Assessing Satellite-Augmented Connected Vehicle Technology for Security Credentials and Traveler Information Delivery

Abstract

Vehicle-to-Everything (V2X) technology has the capability to enhance road safety by enabling wireless exchange of telematics and spatiotemporal information between connected vehicles (CVs). Effective V2X communication depends on rapid information sharing between Roadside Units (RSUs), in-vehicle On-Board Units (OBUs), and other connected infrastructure. However, there are increasing concerns with RSUs related to installation needs, reliability, and coverage, especially on rural roadways. This study aims to evaluate the benefits of augmenting CV infrastructure with satellite technology in situations where RSU access or coverage is limited while maintaining V2X security protocols and critical information exchange. The study utilizes data from over 400 personal, fleet, and commercial CVs collected during two real-world pilot deployments in the United States, one in an urban environment in Florida and one in a rural environment in Wyoming. The analysis performed shows that the delivery of critical security credential information and traveler information messages (TIMs) to CVs is dependent on a multitude of environmental and operational reliability factors. Overall, information delivery is faster with dense RSU infrastructure as compared to satellites. However, we show that by augmenting RSU infrastructure with satellite technology, the delivery of information is more robust, improving V2X system reliability, security, and overall road safety.

1. Introduction

Vehicle-to-Everything (V2X) technology has the capability to increase safety by enabling wireless exchange of telematics and spatiotemporal information between connected vehicles (CVs). An in-vehicle On-Board Unit (OBU) and software compute if nearby vehicles pose a collision threat and provide advance warning to the driver or, for vehicles with autonomous capability, enable the vehicle to avoid the collision on its own. Paired with intelligent infrastructure like Roadside Units (RSUs), OBUs can communicate with infrastructure to receive security information and Traveler Information Messages (TIMs) containing weather alerts or notifications about impending roadway conditions to further improve safety [1].

Intrinsic to this technology is the handling of secure transmission of data [2,3,4,5]. In the CV ecosystem, the security of the message exchange protocol is ensured by a Security Credential Management System (SCMS). The SCMS provides and manages the infrastructure for V2X message security. It consists of back-end systems that generate and distribute certificates that are used by OBUs to sign and validate V2X messages [3,4]. The SCMS also collects reports of misbehavior and uses that information to create a list of revoked certificates or Certificate Revocation List (CRL). In addition, the SCMS generates and distributes a policy file that includes security policies with rules and limits about how certificates and the CRL are to be used [6,7]. For example, one policy that it includes states how old a CRL may be and if it is still considered to be current (currently set to 2 weeks). Based on this policy, an OBU that has not been in proximity to an RSU for a long enough time and, therefore, has not downloaded a new CRL in two weeks would have to stop validating and acting on V2X messages received. Receiving this critical CRL via an alternative method can be of advantage for this OBU as the CRL can be received outside RSU coverage areas [8,9].

In prioritizing RSU deployment, there is often a focus on locations with access to the electrical grid and connectivity, making installations more efficient. However, this approach can lead to communication dead zones, causing CRLs to expire without timely updates [10]. Given this approach to testing and deploying CV technology, rural areas may face limited RSU installations, potentially reducing the effectiveness and reach of these systems. Rural areas have also been noted to experience a disproportionately large percentage of traffic fatalities as compared to urban roads, highlighting the immediate need for V2X [11,12,13]. In 2019, prior to the onset of the COVID-19 pandemic, out of 36,096 motor vehicle traffic fatalities, 16,340 (45%) occurred in rural areas. Considering that only 39 percent of the total U.S. vehicle miles of travel were in rural areas, this translates into a fatality rate, measured in 100 million vehicle miles of travel, that is over 1.9 times higher than that of urban areas (1.66 versus 0.86).

The United States Department of Transportation (USDOT) anticipated this issue by considering augmenting RSUs with additional communications resources, such as cellular and satellite [14]. Augmentation of V2X with an existing communications infrastructure could enable states and local agencies with limited funds to build out V2X to realize the full benefits in both rural and suburban areas. The USDOT has funded large-scale V2X pilots to capture operational metrics of the V2X system in different deployment environments and to quantify aspects of system performance and safety benefits [15]. These pilots were equipped in part with OBUs incorporating satellite receivers to augment RSU communications.

