A Systematic Literature Network Analysis: Research Mapping of International Roughness Index

2.1 International roughness index

The level of road surface roughness is an essential indicator of road performance and a crucial pavement design criterion [30]. In MEPDG, the level of road roughness is indicated by IRI [24]. IRI, the International Roughness Index, is commonly used to measure road roughness, calculated as the accumulated vertical suspension movement divided by the distance traveled by a simulated quarter-car at a measuring profile [17, 18, 31-33]. In general, road roughness characterization indicators are categorized into two classes: profile-based index and response-based index [12]. In this section, we will explore various aspects related to IRI, including measurement methodologies, influencing factors, pavement types, and the significance of IRI in road maintenance.

a) Measurement methodologies

To measure the IRI value, different methodologies have been employed. Manual methods involve tools such as Maymester, NAASRA Roughness Meter, Bump Integrator, BPR roughmeter, Rod and Level Survey, TRRL beam, APL trailer, GMR Profilometer [34], and RMD [35]. Additionally, deflection data collected using the Falling Weight Deflectometer (FWD) or Heavy Weight Deflectometer (HWD) tools are utilized for IRI measurement [36].

b) Factors influencing IRI

The IRI value is influenced by various factors, including traffic load, environmental impact, pavement roughness, and materials used for pavement construction [37]. Globally, asphalt is commonly used for flexible pavements, while cement concrete is preferred for rigid pavements [38, 39]. Flexible pavements consist of layers such as subgrade, subbase course, base course, and surface course, which support and distribute traffic loads. In contrast, rigid pavements rely on a layer of concrete to channel the traffic load to the subgrade soil [9].

c) Studies on IRI

Numerous studies have been conducted to explore different aspects of IRI. Here are some noteworthy findings:

· Prediction Models:

  - A study in Brazil developed a mechanistic methodology to predict IRI using fatigue damage parameters calculated with FlexPAVETM software [40].

  - Several studies have investigated the relationship between IRI and Pavement Condition Index (PCI). The findings indicate that the relationship can vary based on location, functional class, and slope [41]. Exponential regression was identified as the best model for connecting IRI and PCI in rural roads in Iran [42].

  - Machine learning algorithms have been utilized to develop prediction models for IRI and PCI. Simpler algorithms, such as linear regression and decision trees, demonstrated higher accuracy in predicting IRI values [43].

  - Creative fuzzy time series analysis was employed to build an IRI prediction technique with improved accuracy compared to existing methods such as polynomial fitting, autoregressive integrated moving average, and backpropagation neural network [44].

· Practical Applications:

  - Studies have focused on practical applications of IRI thresholds for road rehabilitation. For example, a study in Taipei recommended an appropriate IRI threshold of 4.50 m/km for road rehabilitation [9].

 - Another study in southern Italy examined the relationship between carrying capacity (dynamic modulus-HWD) and IRI on airport pavements [45].

 - A survey of IRI-based specifications from several nations provided insights into national standards, variances, factors influencing specifications, and methodology discrepancies [17].

· Machine Learning and IRI:

  - Systematic literature reviews have explored the use of machine learning for predicting IRI values. Artificial Neural Networks (ANN) and regression emerged as the dominant methods used by researchers worldwide, with the LTPP databases being the primary data sources. However, various countries also utilize their own national datasets [46].

· Spatial Data Forecasting:

  - In the context of toll roads in Indonesia, studies have examined the estimation of IRI. Grey forecasting and the Spatial Statistical Dynamic (SSD) model were found to increase accuracy in forecasting spatial data, particularly for the Pondok Aren-Serpong toll road [47].

In this section, we will delve deeper into the literature to explore various studies on IRI, critically analyzing their strengths, limitations, and gaps, and examining how they collectively contribute to our understanding of IRI.

· Strengths:

  - IRI is the most widely used index for indicating road surface roughness [48].

  - State transportation departments use the mean IRI (MIRI) to monitor the roughness of their pavement network [49].

· While IRI is a widely used index for indicating road surface roughness, there are some limitations:

  - Manual methods, such as Maymester, NAASRA Roughness Meter, and Bump Integrator, may offer cost-effective solutions but can be time-consuming and subject to user bias [34]. On the other hand, using FWD or HWD tools provides objective deflection data but requires specialized equipment and may cause disruptions in traffic flow [36].

