Technical sciences

Melnyk Ihor

Master’s Degree in Law

Luhansk State University of Internal Affairs named after E.O. Didorenko

ORCID: 0009-0001-3690-7611

https://www.doi.org/10.25313/2524-2695-2026-5-02-53

DYNAMIC DRIVER RISK SCORING BASED ON TELEMATICS DATA IN RENTAL FLEETS

Summary. The article is devoted to the search for algorithms that allow drivers and fleet operators to reduce road risks based on telematics data, behavioral analysis and risk scoring. The relevance of the topic is due to the fact that in the conditions of digitalization of the short-term car rental market, driving safety can no longer be ensured only by traditional control of violations or technical monitoring of the vehicle. Dynamic assessment of driver behavior is becoming more important, which allows identifying risk patterns directly during the trip. The purpose of the article is to substantiate the features of dynamic assessment of driving risk based on telematics data in short-term car rental fleets on the US market. The study used general scientific methods of analysis, generalization, comparison and systematization of scientific literature. The author’s methodology for comprehensive risk assessment is also presented, which can be adapted to the conditions of the US market and integrated into modern telematics, insurance and fleet-management systems. The article also shows that approaches to assessing driving risk have different construction logics. Some models are focused on the driver’s perception of danger, while others use sensory signals, a smartphone, GPS or OBD-II data to recognize driving style. This allowed us to show that it is advisable to assess the risk in rental fleets not as a single event, but as a sequence of driver behavioral reactions within a specific trip. It is this approach that creates the basis for earlier detection of unstable or aggressive driving behavior. The practical significance of the study is that the proposed methodology can be used by international fleets, short-term car rental companies, insurance companies that use usage-based insurance models, as well as state and municipal transport systems. Its strengths include the ability to prevent fleet accidents through early detection of risky behavior, reduce insurance and repair costs, create an AI-model of behavioral risk scoring of the driver, scale to various telematics and fleet-management platforms, and transform fleets into data-driven safety and risk management systems.

Key words: fleets, risks, car rental, driving safety.

Introduction. The relevance of the research topic is due to the fact that transport safety in the conditions of digitalization of the mobility sector is gradually moving from classic control of violations to predicting risky driver behavior. For the car rental market, this is of particular importance, since the vehicle is transferred for use to different drivers who have different experience, driving style and level of adaptation to the car. Risk perception depends not only on the road situation, but also on the individual characteristics of the driver, speed, distance and workload [1]. Therefore, risk assessment in a rented car fleet cannot be limited only to the technical condition of the car or the formal history of violations. An additional factor of relevance is the change in the very model of the car rental market. According to Grand View Research and IMARC Group, this market is actively moving to digital channels of interaction with customers, automated booking, mobile services, contactless access to the car and telematic fleet management [7; 8]. In such conditions, companies receive more data about the process of using transport, but at the same time they are faced with the need to correctly interpret this data. The fact of digitalization itself does not reduce the accident rate. Its practical value arises when the collected data is transformed into a tool for early risk detection. 

Literature Review. The issue of dynamic risk assessment of driving in short-term rental fleets is not sufficiently covered in detail in the scientific literature. Nevertheless, individual components of this topic are actively considered by scientists. In particular, the problem of driver risk perception was studied by S. S. B. Masud et al. [1], X. He, J. Stapel, M. Wang and R. Happee [3], M. Kokubun et al. [4], P. Ping et al. [5], X. Zhao et al. [6]. These authors analyzed how the driver assesses the danger while driving, how this is influenced by speed, distance, maneuvers of other vehicles, trust in automated systems and the microscopic state of the road situation [1; 3; 4; 5; 6].

As for the relationship between risky behavior and the psychophysiological load of the driver, this issue was investigated by L. L. Di Stasi, V. Álvarez-Valbuena, J. J. Cañas, A. Maldonado, A. Catena, A. Antolí and A. Candido [2]. In their study, risky behavior is considered in relation to mental workload, which is important for assessing the driver’s state in conditions of tense or unstable driving [2]. Within this framework, S. S. B. Masud et al. also show that changes in workload can affect the perception of risk, and therefore the driver’s ability to respond in a timely manner to a dangerous situation [1]. This is particularly important for rental cars, as the user is often driving a vehicle to which he has not yet fully adapted.

