Data-Driven Approaches for Efficient Vehicle Driving Analysis: A Survey

Abstract

Efficient vehicle driving generally intends to reduce fuel consumption, emissions of harmful substances, and accident rates based on energy-efficient driving patterns as a set of parameters defining optimal vehicle and route characteristics, together with specific ways of driving a vehicle that the particular driver applies. To gain environmental friendliness in driving, two main approaches can be outlined: optimal route planning and driver training based on the principles of ecological driving. The latter can be supported by using software for real-time, efficient vehicle driving recommendations. In order to develop the principles of ecological driving as well as generate relevant real-time recommendations, it is necessary to identify the specific parameters required to analyze driver behavior and vehicle performance, determine the corresponding energy consumption, and understand the influence of route and environmental conditions on overall efficient vehicle driving. These tasks require a large amount of data, often obtained from heterogeneous sources, which, when publicly available, are complex for consolidation, transmission, and processing, not to mention the complexity of the data model itself. This study provides a thorough review of the current data sources and techniques for efficient vehicle driving analysis, focusing on the availability and relevance of dataset sources and repositories. The categorization of parameters and data processing techniques enabling efficient vehicle driving analysis is carried out according to efficiency types such as driver’s efficiency, resource consumption efficiency, and route planning efficiency. For each type of efficiency, we provide a list of contextual groups and features, identifying the dataset containing the necessary feature, making it possible not only to determine the parameters defining, for example, driver efficiency, but also locate the corresponding dataset serving as a stepping stone for researchers and practitioners to join the community investigating efficient vehicle driving analysis. We also discuss future trends and perspectives, identifying alternative data sources for efficient vehicle driving analysis, and focus on data collection issues revealed by the practical use case of collecting data from mobile phone sensors.

1. Introduction

Today, the worldwide community is characterized by growing economic, social, and energy needs, the provision of which has led to increased emissions, the greenhouse effect, and climate change in general. Air pollution remains one of the greatest threats to human health and the environment. Thus, according to the State of Global Air Report 2024 (https://www.stateofglobalair.org/resources/report/state-global-air-report-2024) (accessed on 22 April 2025), air pollution is the second-highest global risk factor for death, and the transport sector (https://ourworldindata.org/air-pollution-sources) (accessed on 22 April 2025) is one of the main sources of pollution. As a result, vehicle emission modeling and specific transportation problems are being investigated with environmental considerations [1]. Therefore, given the need to reduce the impact of the transport sector on the environment, the problem of energy-efficient driving is becoming more relevant today.

One of the obvious solutions to this problem is the use of electric vehicles. Unfortunately, the transition process to a more ecological type of vehicle can take a significant time, and it is also necessary to consider the impact of energy production and the recycling of electric vehicle batteries.

For their part, auto manufacturers are currently implementing various technical vehicle improvements to reduce the negative impact of road transport. Hybrid cars came to the market together with more energy-efficient fuel and parts, and aerodynamic, engine, and transmission characteristics are being improved. However, despite such technical improvements, in addition to the vehicle, the driver’s behavior also has a significant impact on fuel consumption and emissions.

Therefore, another approach to solving the problem of the negative impact of transport on the environment is to introduce the concept of energy-efficient driving directly into the vehicle driving process, namely the approaches of “energy-efficient driving” or “eco-driving”.

Efficient driving in this context is considered as a driving style that reduces fuel consumption, harmful substances emissions, and accident rates, and the energy-efficient driving pattern itself is considered as a set of parameters defining optimal vehicle and route characteristics to reduce energy consumption, as well as specific ways of driving a vehicle that the particular driver applies. If applied, the methods of the mentioned concept can help the driver increase energy efficiency and reduce fuel consumption by 5–25% [2].

Currently, there are two main approaches to increasing the environmental friendliness of driving: optimal route planning and driver training based on the principles of ecological driving. Optimal route planning addresses the problem of creating a route with the lowest fuel consumption. This is achieved by creating a route through roads with less congestion, fewer stops at traffic lights, and optimal speed. This method requires a comprehensive analysis and study of the traffic situation in a particular city, as well as solving the problem of forecasting the road situation. The driver’s training strategy focuses on increasing the energy efficiency of each particular driver. In addition, the implementation of this approach can involve using information and communication technologies to improve the driver’s activity efficiency in real-time by providing efficient driving recommendations.

Specific software deployed on mobile devices can be used to provide real-time driving recommendations. These real-time recommendations can have a significant impact on efficient vehicle driving. Research shows that following ecological driving recommendations can reduce fuel consumption by 4.5–13% [3].

The aforementioned software for eco-driving recommendations requires a large amount of data, often obtained from heterogeneous sources, which, when publicly available, are complex for consolidation, transmission, and processing, not to mention the complexity of the data model itself. In addition, driver behavior cognition to generate appropriate recommendations also requires large amounts of data, both for model training in the case of machine learning methods and for applying the software by drivers in real-time.

Therefore, identifying high-quality data volumes and studies focused on improving efficient vehicle driving analysis is a clear need, mainly to support research and industrial communities in taking the next steps. With this motivation, this paper provides a comprehensive survey of datasets and data processing techniques that, in conjunction, form data-driven approaches oriented to efficient vehicle driving analysis based on machine learning and optimization techniques, which are currently state of the art.

This review aims to emphasize data availability, catalog known data repositories, identify data generation trends, and explore new directions and unexplored issues of data collection. Furthermore, the survey aims to identify specific parameters needed to analyze driver behavior and vehicle performance, as well as to determine the corresponding energy consumption and the influence of route and environmental conditions on overall efficient vehicle driving. The main contributions of this article are listed below.

• We provide a detailed review of current data sources and techniques for efficient vehicle driving analysis, focusing on the availability and relevance of dataset sources and repositories.
• We systematically review the parameters that define efficient vehicle driving by grouping them according to the driver’s efficiency, resource consumption, and route planning efficiency, with further categorization of relevant features. For each type of efficiency, we provide a list of contextual groups and features, identifying the dataset containing the necessary feature if freely available.
• We discuss future trends and perspectives, identifying alternative data sources for efficient vehicle driving analysis, together with perspectives for the development of transportation systems. We also focus on data collection issues revealed by the practical use case on road damage detection while collecting data from mobile phone sensors.

The remainder of this survey is organized as follows. Section 2 describes the methodologies employed for collecting, filtering, and organizing the references considered in this review. Section 3 is centered around the datasets and features for efficient driving analysis, while Section 4 is devoted to data processing techniques enabling efficient vehicle driving analysis. In Section 5, future trends and perspectives are identified, proposing new working lines. As a practical note, use cases are presented in Section 6, where driving efficiency data parameters are collected and processed. Finally, Section 7 contains the main conclusions and future tasks. The paper also includes six appendices containing information on the features required to assess drivers, vehicles, and route efficiency, as well as methods of analysis.

2. Methodology

In order to carry out the related literature review systematically, the methodology proposed by [4] was followed. Thus, we set out a series of initial research questions (RQs) to focus on the relevant issues. The research questions formulated in this study are as follows:

RQ1:

Which input data are relevant for efficient vehicle driving analysis?

RQ2:

Are there freely available datasets?

RQ3:

What techniques are being applied to efficient vehicle driving analysis?

RQ4:

What are the alternative data sources for efficient vehicle driving analysis?

RQ5:

What are the issues concerning data collection?