  - IRI values are sensitive to different factors, such as vehicle speed, wheelbase, and suspension [50].

  - Inaccuracy between the number of vehicles' wheels can affect the accuracy of IRI values [50].

  - The relationship between IRI and other measures of roughness and riding quality is not always clear [9].

· Gaps in the studies related to IRI, are: 

  - There is a need to address the sensitivity of IRI values, inaccuracy between the number of vehicles' wheels, and the characteristics of Android-based apps for measuring IRI [50].

  - More research is needed on the effects of specific subgrade, climate, traffic, and design factors on pavement performance and their impact on IRI [49].

  - There is a need to establish appropriate IRI thresholds for road rehabilitation in metropolitan areas [9].

2.2 Systematic literature network analysis

SLNA, which has been widely utilized by researchers from various fields and countries, has been employed in several studies. For instance, in a literature review focusing on the role of information systems in sustainable business development, SLNA with citation network analysis was used to identify research areas and development directions [51]. Another study used SLNA to analyze the development of supply chain risk management research over time, mapping influential studies and identifying future research directions [52]. Similarly, a previous SLNA study explored the evolution of the body of knowledge in the field of circular economy using citation network analysis, co-occurrence keyword network analysis, and global citation score analysis [29]. Safety climate knowledge was examined using SLNA with bibliometrics and citation network analysis, identifying influential authors and research works [53]. In the supply chain field, SLNA was employed to identify critical research areas on information sharing and associated risks, emphasizing current issues and guiding future research [54]. Additionally, a study combined SLNA with systematic literature review and citation network analysis to investigate the role of artificial intelligence in innovative city sectors and uncover the dynamics of the research field [55]. Another SLNA study analyzed the literature landscape on the concept of "Smart Factory," identifying research directions and essential areas of development [56].

SLNA is a quantitative research method that builds upon systematic literature reviews and bibliographic network analyses conducted by previous researchers [54]. Its quantitative approach enables a broader view of the knowledge domain and facilitates the identification of key pathways and critical issues in a more rigorous manner [56]. By using objective measures and algorithms, SLNA provides a comprehensive understanding of a specific field through the analysis of published literature [53]. Bibliometric analysis, as part of SLNA, extracts metadata from literature databases to evaluate and monitor research, contributing to a systematic and transparent review process that enhances the quality of the review [57]. The quantitative approach of bibliometric analysis is particularly helpful in guiding researchers to influential publications and objectively mapping the study topic before delving into detailed reading [58]. With the exponential growth of scientific publications, bibliometric analysis serves as a valuable tool for filtering and accessing fundamental previous research findings efficiently. Unfortunately, there are no search results that directly answer the question of whether there are systematic literature reviews or bibliometric analyses specifically focused on IRI.

4.1 State-of-the-art international roughness index research

tu_1.png

Figure 1. Annual scientific production

Based on the annual scientific production graph in Figure 1, during 2000-2021, there was a significant increase in the number of publications on the topic of IRI with annual growth rate: 14.23%. This result shows that research in this area remains very productive and in demand, with 73, 87 and 98 documents published in the last three years, namely in 2019, 2020, and 2021, respectively in the form of articles in journals, book chapters, conference papers, and conference reviews.

The graph of the most relevant sources in Figure 2 shows three outlets during 2000-2021 that have published the most scientific writings, including Transportation Research Record journal with 156 documents, International Journal of Pavement Engineering with 59 documents, and Journal of Transportation Engineering with 31 documents. These journals are the primary outlets for civil and structural engineering scientific publications, especially in the transportation and pavement categories with H-index, quartiles, and SJR 2020 each of 119-Q2-0.62; 46-Q1-0.92; and 75-Q2-0.57, respectively.

tu_2.png

Figure 2. Most relevant sources

Identifying the most prolific writers in a field will be beneficial in identifying individual researchers/scholars who have made a significant contribution to the growth and evolution of that research field. Knowing the predecessors who pioneered studies in a particular domain will be helpful for future researchers to decide who might be approached for research policy formulation and who can be asked for their assistance in overcoming problems experienced by the organization [60].