The goal of this study is to present the analysis results of the satellite and RSU reception metrics logged by participant vehicles during the operation of these pilots. Satellite and RSU reception metrics for the security-related CRL were captured in the Tampa Hillsborough Expressway Authority CV Pilot (THEA CV Pilot), while TIM reception metrics were captured in the Wyoming Department of Transportation (WYDOT) Connected Vehicle Pilot (WYDOT CV Pilot) [16,17].

The manuscript is presented in five sections. Following the introduction, a comprehensive literature review detailing V2X systems and processes is presented. The data and methods section then outlines the experimental setup for the two distinct pilot studies, contrasting the urban core of downtown Tampa, Florida, with the expansive Wyoming Interstate 80 (I-80) corridor. The results section first compares the performance of RSU and satellite technology in terms of delivering the necessary security services using real-world CV deployment data collected from a panel of participants in the THEA CV Pilot. This is then followed by assessing a vast array of TIM alerts generated by the WYDOT CV Pilot and dispatched to travelers via both satellite and RSUs, providing timely weather and traffic alerts to travelers along the I-80 corridor. The discussion then proceeds to explore the study’s limitations, outlines potential areas for future work, and concludes with a summary of key lessons learned.

2. Literature and Research Need

In the V2X system, the OBU broadcasts a Basic Safety Message (BSM) up to ten times per second, which provides the vehicle location, direction, and speed to nearby OBU-equipped vehicles and RSUs. RSUs typically communicate via Wireless Local Area Network (i.e., Dedicated Shortrange Communication (DSRC)) or Cellular network (i.e., Cellular Vehicle-to-Everything (C-V2X)) and serve a wide range of functions in addition to broadcasting BSMs and back-end data transfer, such as delivering high-frequency Signal Phase and Timing (SPaT), personal safety messages (PSMs), and intersection geometry (MAP) messages essential for numerous V2X mobility and safety applications [18,19]. They also support various informational services, including TIMs, which assist with traffic speed harmonization and overall situational awareness. Unlike RSUs, other forms of delivery, such as satellite communication, cannot support the low latency and high-frequency message exchange essential for certain safety applications [20]. However, satellite communication can be effective in the delivery of other critical security and informational messages, enhancing system resilience. Figure 1 provides an overview of each component’s role and the general structure of message exchange within the V2X environment.

In any V2X system, the generated messages are secured through the SCMS, which issues unique security certificates to each OBU with a requirement for the OBU to append a certificate to each outgoing BSM. The SCMS removes bad actors or mis-operating OBUs that may disrupt the system by maintaining a CRL, which is updated as misbehavior and is detected and distributed to all OBUs via available vehicle-to-infrastructure (V2I) communications. Misbehavior is defined as any case where an OBU generates a BSM containing one or more values inconsistent with the corresponding vehicle’s true status, position, or behavior [6,21]. OBUs determining incoming BSMs can be trusted by validating their certificate and by checking them against the CRL. The primary risk of misbehavior lies in the possibility of authenticated bogus messages, which can generate false alerts and undermine system trustworthiness [3,7].

Successful CRL delivery hinges upon the availability of CV RSU infrastructure to ensure proper coverage. Proper coverage ensures that vehicles are always within range to receive the latest security credentials, which is essential for maintaining system integrity and protecting against misbehavior in the network [3,22]. A 2017 study used to support the USDOT-proposed rule to mandate vehicle-to-vehicle (V2V) communication in light-duty vehicles estimated that V2V RSU build-out would have a linear phase in a period of over 15 years [14]. As a result, over the first few years of V2V system operation, the RSU density would be highly limited, during which time early OBU deployments could benefit from in-vehicle systems that leverage an alternate communications infrastructure to augment the RSU network. The study projected that, at full build-out, the RSU network would comprise 19,750 RSUs, covering less than 75 percent of the nation’s population. States with limited budgets may elect to locate RSU installations in high-traffic density areas to maximize the number of vehicles covered daily at the expense of rural area coverage [14].

The advantages of augmenting the V2X RSU network with satellite broadcasts have been described by SiriusXM. SiriusXM is a subscription satellite radio and infotainment service provider in North America that operates a dedicated broadcast system consisting of two geostationary satellites augmented with approximately 600 terrestrial repeaters to provide robust signal delivery throughout urban and rural areas. SiriusXM has proposed to dedicate a nationwide data channel on their satellite network to support the distribution of multiple data services to OBUs without the requirement of a subscription. According to SiriusXM, the satellite broadcasts provide seamless in-vehicle coverage of the contiguous U.S. with approximately 99% reliability [23]. This coverage footprint would provide a communication path to OBUs in rural areas and other areas outside of the RSU range. Since vehicle original equipment manufacturers (OEMs) are factory-installing satellite radios in over 80% of new vehicles sold in the U.S., the hardware cost impact for OBUs to leverage data from the satellite broadcast would be reduced. SiriusXM joined the USDOT’s CV Affiliated Test Bed Program in 2014 and has developed a satellite-augmented V2X OBU to support various USDOT and State Pilots.