Then, a series of primary and secondary keywords were identified and applied to search and filter related articles to be thoroughly examined in a second phase. The systematic search was conducted on the leading reference scientific platforms in the target area of this survey for the period 2001–2025:

IEEE Xplore (https://ieeexplore.ieee.org/) (accessed on 22 April 2025),

Scopus (https://www.scopus.com/) (accessed on 22 April 2025),

Lens (https://www.lens.org/) (accessed on 22 April 2025),

ACM Digital Library (https://dl.acm.org/) (accessed on 22 April 2025),

MDPI (https://www.mdpi.com/) (accessed on 22 April 2025).

A flow chart of the systematic review protocol is shown in Figure 1.

We applied five queries and one filter to the above-mentioned search engines. The main keywords were “energy-efficient driving”. Secondary keywords were chosen as follows: “transport”, “fuel consumption”, “driver behavior”, and “range estimation”, filtered by the subject area of “computer science and/or software” (depending on the platform).

Figure 1 shows the results obtained in each search engine according to the keywords. Thus, for example, after applying the filter “computer science and/or software”, we obtained a minimum of 252 (MDPI) and a maximum of 1498 (Scopus) search results.

After each query, the abstract of the article and keywords were screened. To do so, we pointed out three contextual groups of keywords: driver behavior, energy-efficient consumption and emissions, and processing techniques.

The most common keywords in the contextual group of driver behavior were driving style and driving strategy. In the group for energy consumption and emissions, the keywords were eco-driving, energy consumption, energy efficiency, and emissions. In the processing techniques group, the keywords were machine learning, optimization, neural networks, and data mining.

After a preliminary analysis of the keywords, a full-text review was conducted. At this point, considering the research questions of the study, the review was focused on the main topic of the paper (driver, resources, and route), driving task (perceiving the environment, vehicle control, and route planning), features (the exact parameters used to reach the results of the paper), the source of data (sensors, simulation, and external sources), reference to datasets (with an emphasis on the availability of datasets), processing methods, and general processing technology (neural networks, optimization, and statistical analysis).

Finally, the survey includes material from 135 articles, which were grouped by driver efficiency (60 articles), resource consumption efficiency (63 articles), and energy-efficient route planning (12 articles). The distribution of articles by chronology is shown in Figure 2.

Once an exhaustive search was conducted and relevant works were examined, a deeper review was carried out to answer the initial research questions. In this sense, in order to approach RQ1 and RQ2, we analyzed whether the datasets included the drivers’ efficiency, resource consumption (electricity and fuel), and energy-efficient route planning, as detailed in Section 3. The format, attributes, structures, and sources of data collection were also explored. In addition, special attention was paid to detecting if free access to datasets was provided.

Similarly, an overview of the data processing methods that allow analyzing efficient vehicle driving in the context of driver efficiency, resource consumption efficiency, and energy-efficient route planning is provided in Section 4 (to answer RQ3).

Future trends and perspectives in emerging new data sources for the analysis of efficient vehicle driving and future application areas are discussed in Section 5 (to answer RQ4). Data sources such as “serious games” and “metaverse” are considered. Section 6 contains practical notes on the data collection process for road damage detection and the construction of energy-efficient routes for electric vehicles responding to RQ5.

As a summary, the general structure of the article is shown in Figure 3, where each research question is mapped to the section where it is approached.

3. Datasets for Efficient Vehicle Driving Analysis

Efficient driving refers to a driving style that reduces fuel consumption and emissions to the environment and increases the safety of road users, ultimately providing social, economic, and environmental benefits. To evaluate efficiency as a driving property, it is necessary to consider the interaction of three driving elements: the driver, the vehicle, and the environment. Efficient vehicle driving, in this case, describes the level of performance of the driving process that uses the minimum amount of resources (fuel, time, etc.) to achieve the best result (longer distance, amount of cargo, etc.) and consists of driver’s efficiency, resource consumption, and route planning efficiency. This section is devoted to reviewing the references in the context of these three driving aspects and considering the availability of relevant datasets.

3.1. Dataset Availability

The analysis of the availability of the datasets basically shows that they are not accessible in 72% of the selected articles. Only 15% of the articles contain links to the datasets supporting the research presented. In 12% of the articles, the authors indicated that data are available upon request. In 4% of the articles, the authors claimed that the data are included in the article.

During the literature review, 21 datasets were examined (

Table 1. Available dataset descriptions according to the provider, format, main attributes, and source of collection.

Refs.	Provider	Format	Attributes	Collection Method	Link
[5]	U.S. Department of Transportation, National Household Travel Survey	CSV, DBF, SAS V2.1, SPSS V2.1	Vehicle parameters, traffic parameters, charging parameters, trip parameters, and driver parameters	Statistical	https://cutt.ly/ewQehHxn (accessed on 10 March 2025)
[6]	SisInfLab Research Group, Polytechnic University of Bari	CSV	Speed parameters, acceleration parameters, engine parameters, and fuel consumption	OBD-II Phone sensors	https://cutt.ly/3wQejm8h (accessed on 10 March 2025)
[7,8,9]	Microsoft Research T-Drive Project	TXT	GPS parameters	GPS sensor	https://cutt.ly/AwQrf2FI (accessed on 10 March 2025)
[10]	Collected by the authors	XLSX, Jupter Notebook	Electric vehicles’ specifications	Web-scraping and text-mining	https://cutt.ly/TwQrheFl (accessed on 10 March 2025)
[11,12,13]	U.S. Department of Transportation, Federal Highway Administration, Next Generation SIMulation Program	CSV, XML, PDF, Video, API, JSON, and TXT	Vehicle trajectory data	Network of synchronized digital video cameras	https://cutt.ly/YwQrj3xo (accessed on 10 March 2025)
[14]	comma.ai	Video and Numpy arrays	Route parameters, speed parameters, acceleration parameters, and GPS parameters	Comma EONs, road-facing camera, 9-axis IMU, CAN bus, and GNSS	https://cutt.ly/LwQrWNhW (accessed on 10 March 2025)
[15]	Teledyne FLIR LLC	TIFF, JPEG, JSON, and Video	Labeled categories: person, bike, car, motorcycle, bus, train, truck, traffic light, fire hydrant, street sign, dog, skateboard, stroller, scooter, and other	Thermal camera and visible camera	https://cutt.ly/zwQrECoW (accessed on 10 March 2025)
[16]	Department of Energy, Office of Energy Efficiency and Renewable Energy	CSV and DBF	Traffic parameters, vehicle parameters driver parameters, GPS parameters, fuel consumption, and weather	Telematics kit	https://cutt.ly/hwQrRkEU (accessed on 10 March 2025)
[17]	U.S. Department of Transportation, Intelligent Transportation Systems	CSV and JSON	Trip parameters, vehicle parameters, GPS parameters, and weather	Board vehicle devices and roadside units	https://cutt.ly/uwQrTGoX (accessed on 10 March 2025)
	University of Michigan	XLSX	Vehicle parameters, GPS parameters, and fuel parameters	OBD-II	https://cutt.ly/CwQrT8Gz (accessed on 10 March 2025)
[18]	Autonomous Vision Group	JPEG	Topics: stereo, flow, sceneflow, depth, odometry, object, tracking, road, semantics, and raw data	Visual (Stereo) camera, LiDAR, and GNSS	https://cutt.ly/DwQrOeyL (accessed on 10 March 2025)
	Motional	JSON and JPEG	Vehicle parameters and map extension data	LiDAR and CAN bus	https://cutt.ly/DwQrPejh (accessed on 10 March 2025)
	Waymo LLC	Vector maps, JPEG, and TFRecord	Perception dataset and motion dataset	LiDAR	https://cutt.ly/kwQrP65R (accessed on 10 March 2025)
	Audi	JSON and JPEG	Labeled categories: vehicles, pedestrians, road infrastructure, nature, sky, buildings, and rain	Visual camera, LiDAR, and CAN bus	https://cutt.ly/swQrA5Es (accessed on 10 March 2025)
[19]	University of Cambridge	JPEG and Video	Labeled categories: moving objects, road, ceiling, and fixed objects	Camera	https://cutt.ly/1wQrSMAl (accessed on 10 March 2025)
	Daimler AG, TU Darmstadt, MPI Informatics, 4TU Dresden	JPEG, Video	Labeled groups: flat, human, vehicle, construction, object, nature, and sky	Camera, CAN bus, and GNSS	https://cutt.ly/BwQrDWPU (accessed on 10 March 2025)
[20]	South Carolina Department of Transportation	GIS data, CSV, PDF, and TXT	Traffic parameters and road parameters	Traffic cameras	https://cutt.ly/OwQrGDLg (accessed on 10 March 2025)
[21]	University of Sydney, Australian Centre for Field Robotics	CSV	Vehicle parameters and trip parameters	OBD-II and GNSS	https://cutt.ly/GwQrH3vJ (accessed on 10 March 2025)
[22]	Kia motors corporation, South Korea	CSV	Engine features, fuel features, and transmission features	OBD-II	https://cutt.ly/PwQrJfUV (accessed on 10 March 2025)
[23]	European Commission, Joint Research Centre	TIFF and OVR	Trip parameters and map data	Satellite	https://cutt.ly/DwQtFx9t (accessed on 10 March 2025)
[24]	Technical University of Denmark	CSV	Vehicle sensor data, road condition parameters, and driving data	AutoPi telematics unit, and CAN bus	https://cutt.ly/4wTE9zt6 (accessed on 10 March 2025)