A separate block of research is devoted to driving style recognition and telematic driver profiling. This issue was investigated by D. A. Johnson and M. M. Trivedi [9], G. Castignani, T. Derrmann, R. Frank and T. Engel [10], J. E. Meseguer Anastasio, C. M. Tavares de Araujo Cesariny Calafate, J. C. Cano Escribá and P. Manzoni [11]. Johnson and Trivedi considered the smartphone as a sensory platform for driving style recognition, in particular through the analysis of signals that allow distinguishing normal from aggressive driving [9]. Castignani et al. investigated driver behavior profiling using smartphones as an accessible monitoring platform, where risk events can be recorded through acceleration, braking and other driving parameters [10]. Meseguer Anastasio et al. proposed an approach to assessing driver behavior using the DrivingStyles mobile application, which takes into account driving style and road context [11].

Statistical and expert sources complement the scientific basis of the study with a market context. According to Grand View Research, the car rental market in the United States is developing in the context of strengthening digital channels, service automation and technological renewal of car fleets [7]. IMARC Group also points to the growing role of digital solutions, mobile services and new models of interaction between the client and the rental operator [8]. These sources confirm that telematics, digital platforms and driver behavior analytics are becoming not secondary tools, but part of the market logic of rental fleet development [7; 8].

Problem Statement. The purpose of the article is to substantiate the features of dynamic assessment of driving risk based on telematic data in short-term car rental fleets. To achieve the goal, the following tasks will be performed during the study: to determine the relevance of the problem of assessing road risks in rental fleets, to characterize modern technologies for reducing accidents in car sharing services, to reveal the logic of the author’s AFSA Method methodology as a behavioral risk scoring system, to determine its applied value for car fleets, insurance models and financial calculations.

Methodology and Materials. The scientific base used for the study does not sufficiently address the issue of early behavioral assessment of the driver in the first minutes of using a rented car. The scientific novelty of this study lies in the substantiation of the author’s AFSA Method methodology, which shifts the emphasis from assessing individual road events to forming the primary profile of a person in conditions of novelty, adaptation and potential stress. Within this approach, telematics data not only record the fact of sharp braking or acceleration, but are used to build a driver’s behavioral history, which can be transmitted to the server, stored in the database and used for fleet analytics, insurance models and financial calculations. To achieve the goal of the study, methods of scientific literature analysis, comparison of technological approaches, generalization of telematics risk assessment models and structural and functional analysis of the author’s AFSA Method architecture were used.

Results and Discussion. The problem of road accidents remains one of the key prerequisites for the development of new transport safety technologies. Not only the recording of the consequences of road accidents, but also the early detection of behavioral signs of risk are of particular importance. As shown by the study by S. S. B. Masud et al., the driver’s perception of risk can vary depending on the speed, distance to other vehicles, workload and individual characteristics of the driver [1]. This means that the risk of an accident is formed not only due to external road conditions, but also due to how the driver himself assesses the danger at a particular moment of movement. In this context, systems that are able not only to analyze past violations, but also to detect risk patterns directly during the trip become relevant [5; 6].

This problem is especially important for the car rental market. A rented vehicle is often used by drivers with different levels of experience, different driving habits and different abilities to adapt to a new car. The first few minutes after the start of the trip can be critical, as the driver is still adapting to the car, the road environment and traffic conditions. In the presented AFSA Method, this stage is defined as the calibration phase, during which the sharpness of accelerations and braking, micro-movements of the steering wheel, the trajectory of movement, the distance to other cars and the reaction to the road situation are analyzed. This allows you to form an initial profile of the driver’s behavior in conditions of novelty and potential stress. This approach is important for rental fleets, since the risk is associated not only with the technical condition of the car, but also with the behavior of a specific user on a specific trip [1; 9; 10; 11]. At the same time, the car rental market is undergoing a stage of active digital transformation. According to Grand View Research, in 2024, 70.1% of all bookings were made through online channels, which indicates the dominance of the digital consumption model [7]. In such conditions, the importance of mobile applications, online platforms, digital payments, remote car selection and contactless access to the vehicle is growing [7]. That is why, among the promising areas of market development, it is advisable to highlight automated fleet management, telematics systems, personalized offers, digital contracts and contactless customer service mechanisms [7; 8]. Another trend is the gradual expansion of the share of electric vehicles in rental fleets. Grand View Research associates this with the growing demand for environmentally friendly transport solutions and the desire of companies to meet the goals of sustainable development [7]. Thus, the modern car rental market is developing not only in the direction of customer convenience, but also in the direction of deeper digital analytics, where telematics and behavioral scoring can become the basis for reducing accidents and optimizing fleet management.