The chart in Figure 3 presents the top four most relevant authors and the number of published articles, including Peter Múčka (27 articles), Mohammad Hossain (15 articles), Baoshan Huang (14 articles), and Oldrich Kropáč (14 articles). The three articles with the highest number of citations were written by Baoshan Huang et al. with 129 citations on the research topics asphalt mixtures, rutting, and textures, followed by Gong et al. with 80 citations on the research topics pavement, road maintenance, and life cycle costs, and Kropáč and Múčka with 55 citations on the research topics pavement roughness and unevenness.

tu_3.png

Figure 3. Most relevant authors

These authors have contributed significantly to the field of IRI research, as evident from the number of citations their publications have received. Their work covers a range of topics such as limit values, spectral characteristics, vibration-based measurements, road roughness evaluation, forecast models, pavement performance, crumb rubber applications, predictive modeling, and statistical analysis. Their contributions have advanced the understanding and application of IRI in various aspects of road engineering and maintenance. The following are some of the publications by the top four authors as first authors and/or as corresponding authors related to the topic of IRI and pavement as can be seen in Table 1.

The graph in Figure 4 presents the corresponding author's countries and the number of documents published, with the top three countries including the USA (268 documents), followed by China (58 documents), and Canada (46 documents). These three countries also have the most research conducted in collaboration with other countries, as shown by the red section in Figure 4.

The word cloud displayed in Figure 5 shows that of the 25 keywords, the seven most extensive keywords are pavements, surface roughness, road and street, pavement performance, asphalt, regression analysis, pavement, and concrete pavement.

The mapping of the keyword network using VOSviewer was performed using the analysis type of co-occurrence, full counting method, all keywords as a unit of analysis, minimum number of occurrences of a keyword of 10. As a result, the network visualization presents 189 keywords and six clusters as shown in Figure 6.

tu_4.png

Figure 4. Corresponding author's country

tu_5.png

Figure 5. The word clouds

Table 1. Top author scientific publications and the number of citations

tu_61.png

Figure 6. Network visualization

As can be seen in Figure 6, the first clusters are in red consisting of 63 keywords with the most common keywords as follows: international roughness index, surface roughness, road and streets, pavement roughness, acceleration, durability, longitudinal profile, smartphones, and vehicle response. The red cluster research focuses on measuring road surface conditions to obtain the IRI value as part of monitoring road surface conditions using a mechanical method in the form of a vibration response received by the vehicle. A recent study revealed that using smartphones and ordinary vehicles to obtain a vibration-based measurement method for monitoring pavement conditions is inexpensive and easy to use [90]. The second clusters are in green, consisting of 55 keywords with the most common keywords as follows: concrete pavement, asphalt, pavement, pavement performance, long-term pavement performance, forecasting, asphalt, pavement, environmental impact, climate change, road construction, and performance prediction. The green cluster research focuses on the impact of surface roughness on the road environment, especially the effect on fuel and truck tires. An example of a study that has been carried out is to prove that fuel consumption increases along with pavement roughness, thus impacting vehicle operating costs. This study used a simulated fleet consisting of 29 different vehicles representing various vehicle categories, technologies, and types of fuel. As a result, vehicle fuel consumption is affected by the vehicle’s rolling resistance. The pavement roughness and vehicle speed characteristics affect the rolling resistance [91].