). Among dataset providers are government agencies, academic institutions, and private companies, as well as the authors of selected articles. Depending on the type of data in the dataset, the formats can be CSV, JPEG, JSON, TXT, or video. Various external sensors (GPS, cameras, LiDAR, mobile phone sensors, and roadside units) and vehicle built-in diagnostic systems (OBD-II) are most often used as means of data collection. The studied datasets contain many parameters necessary for efficient vehicle driving analysis, including vehicle, traffic and battery charge parameters, fuel consumption, driver characteristics, and weather conditions. The selection of an appropriate dataset depends on the subject area and the purpose of the analysis. The results of the research on the relevant parameters in the context of the efficiency type and availability are presented in the following subsections.

3.2. Driver’s Efficiency Analysis Features

The task of evaluating the driver’s efficiency is quite complex, since a large number of different types and sources of data are used for calculations. Thus, according to [25], the driving time when changing lanes, lateral deviation, speed, and the vehicle turning angle are used to analyze and model the driver’s behavior. The study carried out in [2] uses vehicle characteristics (engine, tires, transmission, and route), traffic parameters (speed, roundabouts, and zones), and driving behavior (speed, limits, and prediction of driver behavior) to determine driving style and traffic conditions. The obtained features are then used to determine the levels of emissions and fuel consumption.

Evaluating efficient driving patterns can also be conducted based on speed, fuel consumption, acceleration, changes in the number of engine revolutions, and the frequency of the oscillation of engine revolutions [26]. In addition, authors of [27] proposed an approach to modeling longitudinal driving behavior and the complexity of driving tasks. Road topology, weather conditions, traffic intensity indicators, and environmental parameters are used. According to [28], speed, acceleration, engine revolutions per minute, GPS data, and route data are used to evaluate driver behavior. Authors of [29] used turn length and radius, slope, and speed limit to detect driving strategies. Authors of [30] evaluated the impact of driver characteristics on emissions and fuel consumption, where driver characteristics, vehicle characteristics, traffic conditions, and GPS data are used as parameters.

The authors of [31] described the two different driver behaviors, namely “normal driving” and “parking search”, defining that factors such as driver age, vehicle class, familiarity with the destination, and rush hour timing influence the search process and the driving style.

In [9], driver behavior was analyzed based only on GPS data. Authors of [32] proposed an approach to modeling energy-efficient driving evaluation based on speed, acceleration, duration of driving modes, and fuel consumption. Authors of [33] used environmental and weather conditions (wind speed and rolling resistance), involving many vehicles and traffic conditions, to estimate the optimal speed profile with the minimum energy and time consumption. In the same year, ref. [34] proposed an approach for profiling and predicting drivers’ behavior based on analyzing acceleration and braking dynamics, road topology, and turns.

Later, ref. [35] proposed an analysis focused on the classification of driving style according to GPS parameters, OBD system data, speed, and acceleration. Similarly, ref. [36] proposed to analyze the driver’s behavior comprehensively, considering the driver and driver–vehicle–environment models. In this paper, vehicle parameters, physiological and psychological parameters of the driver, behavior parameters, and subjective parameters are used for calculations. Authors of [37] proposed an approach to assessing and analyzing drivers’ behavior in the context of energy-efficient driving. The following parameters are used for the analysis: measurement time, average speed, fuel consumption in measuring, average fuel consumption in liters per 100 km, average CO₂ emissions, average position of the accelerator pedal, maximum position of the accelerator pedal, vehicle operation time without using the gas pedal (rolling), time of using the brake pedal, total distance without using the accelerator pedal, total distance using the brake pedal, braking frequency, stopping frequency, stopping time, gear shifting frequency, the total number of engine revolutions, and the average engine speed. From a different point of view, ref. [8] proposed to detect the abnormality in a driver’s manoeuveres (rapid acceleration, sudden braking, and turning) based on the force of pressing the brake, the fluctuation in vehicle speed, the angles of the wheels, and the movement of the steering wheel. In that study, an approach for analyzing speed-related driver actions and direction-related driver actions based on GPS data was proposed. Authors of [38] suggested the analysis of the driver’s misbehavior at roundabouts from the point of view of emissions, uneven vehicle movement, and conflict situations. Data required for analysis include vehicle parameters, a basic traffic model, speed profiles, and roundabout specifications. When considering the issue of adaptive classification of driving styles, ref. [39] used longitudinal and latitudinal acceleration, speed, direction, vehicle parameters, and driver characteristics (dynamic, average, and calm).

The literature review shows that the driver efficiency analysis requires vehicle technical characteristics, vehicle dynamics, driver, infrastructure, traffic, and environment parameters. These features were collected and organized in a series of tables in Appendix A in order to alleviate the contents of the paper so that they can be referred to from them. In this sense, the main aspect of vehicle technical characteristics (

Table A1. Driver efficiency: technical vehicle characteristics.

Contextual Group	Features	References	Available in Refs.
battery parameters	-	[79,80,81]	-
vehicle description parameters	centroid height, drivetrain, gross mass, motor maximum/ rated speed, motor parameters, vehicle body parameters, vehicle data, vehicle length, vehicle parameters, vehicle type, wheel rolling radius, wheel rotational inertia, wheelbase (front, rear), windward area	[2,13,30,38,39,79,81,82,83,84,85,86,87,88,89]	Direct links: - Upon request: [88]

) is the use of vehicle parameters. As seen in

Table A2. Driver efficiency: vehicle dynamics characteristics.