There is significant interest in deploying CV technologies on rural corridors, which typically feature long stretches of roadway through sparsely populated areas with limited power, communications, Intelligent Transportation Systems (ITSs), and, thus, RSU infrastructure. This need is highlighted in the USDOT National Plan released in mid-2024, aimed at saving lives through improved connectivity by accelerating V2X deployment [24]. Agencies servicing rural regions need scalable solutions to deploy CV technologies and extend the safety benefits (i.e., information on approaching road hazards, maximum safe speeds, road closures, incidents and other emergencies, and traffic congestion) to travelers along their corridors [25,26]. Based on the noted challenges in V2X connectivity, USDOT National outlook, and technological advances, we utilize data generated from real-world CVs to explore the security and operational benefits of satellite augmentation in the V2X ecosystem.

3. Data and Methods

The study employs data from two CV Pilot deployments, the THEA CV Pilot in Tampa, Florida, and the WYDOT CV Pilot along the I-80 corridor in Wyoming. Downtown Tampa and the I-80 corridor in Wyoming offer contrasting environments, which are particularly valuable for the purposes of this study. Downtown Tampa and its surroundings represent a densely populated and infrastructure-rich urban setting with a high concentration of RSUs. This is characterized by significant variations in traffic flows due to tourism and sports arenas, complex signalized intersections, and frequent interactions among vehicles and vulnerable road users. In contrast, the I-80 corridor in Wyoming is a sparsely populated rural environment characterized by long stretches of highway and high annual truck traffic volume (30% to 55% of total volume) [27].

First, we use the THEA CV Pilot dataset to test the hypothesis of satellite technology being capable of timely delivery of CRL certificate when the OBUs are out of range of an RSU and compare the performance of satellite delivery when OBUs enter in range, i.e., the vehicles approach the urban area where the RSU infrastructure is installed. Second, we use the WYDOT CV Pilot data to analyze the performance of TIM delivery via RSUs coupled with augmented delivery via satellite dispatch. Figure 2 provides a methodological overview of the study, highlighting the various CV infrastructures implemented during the two pilot deployments.

3.1. THEA CV Pilot

The THEA CV Pilot implemented an experimental design to enroll more than 1000 private citizens to experience the functionality and impact of several V2V and V2I applications by having their vehicles retrofitted with OBUs and audiovisual human machine interface (HMI) devices to deliver a host of safety and mobility advisories [10]. Drivers were selected to represent commuters traveling from all over the Tampa Bay Area to downtown Tampa, Florida. Participant selection was performed using a stratified sample based on travel behavior (e.g., frequency of travel to the study area) and socio-demographics (i.e., age, race, gender, education, and income).

3.1.1. Data Generation

The THEA CV Pilot OBUs had the capability to record and locally store all data generated by the vehicle and any information received by nearby CV infrastructure, such as SPaT, MAP messages (provides intersection and roadway geometry information), and TIMs from CV-equipped signalized intersections, and other OBU-equipped vehicles (i.e., sent and received BSMs). These data were packed and compressed into the so-called OBU Data logs following Society of Automotive Engineers (SAE International) standards [28,29]. Unique to the THEA CV Pilot, as the participant vehicles entered into the range of an RSU, the OBUs uploaded the logs over the air to a secure server. Each log included all data generated and received while the vehicle was operational. These logs can be thought of as ongoing travel diaries, with information recorded from one to up to ten times per second (10 Hertz) to enable assessing the safety and mobility impact of ten V2V and V2I applications [10] and report on system performance [30] or the impact of exogenous short- and long-run shocks on the demand for travel.

During the last two phases of the Pilot (Phase 3 and 4), an over-the-air firmware update was installed on a select pool of 100 aftermarket OBU devices to allow logging of CRL downloads. This provision, unique only to the THEA CV Pilot, allowed the testing and comparison of dispatching CRLs via satellite versus via RSUs. Figure 3 illustrates the layout of existing RSU infrastructure in downtown Tampa, highlighting the geofenced study area and system status indicators (i.e., real-time operational status of RSUs and spatial distribution of BSMs generated by OBU-equipped vehicles).

