Pavement asset management strategies play an important role in the maintenance of roadway systems. A classic pavement asset management system involves systematic processes and tools for optimizing pavement conditions by scheduling maintenance and repairs effectively within budgetary constraints [1] . In 2023, $68.9 billion federal funding was allocated to modernize, repair and improve the safety and efficiency of roads and bridges in the United States [2] . With more than 3.2 million miles of highways in the United States [3] , there is a need to have a data-driven and proactive approach to prioritize asset management strategies on roadway segments that need repair and improvement.

In the state of Indiana, the pavement asset management strategy is unique to pavement type and road category. Several factors impact the decision-making process including pavement roughness, rutting, surface distress, number of preventive measures, age of the pavement and age of the roadway surface. Deterioration curves are developed based on statistical quartiles and a pavement quality index (PQI) is computed for various components that normalize pavement condition across several road categories. An overall pavement condition is generated using a composite rating of component indices weighted most to roughness and least to rutting/faulting as shown in the below equation.

$\begin{array}{l}PQ{I}_{overall}=\left(0.5\ast PQ{I}_{IRI}\right)+\left(0.3\ast PQ{I}_{\%cracking}\right)+\left(0.2\ast PQ{I}_{RUT}\right)\text{for HMA}\\ PQ{I}_{overall}=\left(0.5\ast PQ{I}_{IRI}\right)+\left(0.3\ast PQ{I}_{\%cracking}\right)+\left(0.2\ast PQ{I}_{Faulting}\right)\text{for Concrete}\end{array}$ (1)

For both hot mix asphalt (HMA) and concrete, pavement roughness contributes to half of the overall pavement quality and is based upon the International Roughness Index (IRI) [4] . IRI is usually estimated using high quality inertial profilers or dedicated pavement data collection vehicles equipped with high-speed digital cameras, GPS, laser systems and accelerometers. Agencies typically inventory IRI once a year since these data collection efforts are expensive and labor intensive [5] [6] . However, pavement conditions change continually throughout the year, particularly during spring freeze thaw cycles or locations experiencing significant localized loading due to construction activities and/or detours. Consequently, there is a need to continuously monitor pavement quality on a system wide basis to implement efficient asset management strategies.

This research proposes a framework to analyze and monitor the pavement roughness using estimated IRI from crowdsourced connected vehicle (CV) data. The ubiquitous nature of this data allows continuous monitoring of pavement conditions both on a systemwide level as well as on selected roads and segments. This methodology also provides a data-driven and systematic approach to target timely mitigation strategies to manage the performance of the pavement and ensure their long-term functionality.

Traditional methods for characterizing pavement conditions involves the use of a high-speed inertial profiler [7] , a vehicle capable of collecting roughness data in terms of IRI. The measurement equipment is mounted on the front or rear of the vehicle and relevant data is collected at the posted speed limit. Several automated profilers are widely available and capable of collecting roughness including the Automated Road Analyzer (ARAN), Pathrunner, Multi-Functional Vehicle (MFV), Road Measurement Data Acquisition System (ROMDAS), and Laser Road Surface Test (Laser RST). Although this data is highly accurate, there are significant operating costs associated with these profilers and they require dedicated personnel to drive all the roadways in the system. Additionally, they are less reliable during low speeds and data extraction is labor intensive and time consuming [5] [6] [8] .

Other studies have used image processing techniques [9] - [11] , LiDAR [12] - [14] and Unmanned Aerial Vehicles (UAV) [15] - [17] to estimate pavement roughness. Maeda et al. used deep neural networks on images collected from smartphones to detect and classify pavement condition [18] . Results showed more than 70% accuracy in detecting and classifying road conditions with an inference time of roughly 1.5 s on a smartphone. In another image processing study, Obunguta et al. built a robust probabilistic pavement management model with a safety metric output based on the pavement condition [19] .

Recently, researchers have also developed smartphone applications to calculate road roughness [20] - [22] . Results show that the accuracy of GPS and accelerometer sensors in the smartphones could affect the roughness calculations [23] . Another study developed a fully convolutional neural network architecture called IRI-Net that estimates IRI from vibration data collected using smartphones [24] . The study also outlines strategies for handling low resolution GPS data obtained from smartphones. Results showed that the developed model was able to accurately estimate IRI under real-world conditions irrespective of the vehicle type, driving speed, and smartphone used for data collection.