Contextual Group	Features	References	Available in Refs.
acceleration	absolute throttle position, accelerator pedal value, average deceleration, lateral acceleration, throttle position, throttle signal, vertical acceleration	[6,11,22,26,28,32,34,35,37,55,79,80,88,90,91,92,93,94,95,96]	Direct links: [6,11], Upon request: [88,96]
air compressor	activation of air compressor, air compressor ON–OFF	[22,55]	Direct links: [22] Upon request: -
air parameters	intake air pressure, intake air temperature, manifold air pressure, mass air flow	[6,22,55]	Direct links: [6,22] Upon request: -
emissions	-	[37,59,94]	-
engine parameters	average rotational engine speed, engine coolant temperatures, engine fuel cut off, engine idle target speed, engine load, engine rotations, engine soaking time, engine speed, engine torque, instant engine, maximum indicated engine torque, minimum indicated engine torque, total number of engine revolutions	[6,22,26,37,55,96]	Direct links: [6,22] Upon request: [96]
fuel (power) parameters	fuel consumption, fuel pressure, fuel trim, instantaneous power consumption, long-term fuel trim bank, mean fuel consumption in liters per 100 km, power consumption, remaining fuel	[6,22,26,32,37,55,59,92,94,97]	Direct links: [6,22], Upon request: -
interaction	lead vehicle characteristics, preceding vehicle on the same lane, rear vehicle on the same lane, time-to-collision	[13,83,85]	-
headway	average space headway, average time headway, headway time	[25,88]	Direct links: - Upon request: [88]
location	altitude, current lane position of the vehicle, GPS, lateral coordinate of the front center of the vehicle, lateral deviation, latitude, longitude, longitudinal coordinate of the front center of the vehicle, position in the lane, position on the road, street name or highway name	[6,8,9,13,25,26,28,30,35,59,80,90,93,97,98]	Direct links: [6,8,9,13] Upon request: -
performance parameters	brake, clutch operation acknowledge, converter clutch, current gear, current spark timing, driving mode duration, gear selection, gas pedal, images, measurement time, measuring length, OBD system data of a hybrid vehicle, percentage of time in different speed intervals, repeatability of braking/gear shifting/gear upshifting/ stopping, revolutions per minute, steering angle of car, steering wheel angle, steering wheel speed, time of a stop, time of the brake pedal use, time stamps, total distance with using the brake pedal, total distance without using accelerator pedal, travel distances, vehicle time in rolling	[6,13,19,25,28,32,35,37,55,57,59,75,80,89,99,100,101]	Direct links: [6,13,19] Upon request: [75]
physical parameters	air drag coefficient, friction torque, flywheel torque, reducer ratio	[22,55,84]	Direct links: [22] Upon request: -
road parameters	calculated road gradient, highway, inclination	[29,55,90]	-
speed	average velocity, maximum velocity, safety gap speed, speed profiles, speeding, wheel velocity front left-hand, wheel velocity front right-hand, wheel velocity rear left-hand, wheel velocity rear right-hand	[6,11,22,25,26,28,32,35,37,38,55,59,75,80,88,90,92,93,96,97,98,101,102]	Direct links: [6,11,22] Upon request: [75,88,96]
vehicle parameters	calculated load value, historical trajectory, length, torque converter speed, transmission oil temperature, vehicle trajectory, vehicle-based measures, vehicle’s surroundings	[18,22,29,36,55,88,98,102,103,104,105]	Direct links: [18,22] Upon request: [88]

, the most common features for vehicle dynamics characteristics in the analyzed articles are acceleration and speed. The most used feature for the driver’s characteristics is the driver’s behavior (

Table A3. Driver efficiency: driver, infrastructure, and environmental characteristics.

Contextual Group	Features	References	Available in Refs.
Driver’s characteristics
driver characteristics	gender, age	[80,101]	-
driver’s behavior	-	[2,30,36,39,87,97,106]	Direct links: - Upon request: [106]
distraction/fatigue parameters	eye movement, percentages of eye closure, percentages of head nodding, percentages of scaling, percentages of yawning	[76,106,107,108]	Direct links: - Upon request: [106,107]
Infrastructure characteristics
road parameters	intersection features, flat road sections, lane identification number, long uphill road sections, number of lanes, road data, road design, road slope, road structure, road types, rolling resistance coefficient, roundabout specifications, speed limit, structure of the intersection, topographic information, track width (front, rear), trip average road grade, turns, turning radius, type of lines, type of road, up–down road slopes, velocity limit	[27,29,34,38,79,80,83,84,86,88,90,93,95,97,99]	Direct links: - Upon request: [88]
infrastructure parameters	recharging points on the path, signal change information	[82,83,97]	-
V2X communication	vehicle-to-infrastructure (V2I), vehicle-to-vehicle (V2V)	[48,83]	-
Traffic characteristics
distance parameters	distance, distance between all drivers	[75,102]	Direct links: - Upon request: [75]
route data	route deviation	[28,75]	Direct links: - Upon request: [75]
traffic data	baseline traffic model, road congestion, traffic measures	[2,27,38,96,97,99,106]	Direct links: - Upon request: [96,106]
Environmental characteristics
environmental data	driving conditions, local wind speed, temperature, weather	[27,30,87,96,99,102]	Direct links: - Upon request: [96]
external driver’s environment	-	[109]	-

). Table A3 also summarises different features related to infrastructure and environment characteristics.

3.3. Resource Consumption Efficiency Analysis Features

Regarding resource consumption efficiency, in a study developed by [40], the issue of fuel efficiency was considered based on coasting. Such parameters as vehicle mass, wheel radius, moment of inertia of the engine, torque converter, transmission, wheel moment of inertia, coefficient of aerodynamic force, coefficient of rolling resistance, working volume of the engine, and gravitational acceleration were stored and used for the calculations.

Authors of [41] considered the issue of reducing fuel consumption and emissions using intelligent transportation systems. An energy-efficient navigation system was proposed where the input data were vehicle parameters, digital maps, and GPS data.

Authors of [42] proposed a recommended energy consumption management system for electric vehicles. For this purpose, the system used data such as route information, current information about vehicle movement, location, driving behavior, and road topology. In this line, ref. [43] determined predictions for an energy consumption management strategy for electric vehicles based on the movement of the front vehicle. The authors proposed vehicle and battery models based on sensor data and vehicle technical characteristics. This model was trained using parameters such as speed, acceleration, relative distance to the preceding vehicle, speed, and acceleration for the current vehicle.

Authors of [44] considered the issue of the optimal management of energy consumption. Parameters such as the discrete gear ratio, vehicle trajectory limitations of infrastructure (road signs) and traffic, nonlinear vehicle dynamics, and GPS and GIS data were used for the relevant calculations.

According to [45], the following parameters are used to organize energy-efficient control of electric vehicles in real-time: V2V wireless communication data, GPS and GIS data, road profiles, predefined restrictions (arrival time and maximum and minimum speed restrictions), information about future traffic flows, speed forecasting, information about the current and preceding vehicle, information about the traffic light timing, technical characteristics of the electric vehicle, and data about the driver’s behavior.

In this sense, refs. [43,45] suggested that, in order to predict the movement of a vehicle and gain real-time control for reduction driving energy consumption, it is essential to consider data from vehicle-to-vehicle (V2V) wireless communication, GPS and GIS systems, road terrain profiles, predefined limitation factors (arrival time and speed limits), future traffic flow information, velocity prediction, current vehicle information, traffic light timing information, full fuel vehicle models, and human driving data.