3.1.2. Experiment Design

The comparative tests were conducted by measuring the time it took to receive a complete CRL file from the moment an OBU-equipped vehicle was turned on. A 400KB mockup file was used to replicate satellite delivery metrics of a fully loaded CRL. Each time the vehicle was started, the OBU initiated a power-up cycle that records the following information in a system monitor message recorded in the OBU Data log: timestamp (UNIX time in milliseconds) the OBU was powered up, timestamp the CRL was received from a satellite, timestamp a given RSU was first seen, and timestamp the OBU was powered down. If the OBU did not receive a CRL during a power cycle either via satellite or RSU, the first two data elements would not be logged. For simplicity, CRL availability from the RSU was assumed to be concurrent with RSU availability.

Before statistical analysis, we eliminate outliers and add weather controls and RSU system performance measures. The weather controls account for the variability typical of Florida’s subtropical climate, characterized by hot and humid conditions from mid-May through mid-October coinciding with the rainy season. Summertime weather is consistent from June through September and is characterized by mid-afternoon thunderstorms with frequent lightning. These thunderstorms may last for only a few moments to several hours or even for an entire day. Localized weather conditions can introduce spatially heterogeneous effects on travel and system behavior that need to be carefully controlled for in the analysis.

To test the hypothesis of rain and cloud cover affecting the transmission and receival of CRLs, the OBU Data are augmented using weather data at 10-minute intervals from World Weather Online, including variables such as cloud cover intensity, rain intensity, and wind bearing.

The performance of the RSU infrastructure must also be accounted for in the analysis. There were 47 dual-band (i.e., DSRC and C-V2X) RSUs installed in the Tampa deployment. At any time, a given RSU might experience failure, needing complete replacement or only needing to have the system rebooted to resume operations. In addition, RSUs vary in terms of signal range because of their installation location and because of the surrounding built-environment infrastructure. To account for this heterogeneity, we measure the RSU range using a 95-percent confidence ellipse around the location of each RSU using the BSMs received over the entire day.

After merging the OBU Data Logs (CRL payload data) with the weather API data and the RSU system performance data and after performing initial data cleaning to remove outliers, the final sample consists of 15,555 observations from a panel of 100 vehicles over the period of 1 June 2021 through 23 June 2022. The following variables were extracted from the final dataset for statistical analysis:

• Dependent variables: crl_sat_time_e (capturing the time, in seconds, elapsed from OBU power-up to receiving a CRL from the satellite); rsu_avl_time_e (capturing the time, in seconds, elapsed from OBU power-up to receiving a CRL from the RSU).
• Independent variables: run_time_tot (total run time from engine on to engine off in minutes); rsu_dist (distance, in meters, to a RSU at CRL receival); rsu_on_day (share of RSUs actively operating daily expressed as a percentage); rsu_off_day (share of daily non-operational RSUs expressed as a percentage); rsu_range_m (average operational coverage range of the RSU delivering a CRL in square miles); cloud_cover (percentage of sky obscured by clouds during a given run); rain_int (volume of rainfall during a given run in cubic millimeters).
• Random effects: OBU ID (static identifier used to capture unobserved heterogeneity between multiple observations from the same vehicle/OBU unit).

3.1.3. Statistical Modeling

The statistical modeling approach employs Ordinary Least Squares (OLS)—Equation (1)—and Random Effects (RE)—Equation (2)—regression models to estimate the impact of various independent variables on the time it takes for an OBU to receive a CRL from either a satellite or RSU. The OLS model offers a straightforward initial estimation by assuming each observation is independent, allowing for an assessment of how each independent variable individually influences the dependent variables in this real-world dataset [32]. However, it does not account for repeated observations from the same OBU over the course of the study, which may lead to bias. The RE model is best suited in this case, given the panel data structure (same OBUs observed over the study period), as it assumes that there exists unobserved heterogeneity that is uncorrelated with the explanatory variables [32,33,34].

$$Y_{it}={\beta }_0+{\beta }_1{X}_{1it}+{\beta }_2{X}_{2it}+\dots +{\beta }_{\mathrm{k}}{X}_{kit}+e_{it}$$

