancing vehicle safety through applications like

driver drowsiness detection, airbag deployment, and

hazard identification on the road. The accelerometer

relies on a MEMS (Micro-Electro-Mechanical

Systems) framework, utilizing piezo resistive

materials to register minute changes in resistance

under strain. The study demonstrated that these

accelerometer-based systems provide high accuracy

and rapid response times, which are critical for

reducing accidents and improving overall road

safety.

Another example of such systems was presented

by Kuladeep Chilamkuri et al. (Chilamkuri & Kone,

2020), who set out to develop a system for real-time

structural health by monitoring vibrations and

displacement of the Varadhi Bridge in India. The

study used the ADXL335 accelerometer integrated

with an Arduino UNO microcontroller to measure

vibrations and displacement on one of the bridge

spans. This setup provided a low-cost, efficient

system for real-time structural health monitoring,

enabling early detection of potential structural issues

on the bridge. Through continuous data collection, the

system identified critical vibration levels, enabling

proactive maintenance alerts when vibrations

exceeded safe thresholds. This study showed that

using such accelerometer based systems can allow

reliable condition monitoring of structures

experiencing high traffic.

Similarly, Fujino Y. et al. (Fujino et al., 2019) in

the study on Japan's infrastructure monitoring,

accelerometers played a crucial role, especially in

assessing the dynamic responses of structures like

long-span bridges and high-rise buildings.

Accelerometers measure vibrations and accelerations

resulting from natural forces such as wind and

seismic activity, providing critical data on structural

behavior under stress. The authors concluded that

Japan's approach to structural monitoring provides

valuable insights, helping maintain resilience and

operational safety in infrastructure. In a further study

Fujino Y. et al. (Fujino & Siringoringo, 2020)

outlined Japan's efforts to improve infrastructure

management through a government-sponsored

research program. In the methodology,

accelerometers were installed at various structural

points to capture real-time data. This setup allowed

researchers to measure natural frequencies, damping

ratios, and vibration patterns, helping to detect

anomalies that could indicate structural weaknesses

or damage. The data was then analyzed to confirm

theoretical models, often aligning well with design

assumptions, thus validating the structures' resilience.

The results underscored the accelerometers'

effectiveness in providing a non-invasive, continuous

monitoring solution.

Furthermore accelerometer based systems were

used by David Krier et al.(Krier et al., 2014), who

aimed to improve the accuracy of vehicle control

systems by directly estimating road-tire forces using

in-tire accelerometers. Typically, these forces are

estimated indirectly, but this study utilized

accelerometers affixed to the tire's inner tread,

capturing detailed tire-road interaction data through a

Principal Component Analysis (PCA) model. The

study concluded that the PCA model provided a

feasible and effective method for capturing the

vertical and longitudinal forces acting on the tire. The

research suggests that this innovative approach could

enhance vehicle safety and performance in real-world

applications, though further work is needed to assess

robustness across varying road conditions. A

continuation of this type of methodology could allow

for more accurate road condition estimation in further

studies.

3 METHODS

3.1 Concept of Accelerometer Usage

for Secondary Event Detection

As we know, perfect roads don't exist, at least not in

our area. Road defects could be monitored using the

vehicles traveling on them.

Smart systems installed in vehicles will generate

timestamps and values of impact events caused by

road defects and detected by sensors (see

Figure 3).

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

590

Figure 3: The concept of road monitoring system.

These data, synchronized with global positioning

system (GPS) data, should be forwarded for analysis

to the relevant authorities, who will assess road

conditions in each section, provide this information to

traffic participants, carry out preventive maintenance,

and plan road repair work. Further we focus on

detecting road defects using a sensor mounted on a

container. The study was conducted blindly, meaning

that conclusions about road defects were drawn solely

based on data obtained from the sensor without

knowing the actual physical condition of the road.