In view of this, the problem of road risks today is not considered only as a matter of individual driver responsibility. It is gradually moving into the realm of technological safety management, which is implemented both at the level of individual fleets and at the level of broader transport systems. For car-sharing services, adaptive driver warning systems based on risk perception assessment, personalized driver profiles, dynamic configuration of safety systems and targeted training programs are of particular importance [1; 5; 6]. Such solutions can take into account age, driving style, workload level, tendency to make sharp maneuvers and reaction to the road environment [1; 2]. The practical significance of these technologies is that they are not limited to passive data collection, but create the basis for real-time warnings, simulation learning, age-based routines and detection of risky behavior patterns. Such directions of reducing risks on the roads for car-sharing services are summarized in Fig. 1.

Fig. 1. Road risk reduction technologies for car-sharing services

In modern car-sharing services, road risk reduction technologies are formed around the idea of ​​early detection of dangerous driver behavior. It is not only about the technical equipment of the car, but also about creating a system that is able to analyze the driver’s actions, the road context and the change in risk while driving. As shown by the study by S. S. B. Masud et al., the perception of risk depends on speed, distance to other vehicles, workload and individual characteristics of the driver [1]. That is why modern solutions are gradually moving from universal warnings to personalized safety models. The main technological directions are summarized in Table 1.

Table 1

Modern technologies for reducing risks on the roads for car-sharing services

Note: compiled by the author based on sources [1; 2; 3; 5; 6; 9; 10; 11] and the author’s development of the AFSA Method

The technologies presented indicate that the car-sharing market is gradually moving towards a more complex safety management model. Its basis is not only the control of individual violations, but also a systemic understanding of driver behavior. At the same time, some innovative implementations have not yet received sufficient coverage in the scientific literature. Such solutions include the AFSA Method, which can be considered as an author’s decision-making support architecture for fleets and insurance models. Its peculiarity is that the system does not interfere with the physical driving of the car, but forms a behavioral risk profile of the driver based on telematics data. The implementation of such a methodology can be carried out by internal AI and engineering teams, external contractors in the field of telematics and data science, or through integration with existing fleet and insurance platforms. That is why it is advisable to consider the AFSA Method not as a separate technical device, but as an example of an applied risk analytics model for digital fleet safety management.

The author’s AFSA Method complements existing road risk reduction technologies by combining early detection of dangerous behavior, dynamic scoring, and subsequent fleet analytics. Its logic is that driving risk is not assessed once. It is formed during a specific trip and depends on how the driver adapts to the car, reacts to the road environment, maintains distance, accelerates, brakes, and controls the trajectory. This approach is consistent with the studies of D. A. Johnson and M. M. Trivedi, G. Castignani et al., and J. E. Meseguer Anastasio et al., in which driver behavior is determined through the analysis of sensory, telematic, and contextual data [9; 10; 11]. At the same time, the AFSA Method has an applied difference: it concentrates the assessment on the first minutes of the trip, when the risk may be higher due to the novelty of the car and the driver’s adaptive load.

For car-sharing services, this method is especially important, since the same car can be used by many drivers with different experience, driving styles and risk perception. In this case, conventional violation control is insufficient. What is needed is a system that detects risk signs before they transform into a road accident or significant operational losses. That is why it is advisable to consider the AFSA Method as a technological solution that combines the functions of ADAS, telematics scoring, behavioral feedback and fleet analytics. A summary of how individual elements of the method affect risk reduction is given in Table 2.

Table 2

Impact of the AFSA Method on reducing road risks in car-sharing services

Note: compiled by the author based on sources [1; 6; 9; 10; 11] and the author’s development of the AFSA Method

Thus, the AFSA Method solves the issue of risk reduction not through one separate tool, but through a consistent architecture of driver behavior assessment. At the first level, the system detects adaptation problems in the first minutes of the trip. At the second level, it converts telematic signals into a risk profile. At the third level, it generates feedback for the driver. Then, these data are used for fleet management, insurance analysis and building predictive models. It is this logic that allows us to move from reactive response to accidents to preventive management of road risks in the field of short-term car rental.

Conclusions. Thus, risk assessment in the field of car rental is a complex applied problem, since the risk cannot be reduced only to the fact of violating traffic rules or the technical condition of the car. It is formed through a combination of driver behavior, road environment, speed, distance, maneuvering, and a person’s ability to correctly perceive danger. That is why various risk assessment models are described in the scientific literature. Some of them are based on the driver’s perception of risk and are used to generate adaptive warnings [1; 5; 6]. Others are based on sensor and telematic data that allow recognizing aggressive driving, sudden braking, speeding, or dangerous lane changes [9; 10; 11].