(1)

where

• $Y_{it}$ is the dependent variable (i.e., crl_sat_time_e or rsu_avl_time_e) for OBU ID i at time t,
• ${\beta }_0$ is the fixed intercept or constant term,
• ${\beta }_1,{\beta }_2,\dots ,{\beta }_k$ are the coefficients for each predictor/independent variable,
• $X_{1it},X_{2it},\dots X_{kit}$ are the independent variables specific to each OBU observation, i, at time t, and
• $e_{it}$ is the error term, assumed to be independent of each observation.

$$Y_{it}={\beta }_0+{\beta }_1{X}_{1it}+{\beta }_2{X}_{2it}+\dots +{\beta }_{\mathrm{k}}{X}_{kit}+u_i+e_{it}$$

(2)

where

$u_i$ represents the random effect associated with the i^th OBU ID, capturing the unobserved heterogeneity across the OBU-equipped vehicles.

3.2. Wyoming CV Pilot

Next, this study looks at how the V2X system performs in terms of broadcasting weather advisories to travelers in rural areas. The goal is to provide perspective on (1) the use of CV technology to improve safety conditions and (2) to verify how the system performed in terms of providing advisories via RSU equipment and satellite technology. Wyoming ranks No. 20 in the U.S. in terms of fatalities (1.72 per 100 million vehicle miles of travel). During the winter season, heavy snow, along with strong wind speeds and wind gusts, are causes of increased crashes, particularly with commercial vehicles [27].

3.2.1. Data Generation

The WYDOT CV Pilot deployed V2X communication technology to implement a broad range of advisories, roadside alerts, parking notifications, and dynamic travel guidance with the goal of improving the safety, mobility, and productivity of the users of I-80 [35,36]. The Pilot installed 75 DSRC-based RSUs (with one additional test RSU not along the interstate) along a 402-mile-stretch of the I-80 and equipped about 320 fleet and commercial vehicles with OBUs and weather sensors, as shown in Figure 4 [27,37]. The OBUs had the capability to receive alerts and advisories via RSU or satellite-based TIMs. The WYDOT CV Pilot used a different system to collect and archive CV data from RSU and OBU-equipped vehicles compared to the THEA CV Pilot. Broadcasted TIMs are uploaded as they are generated and made available to the research community via the USDOT ITS DataHub [38].

3.2.2. Experiment Design

To meet the goals of improving safety and mobility by addressing adverse weather conditions and emergency situations, the WYDOT CV Pilot deployed a suite of services, alerts, and advisories throughout the I-80 corridor. The service delivery system is through the dissemination of TIMs using RSU and satellite. TIMs consist of standard International Traveler Information Systems (ITIS) codes and are broadcast with a periodicity of 1 s [40]. Based on the existing or future conditions, ITIS codes relevant to a specific region are configured by the traffic management center controller and encoded as a TIM, which is then broadcast by the RSU or satellite. The TIMs for this region contain various ITIS codes and highlight how the CV system was distributing real-time information from the road conditions sensor system related to adverse weather and emergencies over the course of the winter and into the spring of 2022.

To evaluate how RSU and satellite infrastructure operated while delivering the various advisories, the WYDOT CV Pilot team provided a sample consisting of TIMs encoded over the period of 1 January 2022 through 1 May 2022. After cleaning the sample for redundant messages, the dataset consisted of 113,056 ITIS-encoded TIMs ready for delivery, as shown in Figure 5.

Note that each advisory can contain one or more ITIS codes within a TIM. The most frequently dispatched TIMs were weather-related safety advisories consisting of icy patches (19%), followed by strong winds (15.2%) and snow (11.7%).

3.2.3. Inference Approach

From 113,056 ITIS-encoded TIMS, 898,373 were readily delivered to road users. It is also observed that satellite delivery accounted for a much larger share of the total TIMs delivered (56.6% compared to RSUs 43.4%). To further establish a preliminary understanding of the rate of delivery of these encoded TIMs, we bin the dataset at 1 s intervals, as shown in Figure 5. This is performed to evaluate the number of TIMs delivered to end users in the shortest possible time from initial encoding, enabling greater situational awareness and prompt response to critical alerts. From Figure 6, it is visually evident that TIMs are delivered at consistently higher rates via satellite throughout the study period, with particularly elevated rates during the harsher winter months from January to March.

The binned dataset is subsequently utilized to generate cumulative distribution functions (CDFs), which provide a detailed comparison of the rate and distribution of TIM delivery to end users via both RSUs and satellites.