The third clusters are in blue, consisting of 33 keywords with the most common keywords as follows: budget control, pavement management system, road maintenance, decision-making, pavement condition, pavement deterioration, pavement management, maintenance and rehabilitation, and regression analysis. The blue cluster research focuses on pavement management, which refers to the process of performing maintenance and rehabilitation to keep road conditions as optimal as possible. A study on the national road network in Iran based on cloud decision tree theory, presenting a model for selecting the most optimal maintenance and rehabilitation strategy, is one example of a study that has been conducted. The results of the cloud decision tree obtained show that fatigue cracking and IRI are the most critical parameters to determine the appropriate maintenance and rehabilitation scenario [92]. The fourth clusters are in yellow, consisting of 17 keywords with the most common keywords as follows: pavement management, pavement maintenance, pavement preservation, cost-benefit analysis, cost, service life, and cost-effectiveness. The study focuses on the problem of road operational and maintenance costs and their relation to road service life. A study that has been carried out related to this topic area applies a life-cycle cost analysis combined with a probabilistic pavement performance model based on survival analysis to evaluate the cost-effectiveness of different overlay strategies. Implementation of optimal pavement rehabilitation is essential for highway agencies to allocate a limited budget. The results apply to wet pavements carrying more than 500,000 single-axle equivalent loads per year [93]. The fifth clusters are in purple, consisting of 16 keywords with the most common keywords, including computer simulation, modeling, ridding qualities, highway systems, correlation methods, mathematical models, statistical analysis, and statistical methods. The focus of the research is computer simulation to get statistical correlation and modelling and driving comfort. An example of a study that has been carried out is using Monte Carlo simulation and road surface models to assess the level of driving comfort in automatic vehicles. The results are compared with the classification based on IRI empirical data using the threshold method and the ISO 2631 standard driving comfort guidelines [94]. The sixth clusters are in light blue, consisting of 5 keywords with the most common keywords as follows: decision trees, learning algorithms, machine learning, mean square error, and neural networks. The light blue cluster research focuses on developing predictive models to identify the IRI value using a variety of road damage factors. Machine learning, neural networks, and regression develop predictive models. An example of a study that has been done is predicting IRI values based on the LTPP database using Extreme gradient boosting (XGBoost). The results show that XGBoost provides a better model fit for the mean absolute error and coefficient of determination than the support vector regression and random forest models [46].

Table 2. Keywords analysis of international roughness index research

Discussion:

The International Roughness Index Research Keywords Analysis Table is a combination of the analysis results of the word cloud and network visualization in the form of a table, was used to identify the state-of-the-art in IRI research to address RQ1. Based on the provided Table 2, the keywords "international roughness index" and "pavements" stand out with high frequencies, occurrences, and total link strength, suggesting their central importance in the keyword network. Other keywords like "surface roughness," "roads and streets," and "pavement" also show substantial associations and occurrences. These keyword analyses provide insights into the frequency, occurrences, and link strength of specific terms, helping researchers understand their significance and relationships within the domain of study.

4.2 The emerging and trending research topics in international roughness index research

The results of keyword mapping using VOSviewer with co-occurrence analysis and full accounting methods resulted in overlay visualization with the following keywords: pavement performance, international roughness index, road and street, pavement, pavement roughness, transportation, forecasting, machine learning, deterioration, decision tree, learning algorithms, and long-term pavement performance. These keywords are still up to date topics from 2010 to 2021 as shown in Figure 7. The results show keyword analysis from the publications of the international roughness index on the temporal dimension. The node colour represents the keyword's average publication year, while the node size represents the keyword's frequency of occurrence [95].

Nearly identical mapping results from keyword mapping using VOSviewer and bibliometric mapping using Biblioshiny are shown by the thematic map in Figure 8. There are keyword clusters for the categories of motor theme, roads and street, transportation, and pavement roughness.

tu_7.png

Figure 7. Overlay visualization

tu_8.png

Figure 8. Thematic map

The keyword clusters of international roughness index, pavement, and pavement performance are in the motor and basic theme categories. These six keywords are research topics that were the driving force from 2000 to 2021.

Discussion:

The analysis of keyword mapping using VOSviewer and Biblioshiny has revealed the emerging and trending research topics in international roughness index (IRI) research. The overlay visualization generated through co-occurrence analysis and full accounting methods identified key keywords such as pavement performance, international roughness index, road and street, pavement, pavement roughness, transportation, forecasting, machine learning, deterioration, decision tree, learning algorithms, and long-term pavement performance. These keywords have remained relevant and up-to-date throughout the period from 2010 to 2021.

The analysis of keyword usage on the temporal dimension has provided insights into the evolution of IRI research topics over time. The visualization, represented by node color indicating the average publication year and node size representing the frequency of occurrence, demonstrates the continued significance of these keywords.

The results obtained from both VOSviewer and Biblioshiny mappings have shown remarkable consistency, further confirming the emerging and trending research topics in IRI research. Thematic mapping has revealed keyword clusters related to motor themes, roads and streets, transportation, and pavement roughness. Notably, the clusters associated with international roughness index, pavement, and pavement performance have emerged as prominent areas of research, falling within the motor and basic theme categories. These six keywords have served as driving forces for research in the field of IRI from 2000 to 2021.

The identified research topics provide valuable insights for researchers and practitioners in the field of IRI. These trends highlight the importance of exploring areas such as pavement performance, machine learning, forecasting, and long-term pavement performance. By focusing on these emerging and trending research topics, future studies can contribute to the advancement of IRI research and its practical applications in transportation infrastructure management.