At the same time, these approaches mostly assess either a separate road situation or specific events during movement. The author’s AFSA Method model has a different logic. Its peculiarity is that the system initially analyzes not only the car and not only the violations, but also the behavior of the person in conditions of novelty, adaptation, and potential stress. In the first minutes of the trip, the initial profile of the driver is formed: how he accelerates, brakes, maintains the trajectory, reacts to the road environment and maintains distance. After that, the data is transmitted to the server, where it can be stored, compared with previous trips and used for further analytics.

The practical significance of this approach is that information about the driver’s behavioral profile is no longer lost after the end of one trip. It becomes part of a database that can be used to assess recurring risk patterns. As a result, the very logic of working with the client changes: the driver is assessed not in the abstract, but on the basis of his own history of behavior in real-world conditions of vehicle operation. This creates the basis for more accurate risk-based fleet management, as well as for the development of insurance models, in particular, usage-based insurance, where not only the fact of using the vehicle, but the nature of such use is important.

References

1. Masud, S. S. B., et al. (2026). Dynamic real-time prediction of drivers’ risk perception using deep learning and driving simulator data. Transportation Research Procedia, 95, 57–64. https://doi.org/10.1016/j.trf.2009.02.004
2. Di Stasi, L. L., Álvarez-Valbuena, V., Cañas, J. J., Maldonado, A., Catena, A., Antolí, A., & Candido, A. (2009). Risk behaviour and mental workload: Multimodal assessment techniques applied to motorbike riding simulation. Transportation Research Part F: Traffic Psychology and Behaviour, 12, 361–370. https://doi.org/10.1016/j.trf.2009.02.004
3. He, X., Stapel, J., Wang, M., & Happee, R. (2022). Modelling perceived risk and trust in driving automation reacting to merging and braking vehicles. Transportation Research Part F: Traffic Psychology and Behaviour, 86, 178–195. https://doi.org/10.1016/j.trf.2022.02.016
4. Kokubun, M., Konishi, H., Higuchi, K., Kurahashi, T., Umemura, Y., & Nishi, H. (2005). Quantitative assessment of driver’s risk perception using a simulator. International Journal of Vehicle Safety, 1, 5–21. https://doi.org/10.1504/IJVS.2005.007534
5. Ping, P., Sheng, Y., Qin, W., Miyajima, C., & Takeda, K. (2018). Modeling driver risk perception on city roads using deep learning. IEEE Access, 6, 68850–68866. https://doi.org/10.1109/ACCESS.2018.2879887
6. Zhao, X., Li, Q., Xie, D., Bi, J., Lu, R., & Li, C. (2018). Risk perception and the warning strategy based on microscopic driving state. Accident Analysis & Prevention, 118, 154–165. https://doi.org/10.1016/j.aap.2018.02.010
7. Grand View Research. (2026). U.S. car rental market size & share | Industry report, 2030. Grand View Research. URL: https://www.grandviewresearch.com/industry-analysis/us-car-rental-market-report
8. IMARC Group. (2026). U.S. car rental market size, share, trends report 2034. IMARC Group. URL: https://www.imarcgroup.com/united-states-car-rental-market
9. Johnson, D. A., & Trivedi, M. M. (2011). Driving style recognition using a smartphone as a sensor platform. IEEE Transactions on Intelligent Transportation Systems, 4, 1609–1615. URL: https://www.academia.edu/4388197/Driving_Style_Recognition_Using_a_Smartphone_as_a_Sensor_Platform
10. Castignani, G., Derrmann, T., Frank, R., & Engel, T. (2015). Driver behavior profiling using smartphones: A low-cost platform for driver monitoring. IEEE Intelligent Transportation Systems Magazine, 1, 91–102. URL: https://www.researchgate.net/publication/273291904_Driver_Behavior_Profiling_Using_Smartphones_A_Low-Cost_Platform_for_Driver_Monitoring
11. Meseguer Anastasio, J. E., Tavares de Araujo Cesariny Calafate, C. M., Cano Escribá, J. C., & Manzoni, P. (2013). Drivingstyles: A smartphone application to assess driver behavior. IEEE. https://doi.org/10.1109/ISCC.2013.6755001
12. Fedyk, M., Kudyn, S., & Bondarchuk, M. (2022). Digitalization of the world business environment under the conditions of global instability. Economics & Education, 7(1), 34-41. URL: http://www.baltijapublishing.lv/index.php/econedu/article/view/1697/1716