4. Results and Discussion

4.1. THEA CV Pilot CRL Downloads Analysis

Table 1. Sample descriptive statistics of CRL data.

Variable	Label	Mean	St. Dev.	Min	Max
crl_sat_time_e	Time elapsed from power-up to receive CRL from satellite (seconds)	270.03	55.93	27.78	500.80
rsu_avl_time_e	Time elapsed from power-up to receive CRL from RSU (seconds)	906.36	616.47	13.88	4012.54
run_time_tot	Total run time (minutes)	46.99	65.79	6.77	582.20
rsu_dist	RSU distance at CRL receival (meters)	8547.48	6432.52	70.06	28,898.37
rsu_on_day	Share of RSUs operational (daily)	79.79	12.74	17.02	91.49
rsu_off_day	Share of RSUs not operational (daily)	20.22	12.73	8.51	82.98
rsu_range_m	Average RSU range (sq. miles)	0.11	0.03	0.05	0.20
cloud_cover	Cloud cover (%)	0.54	0.24	0.03	1.00
rain_int	Rain volume (cubic mm)	0.004	0.042	0.000	1.525

Note: St. Dev. is the standard deviation.

provides sample descriptive statistics. The table shows that, on average, when a vehicle was powered up and not in range of an RSU, it took about 270 s to receive a CRL via satellite. On the other hand, given the fixed location of the RSUs, mostly clustered in the study area (i.e., Tampa downtown), it took a vehicle, on average, approximately 906 s (about 15 min) to latch onto an RSU and (as assumed) receive a CRL. By looking at the standard deviation and range, it appears that starting a vehicle within the study area and in proximity to an RSU could, in some instances, reduce the time to obtain a CRL to a mere 13.88 s. This is also confirmed by the variable, rsu_dist, measuring distance to the nearest RSU at the time of CRL receipt, with an average distance of about 8548 m.

The augmenting variables to control for RSU performance are also important to note. On average, for the entirety of the study period, RSUs were not operational 20.2 percent of the time. The mean range of an RSU is about 0.11 square miles. The weather variables reflect the range of conditions throughout the study period and the Tampa Bay area.

Next, we compute and plot the CDFs of satellite and RSU performance in broadcasting CRL payloads. By plotting the cumulative probability, the CDFs assess how quickly CRLs are typically delivered and the variability in delivery times. This visual representation aids in comparing the proportion of observations that fall below specific time thresholds, offering insights into the efficiency and consistency of both delivery methods. Figure 7 shows that satellite-based CRL delivery times remain relatively consistent, with minimal variation across 10% to 95% of observations (ranging from 220 to 320 s). In contrast, the CDF for CRL delivery via RSU shows greater variability, with delivery times being more dispersed. Specifically, half of the RSU observations cluster around a median time of approximately 820 s, indicating a wider range in delivery efficiency compared to the satellite. This suggests that, on average, RSUs have a lower CRL delivery efficiency.

Additionally, when vehicles transition or are powered up within the downtown Tampa area where most RSUs were located, the CDF (shown in Figure 8) of CRL received via the satellite remained consistent with the overall observations shown in Figure 6. However, the CDF of CRL received via RSU demonstrates that up to 90% are received within 225 s, with a median value of 31 s, significantly faster than outside this area and satellite delivery. The noted differences warrant exploring the relationship between observed CRL received times and the observed factors that might be affecting it, such as RSU location, distance, coverage and operational reliability, and weather conditions. This analysis is carried out by regressing the CRL received times via satellite (crl_sat_time_e) and via RSU (rsu_avl_time_e) against these controls.

Table 2. Results of the regression analysis.

Variable	Time to Receive CRL
	Satellite		RSU
	(1)	(2) ƚ	(3)	(4) ƚ
Total run time (minutes)	0.00358	0.0123 *	1.006 ***	0.815 ***
	(0.00686)	(0.00715)	(0.0562)	(0.0555)
RSU distance at CRL receival (meters)	−0.00000944	0.000103	0.0570 ***	0.0566 ***
	(0.0000735)	(0.0000847)	(0.000602)	(0.000670)
Share of RSUs not operational (daily)	−0.00104	0.00980	2.038 ***	2.110 ***
	(0.0456)	(0.0449)	(0.373)	(0.345)
Average RSU range (sq. miles)	34.49	34.98	−141.2	−160.1
	(32.86)	(32.37)	(269.0)	(248.9)
Cloud cover (%)	4.052 **	4.262 **	23.85	28.74 **
	(1.925)	(1.894)	(15.76)	(14.56)
Rain volume (cubic mm)	−14.85	−16.82	−70.97	−109.7
	(10.72)	(10.54)	(87.77)	(81.03)
Within study area (1 yes, 0 otherwise)	17.10 ***	20.91 ***	−350.7 ***	−362.4 ***
	(2.108)	(2.348)	(17.26)	(18.46)
Constant term	261.6 ***	259.6 ***	319.1 ***	349.4 ***
	(4.946)	(5.000)	(40.50)	(42.10)
Observations	15,555	15,555	15,555	15,555
R-squared	0.10	0.09	0.45	0.394