The conducted literature review shows that resource consumption efficiency analysis requires a series of characteristics, which have been thoroughly gathered in Appendix B with their available datasets. In particular, vehicle technical characteristics (

Table A4. Resource consumption efficiency: technical vehicle characteristics.

Contextual Group	Features	References	Available in Refs.
battery parameters	capacity, charge, cooling mechanism, current, diameter of the cylinder, initial capacity, input currents, number of cells and modules, real-time status, state of charge, thermal and health model, traction battery current and voltage, type, voltage	[5,10,14,54,56,110,111,112,113,114,115,116,117,118,119]	Direct links: [5,10,14], Upon request: [116]
emissions	-	[120,121,122]	-
energy consumption	-	[110,123,124]	-
engine parameters	capacity, efficiency, electric motor parameters, model, number of cylinders, power, rated capacity, speed, torque	[52,110,113,115,117,125,126,127]	Direct links: - Upon request: [125]
fuel parameters	density, lower heating value	[117]	-
gearbox parameters	power loss, efficiency index, gear ratio	[44,115,117]	-
physical parameters	drag coefficient, required force and torque at the wheels, rolling friction coefficient, rolling resistance force	[24,113,117,128]	Direct links: [24] Upon request: -
vehicle description parameters	load, mass, model, structure, wheel radius, equipped sensors/radars, total mileage, curb weight, vehicle data, frontal area, transmission parameters, powertrain, electric vehicle body shape, electric vehicle year, average vehicle length, full electric vehicle model	[5,10,17,24,40,41,45,47,52,73,111,113,115,117,124,126,127,128,129,130,131,132,133,134,135,136,137,138,139]	Direct links: [5,10,17,24] Upon request: [136,139]

), vehicle dynamics characteristics (

Table A5. Resource consumption efficiency: vehicle dynamics characteristics.

Contextual Group	Features	References	Available in Refs.
acceleration	-	[7,10,24,52,110,119,120,123,125,128,134,140,141,142]	Direct links: [7,10,24] Upon request: [125]
engine parameters	load rate, revolution, speed, stroke, torque, revolutions per minute	[52,116,117,125,143]	Direct links: - Upon request: [116,125]
fuel consumption	power consumption	[14,24,52,120,143,144]	Direct links: [14,24] Upon request: -
location	GPS, azimuth, time-stamped location, longitude, latitude	[7,24,41,42,44,45,111,120,123,125,132,140,145]	Direct links: [7,24] Upon request: [125,145]
performance parameters	brake usage, deceleration, driving data, frequency of stopping, idling, OBD data, percentage of traction time, state of air conditioning, stop signs, throttle position sensor, turn off time, turn on time, vehicle motion, vehicle telemetry data, vehicle state	[16,42,47,51,58,111,113,117,120,125,141,143,145]	Direct links: [16] Upon request: [58,125,145]
physical parameters	aerodynamic force, attractive force, braking torque, final drive ratio, gross weight, nonlinear dynamics of the vehicle, power, torque, traction power, vehicle description parameters, wheel torque	[24,44,115,116,117,128]	Direct links: [24] Upon request: [116]
interaction	autonomous vehicles, current preceding vehicle information, data from connected and automated vehicles, average time gap (headway) between vehicles, driving distance, position and velocity of the host vehicle, position and velocity of the preceding vehicle, vehicles in a system, vehicles not registered as an autonomous vehicle	[45,51,52,131,134,146],	-
speed	vehicle speed, speed of a link, speed of a rarefaction wave, spot speed, moving speed, speed–time series, velocity prediction	[7,14,24,45,51,52,110,111,113,116,119,120,123,125,134,140,141,144]	Direct links: [7,14,24] Upon request: [116]
traffic rules	advisory speed limit	[134]	-

), trip characteristics (

Table A6. Resource consumption efficiency: trip and traffic characteristics.

Contextual Group	Features	References	Available in Refs.
Trip characteristics
current vehicle driving information	real-time data of travel status	[42,54]	-
elevation and street-level maps	-	[16]	Direct links: [16] Upon request: -
historical data	including GPS data	[74,131]	-
route parameters	arrival time, distance, route, trip time, vehicle trajectory	[42,44,45,56,117,121,142]	-
Traffic characteristics
density parameters	critical density, density of a link, downstream traffic density, jam density, upper stream traffic density	[134]	-
number of vehicles	multitude of vehicle	[33,140]	-
traffic data	future traffic flow information, minimum vehicle headway, real-time traffic conditions, traffic information, upstream loop detector location	[5,16,33,45,58,114,129,131,133,134,137,147]	Direct links: [5,16] Upon request: [58,147]
traffic laws	-	[47]	-
vehicle information	-	[137]	-

), driver’s characteristics (

Table A7. Resource consumption efficiency: driver, infrastructure, and environmental characteristics.

Contextual Group	Features	References	Available in Refs.
Driver’s characteristics
driver characteristics	gender, age, occupation, individual-specific characteristics, observation-specific characteristics	[42,113,130,133]	-
human driving data	-	[45]	-
travel behavior	-	[130]	-
Infrastructure characteristics
GIS data	-	[44,45]	-
road parameters	conditions information, control speed at the intersection, force related to road steepness, intersection location, length of road, max and min speed limits, road data, road geometry, road network topology, road roughness, road slope, road speed limit, road terrain profile, roadways, slope angle, speed limits, top speed	[10,17,24,42,45,47,52,113,116,128,129,132,133,134,140,142,146,147]	Direct links: [10,17,24] Upon request: [116,147]
safety concerns	-	[47]	-
traffic lights parameters	green time interval in signal cycle, signal cycle, signal information, signal phase, timing information from the roadside equipment unit, timing information, yellow time interval in signal cycle	[45,47,131,132,134]	-
V2X communication	vehicle-to-infrastructure (V2I), vehicle-to-vehicle (V2V)	[43,45,131,132]	-
Environmental characteristics
weather data	altitude, ambient air density, date-specific environmental attributes, environmental variations, operation conditions, rolling resistance, wind speed	[16,17,33,47,52,58,113,117,130,133,147]	Direct links: [16,17] Upon request: [58,147]

), infrastructure characteristics (Table A7), traffic characteristics (Table A6), and environmental characteristics (Table A7) are considered.

3.4. Route Efficiency Analysis Features

Route planning efficiency also contributes to overall vehicle driving efficiency. To determine energy-efficient routes and range estimation models, according to [46], it is necessary to consider historical data of average speed, accelerations and road gradient, future data of driving speed, related parameters to the vehicle, road, traffic, weather, and driver profiles, dynamic traffic data, and battery models (applied to electric vehicles, EV).

Regarding time and energy-efficient routes, efficient speed profiles, detailed vehicle models, and routing algorithms for time and/or energy-saving concerning EV, ref. [21] suggest using historical driving data, an electric vehicle model, max and min battery capacity, vehicle parameters, road conditions, grid-to-vehicle, vehicle-to-grid services, speed variations, force models, battery model, and driving and vehicle attributes (vehicle mass and cargo, temperature, time of travel, traffic level, and propulsion motor power).

The energy-efficient trajectory can be defined by physical traits (powertrain, environmental variations, traffic laws, and safety concerns), road slopes, speed limits, safety distance, stop signs, and traffic lights [47].