4.3 Potential future work in international roughness index research

The analysis results using bibliometric analysis using Biblioshiny are obtained in the trend topics graph for the last five years. At the top, the keywords include roughness index, predictive analytics, decision tree, machine learning, and roughness prediction, as can be seen in Figure 9 below. The co-occurrence analysis of the overlay visualization results in Figure 10 shows that the keywords machine learning and learning algorithms, as well as up to date topics with the yellow colour representation above 2020, are closely related to keywords forecasting and the international roughness index. Because the IRI model is related or influenced by several parameters, it will be reviewed in the following paragraph to provide a deeper view of future work.

An estimate/prediction model that can describe the condition of road infrastructure in the future is measuring or determining pavement grades as one of the essential elements at the network level and project level in the Pavement Management System [19, 46, 96]. The suitable model will provide accurate prediction results that can save construction costs effectively and efficiently during the road maintenance period [97].

tu_9.png

Figure 9. Trend topics

tu_10.png

Figure 10. Trending topics with overlay visualization

In 2022 there will be two IRI research publications; the first is a study in Laos to predict IRI models using ANN for double bituminous surface treatment and asphalt concrete pavements. This model uses Age, CESAL and YESAL parameters as input, with the result that the ANN model is more accurate than the multiple linear regression model [98]. The second study used IRI prediction models using several soft computing techniques, including gradient boosting method and random forest for flexible pavement to determine the effect of maintenance, traffic, and climate conditions. The results of the study reveal that SN is the most influential parameter based on the variable importance score [99]. These models generally relate IRI to pavement distresses, site conditions, climatic conditions, traffic levels, and structural parameters, such as layer thickness. Distress-based indices, such as rut depth and crack length, are used to determine the pavement coating’s rehabilitation time or service life. Therefore, good quality performance data combined with rational and accurate damage models are essential to accurately capture the total cost of living, energy consumption, and emissions [100]. The original IRI value is used as one of the quality assurance criteria by almost all road authorities across the globe, while the existing/current IRI indicates necessary pavement maintenance measures or rebuilding requirements [101, 102]. Based on the results of previous studies, it was found that the future direction where the current trend is the utilization of big data to measure IRI with an artificial intelligence prediction model approach, so that the measurement of IRI values becomes easier and cheaper but still accurate and reliable, of course using reliable data sources such as those sourced from the LTPP database.

Discussion:

Future work in international roughness index (IRI) research can focus on several key areas identified through bibliometric analysis and trend analysis. Recent trends, as depicted in the trend topics graph and co-occurrence analysis, highlight keywords such as roughness index, predictive analytics, decision tree, machine learning, and roughness prediction. These topics, particularly machine learning and learning algorithms, demonstrate a strong association with forecasting and the international roughness index, indicating their relevance for further investigation.

One important aspect for future work is the development of prediction models that accurately describe road infrastructure conditions, specifically pavement grades, at both the network and project levels within the Pavement Management System. These models should provide precise predictions to effectively manage construction costs during road maintenance periods. Recent publications have explored the use of artificial neural network (ANN) models for predicting IRI in different pavement types, demonstrating their superiority over traditional multiple linear regression models. Additionally, soft computing techniques like gradient boosting and random forest have been utilized to assess the effects of maintenance, traffic, and climate conditions on IRI prediction models for flexible pavements.

To ensure accurate evaluation of the total cost, energy consumption, and emissions associated with pavement performance, it is crucial to combine high-quality performance data with reliable damage models. The original IRI value serves as a quality assurance criterion for road authorities worldwide, while the existing IRI provides insights into necessary pavement maintenance measures or rebuilding requirements. As the field advances, the utilization of big data and artificial intelligence prediction models, leveraging sources like the Long-Term Pavement Performance (LTPP) database, is expected to simplify and streamline IRI measurement while maintaining accuracy and reliability.

Future research in IRI should concentrate on refining prediction models, incorporating advanced techniques such as machine learning, leveraging big data sources, and ensuring the availability of comprehensive and reliable performance data to support informed decision-making in road infrastructure management.

The authors would like to thank Palangka Raya University for funding the research and writing of this article with the decision letter of study aid/grant recipient.