Standard errors in parentheses; * p < 0.05, ** p < 0.01, *** p < 0.001; month dummy variables not reported for brevity; ^ƚ Hausman test: Model (2): chi²(17), Prob > chi² = 0.378; Model (4) chi²(17), Prob > chi² = 0.904.

reports the results. The naïve estimators of Model (1) and Model (3) employ OLS regression. The preferred models are Model (2) and Model (4), which are regressed using RE panel regression. Hausman tests against the fixed-effect alternative fail to reject the null hypothesis [41]. Model (2) estimates the time to receive a CRL via satellite after controlling for the CV infrastructure, weather conditions, and time trends to control for seasonality.

As expected, in Model (2), the RSU infrastructure (RSU distance and operational performance) has no effect on CRL receival time via satellite. The dichotomous variable to flag the CRL being received within the THEA CV Pilot study area is statistically significant and has a positive sign. The magnitude of the estimated parameter suggests that all else being equal, CRL messages downloaded within the study area take, on average, 17 more seconds. Weather conditions, such as increased cloud cover, affect CRL reception via satellite by causing longer download times.

Model (4) regresses the time to receive a CRL via RSUs, controlling for the same factors. All RSU infrastructure controls are statistically significant. As the distance from an RSU increases, so does the CRL download time, noting that this is subject to threshold value due to RSU maximum broadcasting ranges. It is relevant to note the impact of RSU’s operational performance. All else equal, the model estimates an increase of about two seconds for each one-percent increase in the percentage of RSU out of commission in a given day. A large magnitude is expected, given the number of RSUs installed (47). As in the case of satellite transmission, cloud cover has an impact on the transmission of CRL to the OBUs. Starting a vehicle within the study area results in a dramatic reduction in download time, as shown by the negative sign and value (−362.4) of the study area dummy variable. In the following subsections, we thoroughly explore the impact of weather and RSU infrastructure on CRL transmission by estimating the marginal effects.

4.1.1. Effects of Weather on Satellite and RSU CRL Transmission

Weather impacts CRL message reception either via satellite or RSU, but the relative magnitude differs. Figure 9 shows the marginal effects of weather by computing the predicted mean of the outcome variables when cloudiness increases. To appreciate the impact, the cloud cover impact is estimated at its distribution quartile increments over the predicted means. The graphs show that as the cloud cover increases, so does the average time to download a CRL, but the effect is less if the CRL is transmitted via satellite compared to RSUs. From an open sky to a completely clouded one, CRL dispatched via RSUs took about 29 s longer compared to a four-second increase if broadcasted via satellite.

4.1.2. Effects of RSU Infrastructure on RSU CRL Transmission

We assess the impact of infrastructure first by comparing the predicted changes in the estimated CRL download times via RSUs in instances where the vehicle was either inside or outside the THEA CV Pilot study area (downtown Tampa). Second, we assess the impact on CRL download times at different rates of RSU performance in terms of the percentage of RSU being operational on a daily basis.

Most of THEA CV Pilot RSUs were clustered downtown in the Tampa Central Business District. In the sample, vehicles traveling within the downtown area are, on average, about 144 m from the nearest RSU. Figure 10a displays the marginal effects. The predicted margins, ceteris paribus, show that OBUs powering up or traveling within the study area could download CRL payloads within 107 s, compared to about 946 s for those OBUs outside the study area. The figure also hints that in urban environments with established CV infrastructure and assuming average operational efficiency, RSUs can perform comparatively better than satellites in dispatching CRLs. On the other end, vehicles outside of the study area were likely to be out of range of RSUs, a scenario similar to the current national infrastructure in rural areas where vehicles, even if equipped with OBUs, are likely to be out of range of any CV or other intelligent infrastructure.