3.2 Measurement Method

When a vehicle's wheel enters a road defect zone, the

trajectory of the wheel's axis usually changes

abruptly, resulting in an impact. This generates

vibrations that propagate throughout the entire

vehicle. These impact-induced vibrations can be

recorded by a sensor installed in the vehicle. The

sensor could be installed either in the vehicle itself or

on the transported cargo. For cargo, marine containers

are particularly suitable because of their large mass—

the amplitude of vibrations caused by impacts will be

greater than with smaller masses, increasing the

likelihood of detecting road defects. Additionally, it

is currently a trend to install smart sensors on

containers, and road defect detection could serve as

an additional feature. For our experiment, it was

decided to mount the sensor on the door of a container

transported by a truck (Figure

), as smart sensors are

most commonly installed in this location.

Figure 4: Sensor mounting position.

The sensor, using an accelerometer, recorded

accelerations along three axes: X – longitudinally in

the direction of the truck's movement, Y –

transversely to the truck, and Z – vertically relative to

the road surface. The sensor processes this

accelerometer data using a special algorithm that

determines whether an impact event occurred. In this

study, accelerometer data with a 100 Hz sampling rate

was used. The data was collected while the truck

transported an empty container in the city of

Klaipėda. The total duration of the collected data (the

truck's journey) was 31 minutes. Two specific

operational periods of the truck were selected for

analysis: the first from the engine being off to the start

of acceleration, and the second during road travel

without active speed changes.

Approaching the resolution of this hypothesis,

further research was focused on detecting road

irregularities using edge computing-based algorithms.

The sensors used in container security and data

recording systems are low-power devices, which also

means that their edge computing capabilities are very

limited. Usually, these limitations are tied to device-

level functionality. For example, in the case of FSM

(Finite State Machine) algorithms, only predefined

features can be used to implement the algorithm. All of

this contributes to the complexity of solving this

hypothesis. Classical detection methods, as mentioned

earlier, cannot essentially be applied, making it

necessary to seek simpler approaches and methods. To

find these solutions, it is first necessary to understand

the data (acceleration data), its nature and

characteristics, identify its distinctive features, and find

a method to extract them using predefined features.

Considering this, a field experimental study was

conducted, using classical data collection equipment to

gather acceleration data.

Evaluation of Detection Approaches for Road Anomalies Based on Accelerometer Data

591

The equipment was mounted on the door of the

container (transported by the truck), simulating the

typical installation position of data recorders. This is

essential for detecting the characteristic properties of

the data from a specific position. Changing the

position is likely to alter the nature of the data (for

example, mounting it at the front of the truck may

affect the suspension impact, resulting in different

patterns for the same anomaly). Therefore, during the

field experiment, the classical device position was

chosen.

4 EXPERIMENTAL RESULTS

4.1 Accelerations at the Start of

Movement

In the vertical acceleration (Z-axis) graph (see Figure

5), during the period from 92.0 s to 96.2 s, the truck's

engine was turned off. In this interval, the

acceleration amplitude was very low, registering only

minor mechanical vibrations (e.g., the driver closing

the door) and the internal noise of the accelerometer.

The constant component of the vertical

acceleration is not exactly 1 g as expected because the

accelerometer is not perfectly oriented; that is, the X-

axis is not perfectly aligned with the zenith, and the

X and Y axes are not perfectly aligned with the

horizontal plane.

At the time point of 96.2 s, the driver started the

truck's internal combustion engine, and during the

period from 96.8 s to 97.2 s, the engine's idle speed

was increased several times. An increase in the

recorded acceleration amplitudes is visible in all

graphs, with the Z-axis showing the highest

amplitude at 0.051 g. While the engine continued to

idle during the period from 97.2 s to 106.7 s, lower

acceleration amplitudes were recorded, with the Z-

axis reaching a maximum of 0.008 g. It can be

observed that, while the truck engine is running, the

sensor registers the strongest vibrations caused by the

engine on the Z-axis. This is likely due to the

structural characteristics of the truck.