An early study conducted in 2014 explored the use of sensor data from crowdsourced CVs for effective pavement data collection and found that new metrics from this source could be implemented for transportation asset management practices within the near term of 3 - 5 years [25] . Bridgelall et al. used data from on-board accelerometers and speed sensors to estimate road roughness indices and found that the average margin of error between the measurements was less than 6% [26] . Researchers have also investigated the possibility of augmenting standard pavement survey data with spatiotemporally continuous sensor data crowdsourced from CVs. Kargah-Ostadi et al. leveraged a physics-integrated machine learning model and quarter-car simulation approach to predict standard IRI values using speed, acceleration and suspension data. The evaluation results yielded high accuracy, precision and generalization capability between the training and test datasets, especially on smoother road profiles [27] .

A recent study exploring the potential uses of CV technologies found that utilizing sensor data from CVs for pavement evaluation offers a cost-effective and efficient approach [28] . Few studies have used enhanced real-world CV data to evaluate pavement quality. A study compared IRI estimates from connected production vehicle data and inertial profilers, and found that there was a good correlation between the two with an R² value of 0.79 [29] . Another study highlighted the scalability of this data by analyzing pavement roughness over 5800 miles of I-80 using CV data from 730,000 crowdsourced segments [30] . Other studies have applied machine learning techniques on data from inertial sensors to estimate roughness index [31] [32] . In general, the results suggest that estimated roughness from crowdsourced CV data is a viable tool for network level screening of pavement quality.

The objectives of this study are the following:

Figure 2 provides an annual comparison of the median roughness across all interstates (by direction) in Indiana during the study period. Table 1 shows the total bi-directional miles of interstate roadways classified by pavement roughness condition between 2022 and 2024. The classification is performed using the IRI thresholds outlined by Federal Highway Administration (FHWA) [36] . In general,

Changes in specific interstates are shown in Figure 2 .

This systemwide overview of pavement roughness at both a system level and route level provides an important high-level status of pavement condition by year. The subsequent sections describe how this data can be further analyzed at more discrete spatial and temporal levels.

Table 1. Pavement roughness by miles of interstates in Indiana and absolute change compared to 2022.

*6 months.

This study used crowdsourced connected vehicle data to develop spatiotemporal assessment of pavement roughness conditions for asset management planning. Over 3 billion connected vehicle records available between January 2022 and June 2024 ( Figure 1 ) in the state of Indiana were linearly mapped to interstate roadways. Annual comparisons across all major interstates in Indiana showed the miles of interstates classified as “Good” decreased from 1896 to 1661 miles between 2022 and 2024. The miles of interstate classified as “Needs Maintenance” increased from 82 miles to 120 during that same period ( Figure 2 and Table 1 ). Annual ( Figure 3 ), monthly ( Figure 6 ) and daily ( Figure 7 ) comparisons showing the variations in pavement conditions on I-65 were presented. Visualizations were also developed to easily identify segments along the interstate with considerable variations in pavement conditions over time ( Figure 5 ). Dashcam images from commercial trucks ( Figure 8 ) and Google Street View ( Figure 4 ) were also used to validate the results generated using the crowdsourced data.

The methodology and framework presented in this study will help agencies understand the time and location when pavement conditions change so they can undertake a proactive approach to asset management planning. The real-time availability of this data will also provide additional opportunities for continuous pavement monitoring. Furthermore, the results from this methodology can be used for quantitative monitoring of pavement conditions and can be used to identify locations for additional targeted and enhanced data collection using advanced equipment. The visualizations and performance measures developed in this study could also aid in the systematic prioritization of pavement management strategies to ensure safety and long-term functionality of roadways.

Future research may be used to perform comprehensive assessments of pavement degradation rates. This will be particularly helpful for agencies to identify critical sections, optimize budget allocations, and prioritize maintenance tasks. By comparing daily, monthly, and yearly degradation rates, agencies can develop effective mitigation strategies at the system, network, or project level. Timely pavement maintenance can also significantly enhance user safety and reduce costs.

The connected vehicle roughness data was provided by NIRA Dynamics AB. Commercial truck dashcam images were provided by Vizzion. This work was supported by the Joint Transportation Research Program and the Indiana Department of Transportation. The contents of this paper reflect the views of the authors, who are responsible for the facts and the accuracy of the data presented herein, and do not necessarily reflect the official views or policies of the sponsoring organization.