As previously shown, the literature review (with references in Appendix C) shows that route planning efficiency analysis requires vehicle technical characteristics (

Table A8. Route efficiency: technical vehicle and dynamics characteristics.

Contextual Group	Features	References	Available in Refs.
Vehicle technical characteristics
battery parameters	battery model, max and min battery capacity, state of charge	[20,21,46,148]	Direct links: [20,21] Upon request: -
emissions	fuel and CO2 emissions cost per liter	[149]	-
physical parameters	coefficient of aerodynamic drag, coefficient of rolling resistance, conversion factor, curb weight, force models, fuel-to-air mass ratio, gravitational constant	[21,149,150]	Direct links: [21] Upon request: -
vehicle description parameters	efficiency parameter for diesel engines, electric vehicle model, engineer displacement, engineer friction factor, engineer speed, frontal surface area, heating value of a typical diesel fuel, mass of vehicle and cargo, propulsion motor power, vehicle drive train efficiency, vehicle drivetrain efficiency	[21,23,46,149,151]	Direct links: [21,23] Upon request: -
Vehicle dynamics characteristics
acceleration	-	[46,50]	-
angular velocity	-	[50]	-
autonomous vehicle status information	-	[152]	-
energy consumption	-	[20,153]	Direct links: [20] Upon request: -
future data of the driving speed	-	[46]	-
OBD data	-	[153]	-

), vehicle dynamics characteristics (Table A8), trip characteristics (

Table A9. Route efficiency: trip, traffic, driver, infrastructure, and environmental characteristics.

Contextual Group	Features	References	Available in Refs.
Trip characteristics
historical driving data	historical data of average speed	[21,46]	Direct links: [21] Upon request: -
trip data	direction, distance traveled, route, spatial graph, time of travel	[20,21,23,49,148,153]	Direct links: [21,23] Upon request: -
Traffic characteristics
surrounding vehicles status information	-	[152]	-
traffic-related parameters	dynamic traffic data, real-time road traffic flow, traffic flow, traffic level	[20,21,46,49]	Direct links: [20,21] Upon request: -
Driver’s characteristics
driver profile-related parameters	-	[46]	-
Infrastructure characteristics
network data	-	[23]	Direct links: [23] Upon request: -
road-related parameters	road conditions, road gradient, road information, road network	[20,21,46,150,151,152]	Direct links: [20,21] Upon request: -
services	grid-to-vehicle, vehicle-to-grid services	[21]	Direct links: [21] Upon request: -
speed limits	upper speed limit, lower speed limit	[149]	-
traffic light	-	[49]	-
Environmental characteristics
weather-related parameters	air density, environmental parameters, temperature	[21,46,148,149,150]	Direct links: [21] Upon request: -

), driver characteristics (Table A9), infrastructure characteristics (Table A9), traffic characteristics (Table A9), and environmental characteristics (Table A9).

In particular, the tables (Table A8 and Table A9) contain the categorization of parameters and also references to the available datasets (both with direct links and upon-request directions).

4. Data Processing Techniques Enabling Efficient Vehicle Driving Analysis

The previous section focused on analyzing the availability of datasets and the different parameters obtained from related studies. This section is oriented to analyze those approaches aimed at extracting value from these data. The distribution of references according to the type of efficiency and general processing method is shown in Figure 4, where relevant techniques oriented to machine learning (ML), optimization methods, and other approaches are studied.

As illustrated in this figure, the literature review shows that optimization approaches are most widely applied to resource and route efficiency analysis, while machine learning-based methods are applied to driver efficiency analysis. This is mainly because the end goals in both applications are very much geared to these types of techniques, the former being suitable for efficiency maximization and the latter for the classification of driver behavior. These relevant techniques are discussed in the following subsections.

4.1. Optimization Methods

Optimization methods, when applied to datasets in the transportation domain, allow us to analyze the influence of driving style and traffic measures on vehicle emissions and fuel consumption [2], discover driving strategies [29], control vehicle speed for eco-driving [48], reduce fuel consumption and exhaust pollutant [41], construct energy efficient routes [49], and plan speed at signalized intersections [11], among others.

According to the literature review conducted, to analyze the driver’s efficiency, the following optimization methods could be applied: the non-dominated sorting genetic algorithm (NSGA-II), multiple objective particle swarm optimisztion (MOPSO), the Pareto envelope-based selection algorithm (PESA-II), the strength Pareto evolutionary algorithm (SPEA 2), and fuel consumption models (

Table A10. Optimization methods for driver and route efficiency analysis.

Reference	Processing Methods	Data Type	Features	Objective
Driver’s Efficiency
[2]	roll-bench emission tests, vehicle simulations, regional emission modeling	vehicle data, traffic data, infrastructure data	vehicle parameters, traffic measures, driving behavior	Perceiving the environment
[83]	microscopic fuel consumption models	data from the simulator (MATLAB application)	V2I data, signal change information, vehicle data, lead vehicle data, intersection features	Vehicle Control
[29]	multiobjective optimization algorithm, nondominated sorting genetic algorithm	sensor data	length, turning radius, inclination, velocity limit	Perceiving the environment
[48]	microscopic fuel model	sensor data, V2X data	V2V/V2I communication, connected vehicles	Perceiving the environment
[84]	optimization mathematics model, offline optimization stream	vehicle data	vehicle parameters	Vehicle Control
[91]	nondominated sorting multi- objective genetic algorithm, strength Pareto evolutionary algorithm	vehicle data	acceleration value(s) with the associated duration(s), controller gains number of accelerations	Perceiving the environment
[30]	comprehensive modal emissions model	driver questionnaire, sensor data	driver characteristics, vehicle characteristics, driving conditions, GPS	Perceiving the environment
[154]	grounded group decision making	sensor data	vehicle parameters, driver’s characteristics, energy consumption	Perceiving the environment
[79]	nondominated sorting genetic algorithm, multiple objective particle swarm optimization, Pareto envelope-based selection algorithm, strength Pareto evolutio- nary algorithm	sensor data, vehicle data	road characteristics, vehicle acceleration sections, motor parameters, battery parameters	Vehicle Control
[92]	multinomial Radau pseudo-spectral method	vehicle sensor data	speed, acceleration, power consumption	Vehicle Control
[11]	traffic efficiency optimization, energy consumption optimization, driver comfort optimization	vehicle data	speed, space difference, acceleration	Vehicle Control
[85]	genetic algorithm	sensor data	time-to-collision, vehicle data	Vehicle Control
[93]	intersection model, vehicle status model	vehicle data, infrastructure data	structure of the intersection, speed limit, safety gap speed, acceleration	Vehicle Control
[98]	regression analysis	vehicle data, traffic data, infrastructure data	vehicle’s position, speed, additional relevant information	Perceiving the environment
[103]	resequencing and platooning algorithm	vehicle data	vehicle parameters	Vehicle Control
Route Efficiency
[151]	two-objective hybrid local search algorithm	vehicle data, road data	vehicle parameters, road gradient	Route Planning
[21]	data mining, constrained shortest path algorithm, dynamic route planning method, swarm optimization method, multi-objec- tive route planning for solar-power- ed electric vehicles	sensor data, vehicle data, force and motion models	historical driving data, vehicle parameters, battery parameters, road conditions	Perceiving the environment
[20]	autoregressive integrated moving average model, non-parametric kernel regression method	vehicle data, traffic data	traffic flow, state of charge, road network, travel time, energy consumption	Route Planning
[149]	improved simulated annealing algorithm	vehicle data	vehicle technical parameters, fuel consumption, speed values	Route Planning
[49]	dynamic space-time route, velocity- space-time three-dimensional network model	vehicle data, infrastructure data	real-time traffic, traffic lights, vehicle driving status	Route Planning
[49]	shortest path algorithms, routing algorithms, temperature-dependent model, graph model, energy consumption predictive model	sensor data	historical data (speed, accelerations, road gradient), vehicle-, road-, traffic-, weather-, driver related parameters, battery model	Route Planning
[153]	maximal-frequented-path-graph shortest-path algorithm	vehicle data, infrastructure data	OBD data, energy consumption of road segments	Route Planning
[148]	nondominated sorting multi-objec- tive genetic algorithm, Pareto envelope-based selection algorithm, Pareto archived evolution strategy, strength Pareto evolutionary algorithm	vehicle data	state of charge, speed, distance, route, environmental parameters	Vehicle Control
[152]	Bezier curve fitted trajectories, artificial potential field method	vehicle data, sensor data	vehicles status information, road information	Route Planning
[150]	MATLAB Simulink model	simulation data	aerodynamics, wind speed, topology of roads	Vehicle Control
[89]	physical-data-driven distributed predictive control strategy	sensor data	vehicle parameters	Vehicle Control