Relaxing the assumption of constant operational efficiency is a necessary next step. We consider the impact of RSU’s operational performance. This is important because there is an interest in understanding the relevance of CV system performance in the context of scalable V2X deployments, especially in early phases characterized by a reduced number of RSUs or in rural areas characterized by lower density of transportation capital infrastructure. In the THEA CV Pilot, RSU operational performance varied throughout the study period. System performance measures indicate that, on average, 20% of the RSUs were not fully operational at any given time, with variability registered monthly. There were multiple causes for this, mostly because replacing an RSU requires obtaining the necessary permits to access intersection poles, thus extending replacement times. In addition, Florida’s tropical weather, characterized by frequent lightning, resulted in a high number of RSUs being replaced at various times.

Figure 10b shows the predicted CRL means at various levels of RSU performance, as measured by the percent of RSU that can be offline at any point. The margins show the predicted impact at various levels of RSU outage. Up to about 20% (i.e., the sample mean), the impact on CRL download times does not increase. As the percentage increases above 25 percent, there is a rapid deterioration in CRL download times. In a CV infrastructure where RSU and satellite technology are simultaneously deployed, variability in operational efficiency can be compensated by the two systems.

4.2. WYDOT CV Pilot TIMs Analysis

To get a more comprehensive overview of satellite and RSU TIM broadcasting, Figure 11 displays the CDF. The data show that satellite delivers TIMs at higher rates than RSUs. At the 50th percentile of the distribution, satellite delivery dispatches up to seven TIMs, while RSUs deliver approximately two TIMs. Beyond the 95th percentile, satellite dispatches up to 21 TIMs compared to only 10 delivered by RSUs to the end users. This highlights the critical role of satellite augmentation in rural and more dispersed V2X environments.

5. Conclusions

Timely delivery of CRL within an established Security Credential Management System is essential to an effective and secure CV system. The performance assessment of CRL delivery via satellite and RSU using real-world data from 100 OBUs in the THEA CV Pilot shows that the two delivery systems can provide complete coverage while overcoming existing CV infrastructure constraints, especially in rural areas. In urban areas, with deployment ramping up or with already established CV infrastructure, RSUs installed at intersections can provide a variety of services, from the delivery of high-frequency SPaT and MAP messages that are critical to several V2X applications to information services and other speed-harmonizing applications through effective delivery of TIMs. At the same time, satellite technology can be a cost-effective solution to overcome RSU operational reliability issues, as demonstrated in this study.

In other less densely populated rural areas, integration of satellite-based technology into the CV ecosystem provides added mobility and safety benefits. In environmental conditions characterized by adverse weather, CV technology proved effective in providing travelers with timely advisories, as exemplified by the WYDOT CV Pilot. Moreover, for local transportation agencies adopting CV technology, having both satellite and RSU infrastructure offers a robust solution to establish and maintain CV coverage across a broader geographical area without widespread RSU installations.

Although the urban core analysis performed in this study was comprehensive, the limitations of the rural dataset should be acknowledged. These limitations prevented the augmentation of other critical data sources, such as environmental barriers (e.g., trees) and terrain variability. These factors could impact the performance of both satellite and RSU technologies in remote areas. Further investigation with a broader dataset is needed to fully understand these dynamics in rural environments. Additionally, the results observed in the urban core may not fully apply to regions with different climates, such as those experiencing snow or prolonged periods of heavy rainfall/winds, which are not typical of Florida.

Future research could focus on integrating additional geographical, environmental, and traffic variability to gain a better understanding of the information delivery reliability associated with RSUs and satellites. Additionally, comparing the performance of traditional RSUs with newer, more efficient, edge-processing-enabled RSUs would provide valuable insights into the potential improvements in delivery reliability. Research focused on the scalability of hybrid RSU–satellite systems across different regions, as well as an analysis of the cost-effectiveness and sustainability of various configurations, would aid in better supporting the decision-making for large-scale implementations.

Overall, the study demonstrates that CRLs are delivered more quickly in an urban core with dense RSU coverage compared to delivery via satellite. However, the limited availability of RSUs in rural areas, ongoing reliability challenges, and dependence on vendor maintenance make satellite delivery of CRLs and TIMs essential for the future of CV deployments. This is particularly evident in regions prone to severe weather, where augmenting CRL and TIM delivery has proven to be highly valuable for ensuring that OBUs stay current, continually delivering safety and mobility benefits.