In the acceleration spectrum (Figure 6), the green

color represents the Z acceleration spectrum when the

engine was off (period from 92.0 s to 96.2 s), while

the blue color represents the spectrum when the

engine was idling and revved several times (period

from 96.8 s to 97.2 s). It can be observed that the

strongest influence of engine vibrations on the road

defect detection algorithm occurs in the high-

frequency range starting from 10 Hz, with amplitudes

not exceeding 0.002 g.

In the case of Z-axis accelerations, an increase in

acceleration amplitudes is also observed across the

entire frequency range. A peak can be identified at

approximately 1 Hz. When examining higher

frequencies, it can be noted that, similar to the engine

on/off case (Figure 6), the slope of the acceleration

amplitudes differs from those observed in the X and

Y axes. This can likely be attributed to the structural

characteristics of the truck and the operating

conditions of the engine.

Figure 5: Start of movement: acceleration Z.

Figure 6: Z acceleration spectrum of the engine off and on.

Figure 7: Z acceleration spectrum without acceleration and

with acceleration.

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

592

Comparing the Z acceleration patterns presented in

Figures 6 and 7 in the high-frequency range, it can be

observed that the envelope characteristics of the

patterns are not identical. Therefore, the influence of

vibrations caused by road irregularities cannot be ruled

out, as the vertical displacement of the truck's wheel

when traveling over road irregularities is expected to

be significantly greater than in the horizontal direction.

This is further confirmed by the observation that, in the

high-frequency range (from ~2 Hz), the acceleration

amplitudes recorded on the Z-axis are significantly

higher than those on the X or Y axes. Thus, it can be

concluded that, for a road defect detection algorithm, it

is sufficient to analyze only the Z-axis accelerations.

4.2 Accelerations While Driving

In the vertical acceleration graph (Figure 8), the green

curve represents the accelerations when the truck was

driving on a road without significant defects, while

the blue curve represents the accelerations when the

truck encountered three road defects. The graph

clearly shows vibrations caused by impacts at the

timestamps 1459.6 s, 1462.3 s, and 1465.2 s. It can be

observed that the bursts of vibrations caused by the

impacts occur in two stages, with the first likely

originating from the front axle of the truck.

Looking at the acceleration patterns (Figure 9),

where the green curve represents the truck driving on

a road without significant defects and the blue curve

represents the truck encountering three defects, the

oscillations in acceleration caused by the road defects

clearly stand out in the frequency range from 9 Hz to

17 Hz.

4.3 Results and Analysis of Road

Irregularities

The data is collected at a low frequency of only 100

Hz, which is typical for such systems. This also

presents a significant challenge, especially from the

perspective of classical event detection methods.

During the experiment, a dataset was collected

from a single container journey. The entire experiment

was filmed to more accurately identify road sections

where anomalies were detected. A more detailed

analysis of the recorded impacts is presented below.

A manhole cover that has sunk into the asphalt

was detected (see Figure 10). This is one of the most

common road defects that can affect cargo. In this

case, the defect is minor and has no significant

impact. However, such road defects, where the

manhole has sunk more than 5 cm, can not only

damage the vehicle's wheels but also affect the cargo.

There have been known cases where similar

defects caused pallets with cargo to topple inside a

container (which could also result from improperly

secured cargo).

Figure 8: Z acceleration graph without impacts and with

impacts.

Figure 9: Z acceleration graph without impacts and with

impacts.

Figure 10: Shock detection - manhole cover.

Another commonly encountered impact-causing

infrastructure element is a railway crossing(see Figure

11). Although this is not a critical element, such a

Evaluation of Detection Approaches for Road Anomalies Based on Accelerometer Data

593

reverse road condition assessment system could help

identify crossings that will require maintenance in the

near future. This would enable the creation of maps of

road sections in need of repair and prioritize them

based on the frequency and intensity of impacts.