). For route efficiency analysis, the following processing optimization methods could be used: the constrained shortest path algorithm, the dynamic route planning method, the swarm optimization method, and multi-objective route planning (Table A10).

Resource consumption efficiency analysis could be conducted by applying such processing methods as infinite-dimensional optimization, discretized optimization models, temporal-based methods and spatio-temporal methods, minimum energy cost with time cost constraint, minimum time cost with energy cost constraint, multi-agent system (MAS) modeling, and the mutant particle swarm optimization (MPSO) algorithm (

Table A11. Optimization methods for resource consumption efficiency analysis.

Reference	Processing Methods	Data Type	Features	Objective
Resource Consumption Efficiency
[41]	route optimization	vehicle data	vehicle parameters, GPS	Vehicle Control
[42]	multi-objective optimization	sensor data	route information, vehicle motion, location data, road geometry	Perceiving the environment
[43]	velocity, time and distance discreti- sation, nonlinear model predictive control	sensor data, vehicle parameters	V2V communication data	Perceiving the environment
[44]	model predictive control, Krylov subspace method, Pontryagin mini- mum principle, dynamic program- ming, numerical optimization techni- ques	sensor data	discrete gear ratio, vehicle trajec- tory, nonlinear vehicle dynamics, GPS data, GIS data	Perceiving the environment
[110]	multiple regression analysis	sensor data, calculations, questionnaries	vehicle parameters, battery pack current, voltage, acceleration and energy consumption (calculated)	Vehicle Control
[129]	optimal path search algorithm, regression-based approach, physics- based approach, Pontryagin mini- mum principle, dynamic program- ming	traffic data, vehicle data	vehicle parameters, road network topology, real-time traffic condi- tions, real-time data from maps web-services	Route Planning
[130]	linear regression, multi-level models	sensor data, CAN data	drivers’ characteristics, car attribu- tes, date-specific environmental attributes, travel behavior	Perceiving the environment
[131]	graph-based model	sensor data, V2X data	connected and automated vehicles parameters, traffic, signal phase and timing information	Route Planning
[33]	piecewise method, infinite-dimen- sional optimization, discretised optimization models, temporally based methods and spatiotemporal methods, minimum energy cost with time cost constraint, minimum time cost with energy cost constraint	traffic data, weather, vehicle sensors	environmental and weather condi- tions (wind speed, rolling resistance), traffic	Route Planning
[111]	Monte Carlo method	vehicle data	battery parameters, location, speed, azimuth	Vehicle Control
[47]	Pontryagin minimum principle, driving mission as an optimal cont- rol problem, dynamic programming	sensor data, topographic information	physical traits, road slopes, speed limits, safety distance, stop signs, traffic lights	Vehicle Control
[123]	Pontryagin minimum principle	simulation data	energy consumption, position, speed, acceleration	Vehicle Control
[146]	model predictive control with fuzzy-tuned weights	simulation data	road slope information, vehicle position and velocity	Vehicle Control
[14]	real-time gang scheduling, dynamic energy optimization, geometric programming	vehicle data	speed, battery charge, power consumption	Vehicle Control
[5]	multiagent system modeling	vehicle data, traffic data	vehicle parameters, traffic parameters, charging parameters	Perceiving the environment
[112]	sliding mode observer	vehicle data	voltages, input currents	Perceiving the environment
[144]	vehicle specific power model	simulation data	speed, fuel consumption	Perceiving the environment
[140]	Helbing’s optimal velocity function, car-following model	sensor data	velocity, acceleration, length of road, number of vehicles, position	Perceiving the environment
[117]	fuel estimation algorithm	vehicle data	engine parameters, fuel parameters, vehicle parameters	Perceiving the environment
[134]	signal control methods, eco-driving algorithms based on the connected vehicles	infrastructure data	intersection parameters, traffic, traffic lights, speed, acceleration	Vehicle Control
[133]	Dijkstra’s algorithm, fastest k-route, eco-route, virtual and adaptive traffic light, advisory system	sensor data	vehicle, traffic, roadways, drivers, weather conditions	Vehicle Control
[132]	multi-objective optimization, clustering methods, Rint model (modeling method of battery)	simulation data	signal information, speed limit, location, V2I and V2V data, vehicle specifications	Vehicle Control
[56]	rule-based power distribution, predictive energy management	vehicle data	driving route, current, charge	Vehicle Control
[127]	dynamic programming	experimental data	vehicle parameters, engine parame- ters, motor parameters, transmis- sion parameters, economic cost parameters	Vehicle Control
[135]	mutant particle swarm optimization algorithm	sensor data	vehicle model parameters (vehicle dynamic model, tire model, motor model)	Vehicle Control
[136]	minimum equivalent fuel consumption model	sensor data	vehicle parameters	Perceiving the environment
[124]	decision-making algorithm and an optimization-based trajectory planner	simulation data	vehicle model, energy consumption model	Route Planning
[138]	methods for optimal control prob- lem with multiple constraints under the model predictive control	sensor data, experimental data	vehicle physical parameters	Perceiving the environment

).

4.2. Machine Learning Methods

When applied in the transportation domain, machine learning methods allow us to estimate target aspects such as road grade [50], minimize the energy loss of driving platoon decisions [51], predict vehicle fuel consumption [52], predict the charging-related state for electric vehicles [53], predict the driving range of electric vehicles [54], analyze driving behavior [55], and predict road conditions and driving style [6].

Although there exist approaches focused on nonsupervised analysis, such as clustering [13] or unsupervised auto-encoders [8], most studies are widely oriented toward supervised prediction models. Among them, the most commonly applied methods for driver efficiency analysis are the following: support vector machine (SVM), neural network and deep learning-based approaches, and decision tree-based methods and ensembles, where random forest proposals stand out for their quantity. These methods are detailed in

Table A12. Machine Learning for driver efficiency analysis.