Road wear, one of the more common defects, is

observed in this case as recurring road depressions

and potholes. These defects cause the container with

cargo to sway. In this instance (see Figure 12), the

damage is not critical; however, the frequency and

severity of such defects along the road segment may

indicate the need for road reconstruction.

Figure 11: Shock detection – railway crossing.

Figure 12: Shock detection – Pothole/uneven road.

Figure 13: Shock detection – Overpass joints.

Another infrastructure element is overpass joints

(see Figure 13). Damage and potholes often develop

in the joints of older overpasses. Similar to railway

crossings, these areas can be monitored, and as the

risk of accidents increases, they can be identified as

problematic locations.

5 CONCLUSIONS

Since many heavy vehicles transporting containers as

cargo travel on the roads, most modern containers are

equipped with data loggers and tracking systems for

security purposes. This equipment is based on edge

computing, focusing on safety and damage detection

issues, such as critical impacts, deviations from the

route, door openings, or lock damages. A significant

portion of modern detection systems uses low-power

edge computing solutions. One such solution is low-

power accelerometers, which can process data locally

and generate interrupts to data transmission

subsystems when predefined thresholds are exceeded.

These subsystems operate in sleep mode due to higher

energy consumption, so they do not perform complex

computations and rely on interrupts generated by

sensors. Such algorithms are usually based on FSM

(Finite State Machine) models. Limited rules and

capabilities drive the need to develop algorithms

based on logic and data analysis. The currently

recorded events, such as impacts, are no exception.

These are typically recorded when the sensor's value

exceeds predefined g-force thresholds. This area

provides significant flexibility and opportunities to

refine event-based system algorithms. Additionally,

there is potential to collect secondary information.

Imagine transforming the vast amounts of

containers on the road into a road anomaly detection

system. This would enable the creation of road

condition maps based not on a single measurement

but on collective events. Currently, road conditions

are primarily assessed using specialized equipment in

targeted evaluations. This approach produces

accurate road models and assesses their condition, but

it only addresses issues with specific roads and does

not contribute to the broader monitoring of road

conditions, which is essential for advancing green

transportation initiatives.

However, this could change if modern IoT

technologies and existing recorders, supplemented

with new algorithms, enable data collection from

these devices. This data would consist of events that

exceed predefined thresholds on the road, allowing

the identification of road issues such as potholes,

uneven surfaces, inclines, and other irregularities. By

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

594

combining this data with geolocation information, we

can create a map of such anomalies. This map would

not rely on isolated events but on recurring ones. For

example, if a certain number of recurring events are

detected at the same coordinates, it can be confirmed

that an anomaly exists. By marking and classifying

anomalies in this way, it becomes possible to create

large-scale road maps. Based on the type of anomaly,

road maintenance services can respond accordingly.

As a secondary use, this data would also benefit

truck drivers and logistics companies. The latter

could use this information to prevent transportation

disruptions. It would work similarly to Waze or

Google alerts. When integrated into appropriate GIS

solutions, this information could warn container truck

drivers about upcoming obstacles (such as road

irregularities) on their route. Unlike Waze or Google,

this information would not only rely on user input but

also on data provided by recorders. This approach

would be not only more accurate but also more

reliable. For evaluating event classification, such as

the size of a pothole or the potential risk of damaging

vehicle wheels or suspension, artificial intelligence

(AI) methods could also be employed. Here’s how

they could work:

The recorder's MCU can save accelerometer data

stored in its buffer after an event and transmit it along

with a notification to a data center. Subsequently, the

data about the recurring event and the accelerometer

readings can be sent to mathematical models

(systems/algorithms) that can determine and evaluate

the impact on the vehicle, the type of impact, and

similar aspects (in the context of road irregularities).

This contextual accelerometer information would

thus serve as material for analyzing and processing

specific events.

Integrating all of this would contribute globally to

green transportation by reducing the number of

incidents caused by road defects.

ACKNOWLEDGEMENTS

An artificial intelligence model (OpenAI GPT-4o)

was employed to assist in the editing of this

manuscript. The model was utilized to refine the

linguistic style and ensure clarity, however, the

authors take full responsibility for the scientific

content of the article.

REFERENCES

Barlogis, R., Montarnal, A., Ouedraogo, C., & Gourc, D.

(2025). Predicting container intermodal transport arrival

times: An approach based on IoT data. Internet of Things,

29, 101460. https://doi.org/10.1016/J.IOT.2024.101460

Boyarchikov, Y., & Martinec, T. (2024). Road Pavement

Monitoring and Analysis Approaches Using Gyroscope

and Accelerometer Data. IFAC-PapersOnLine, 58(9),

138–142. https://doi.org/10.1016/J.IFACOL.2024.07.385

Chilamkuri, K., & Kone, V. (2020). Monitoring of varadhi

road bridge using accelerometer sensor. Materials

Today: Proceedings, 33, 367–371. https://doi.org/10.10

16/J.MATPR.2020.04.159

Deepa, D., & Sivasangari, A. (2023). An effective detection

and classification of road damages using hybrid deep

learning framework. Multimedia Tools and Applications,

82(12), 18151–18184. https://doi.org/10.1007/s11042-

022-14001-9

Fujino, Y., & Siringoringo, D. M. (2020). Recent research

and development programs for infrastructures

maintenance, renovation and management in Japan.

Structure and Infrastructure Engineering, 16(1), 3–25.

https://doi.org/10.1080/15732479.2019.1650077

Fujino, Y., Siringoringo, D. M., Ikeda, Y., Nagayama, T., &

Mizutani, T. (2019). Research and Implementations of

Structural Monitoring for Bridges and Buildings in

Japan. Engineering, 5(6), 1093–1119. https://doi.org/

10.1016/J.ENG.2019.09.006

Gupta, J., Gupta, N., Duggal, R., & Saxena, A. (2024). An

Introduction to IoT Based Sustainable Road Safety and

Underlying Sensor Technology. Encyclopedia of

Sustainable Technologies, 711–726. https://doi.org/

10.1016/B978-0-323-90386-8.00029-2

Krier, D., Zanardo, G. S., & Del Re, L. (2014). A PCA-based

Modeling Approach for Estimation of Road-tire Forces

by In-tire Accelerometers. IFAC Proceedings Volumes,

47(3), 12029–12034. https://doi.org/10.3182/20140824-

6-ZA-1003.02293

Mattes, P., Richter, R., & Döllner, J. (2023). Detecting Road

Damages in Mobile Mapping Point Clouds using

Competitive Reconstruction Networks. AGILE:

GIScience Series, 4, 1–13. https://doi.org/10.5194/agile-

giss-4-7-2023

Yang, X., Zhang, J., Liu, W., Jing, J., Zheng, H., & Xu, W.

(2024). Automation in road distress detection, diagnosis

and treatment. Journal of Road Engineering, 4(1), 1–26.

https://doi.org/10.1016/J.JRENG.2024.01.005

Yarmukhamedov, S., Smith, A. S. J., & Thiebaud, J. C.

(2020). Competitive tendering, ownership and cost

efficiency in road maintenance services in Sweden: A

panel data analysis. Transportation Research Part A:

Policy and Practice, 136, 194–204. https://doi.org/10.10

16/J.TRA.2020.03.004

Ye, Z., Wei, Y., Yang, S., Li, P., Yang, F., Yang, B., & Wang,

L. (2024). IoT-enhanced smart road infrastructure

systems for comprehensive real-time monitoring.

Internet of Things and Cyber-Physical Systems, 4, 235–

249. https://doi.org/10.1016/J.IOTCPS.2024.01.002

Evaluation of Detection Approaches for Road Anomalies Based on Accelerometer Data

595