Reference	Processing Methods	Data Type	Features	Objective
Driver’s Efficiency
[25]	Neisser’s perceptual cycle	sensor data	headway time, lateral deviation, velocity, steering angle of car	Perceiving the environment
[82]	data fusion	sensor data	vehicle parameters, infrastructure parameters	Perceiving the environment
[27]	Gazis-Herman-Rothery car- following model, driver model, neurofuzzy modeling	vehicle data, traffic data, weather data	road design, weather, traffic, environment	Vehicle Control
[28]	data mining	sensor data, OBD data	speed, acceleration, engine parameters, position, route data	Perceiving the environment
[9]	auto-encoder model	sensor data	GPS trajectories	Perceiving the environment
[34]	Bayesian network trees, multi-layer perceptron	sensor data	accelerations, brakes, road structure, turns	Route Planning
[19]	attention-guided lightweight network	graphic data	images	Vehicle Control
[57]	hybrid convolutional-recurrent deep network	graphic data	frontal images	Vehicle Control
[36]	support vector machines, decision trees, Bayesian learners, ensemble learners, evolutionary algorithms	vehicle data, driver data, environment data	vehicle-based measures, physiolo- gical-based measures, behavioral- based measures	Perceiving the environment
[18]	deep learning, data representation methods, data feature extraction methods, detection and prediction methods	LiDAR data	vehicle’s surroundings	Vehicle Control
[15]	object detection in thermal images through style consistency	thermal images	object classes	Vehicle Control
[8]	unsupervised deep auto-encoder peer dependency	sensor data	GPS vehicle data, GPS trajectories	Perceiving the environment
[6]	random forests, decision trees, support vector machine algorithms	in-vehicle sensor data	speed, engine parameters, fuel consumption	Perceiving the environment
[39]	transfer learning, multi-layer- perceptron, ReLU-activation, stochastic gradient descent training	vehicle data, driver’s data	vehicle parameters, driver’s characteristics	Perceiving the environment
[80]	support vector machine algorithms	vehicle data, driver’s data	location, gender, battery para- meters, speed, acceleration, deceleration, road types	Perceiving the environment
[100]	bioinspired approach-sensitive neural network	graphic data	images	Vehicle Control
[101]	random forest, Naive Bayes classifiers	vehicle data	age, acceleration, brake, speed variation	Perceiving the environment
[102]	long short-term memory network	vehicle data	speed, distance, historical trajectory, environmental data	Vehicle Control
[12]	deep neural network-based reward function for inverse reinforcement learning with efficient model personalization via machine unlearning, ConvLSTM-based RewardNet	vehicle data	vehicle trajectory	Perceiving the environment
[22]	supervised learning algorithms	vehicle data	acceleration, engine parameters, fuel consumption parameters, speed parameters	Vehicle Control
[13]	clustering analysis	vehicle trajectories	vehicle characteristics, position, location	Perceiving the environment
[55]	support vector machines, AdaBoost, random forest	OBD data	acceleration, engine parameters, fuel consumption, speed, vehicle operation parameters	Perceiving the environment
[107]	OpenCV, Viola Jones algorithm	camera data	eye closure, yawning, head nodding, scaling	Perceiving the environment
[96]	regression, classification, clustering	experimental data, simulation data	vehicle parameters, vehicle operation parameters, weather, traffic signals	Vehicle Control
[108]	deep learning model	sensor data	eye closure, open-eye state, yaw- ning, and non-yawning instances	Perceiving the environment
[76]	flexible multimodal federated learning method	sensor data, camera, physiological sensor	driver’s characteristics	Perceiving the environment
[104]	fuzzy logic estimator based on time-to-collision and time-to-gap	simulation data	vehicle parameters, trajectories	Vehicle Control
[105]	deep learning	simulation data	vehicle parameters, trajectories	Vehicle Control

with regard to the main features they consider and the target features guiding the search.

Resource efficiency analysis is conducted by applying techniques such as the gradient-boosting decision tree algorithm, backpropagation neural network, recurrent neural network, and deep reinforcement learning. More details on the processing methods, data types, and necessary features can be found in

Table A13. Machine Learning methods for resource consumption and route efficiency analysis.

Reference	Processing Methods	Data Type	Features	Objective
Resource Consumption Efficiency
[45]	hybrid model predictive controller, explicit model predictive control method, Bayes Network model	V2X data	location, road profile, limitations, traffic, velocity prediction, traffic lights, full electric, vehicle model, human driving data	Perceiving the environment
[54]	machine learning, multiple linear regression method, gradient boosting decision tree algorithm	vehicle data	real-time data of travel and battery status	Vehicle Control
[125]	back propagation neural network, support vector regression, random forests	sensor data, OBD data	speed, acceleration, position, torque, revolutions per minute, state of air conditioning	Vehicle Control
[143]	support vector machine	sensor data, OBD data	fuel consumption, revolutions per minute, throttle position sensor	Perceiving the environment
[17]	deep learning model, artificial neural network, recurrent neural network, long short-term memory	vehicle data, weather data	vehicle parameters, road data, weather characteristics	Perceiving the environment
[16]	deep neural networks, linear regression, decision tree	vehicle data, weather data, traffic data	vehicle telemetry data, elevation and street-level maps, weather characteristics, traffic	Route Planning
[74]	Markov chain neural network	sensor data	historical GPS data	Vehicle Control
[58]	neural network with attention mechanism, deep neural network	vehicle data	driving data, operation conditions, traffic information of road network, vehicle state	Perceiving the environment
[10]	support vector machine regression, elastic net, data-driven regression modeling, web-scraping, text- mining techniques	vehicle data	electric vehicle characteristics, battery characteristics, performance specifications	Perceiving the environment
[114]	autoregressive network with exogeneous input	vehicle data	state of charge, traffic information	Perceiving the environment
[128]	machine learning algorithm	vehicle data	rolling resistance force, force related to road steepness, aerodynamic force, acceleration force, vehicle parameters	Perceiving the environment
[126]	deep reinforcement learning	vehicle data	vehicle parameters, electric motor parameters, internal combustion engine parameters	Vehicle Control
[53]	deep learning model, recurrent neural network, long short-term memory	vehicle data	state of charge	Vehicle Control
[113]	machine learning	vehicle data, driver’s data	battery characteristics, road topology, slope angle, wind speed, vehicle load	Perceiving the environment
[115]	deep reinforcement learning	vehicle data	vehicle parameters, engine parameters, battery parameters	Vehicle Control
[137]	reinforcement learning	map data, sensor data, V2X, V2V	traffic information, vehicle information	Vehicle Control
[52]	support vector machine, random forest, artificial neural network, deep neural network	sensor data, OBD data	vehicle characteristics, driver’s characteristics, weather	Perceiving the environment
[51]	reinforcement learning	sensor data	traffic, speed, acceleration, deceleration	Vehicle Control
[73]	integration of machine learning and physics-based models	on-board sensors, simulation data, data augmentation	vehicle parameters	Vehicle Control
[118]	adaptive network-based fuzzy inference system	vehicle parameters	historical vehicle velocity, accelerations, battery state of charge trajectory	Vehicle Control
[142]	deep reinforcement learning	sensor data	rolling resistance, acceleration, travelled distance	Vehicle Control
Route Efficiency
[50]	sensor fusion techniques (Kalman and complimentary filters)	phone sensors, vehicle data	acceleration, angular velocity	Vehicle Control

. When analyzing route efficiency, sensor fusion techniques could be applied (Table A13).

4.3. Other Processing Methods

In addition to optimization and machine learning methods for efficient vehicle driving analysis, the literature review showed a wide range of other data processing techniques. Thus, the following methods could be applied for driver efficiency analysis: cellular automaton, simulation methods, mathematical modeling, and statistical analysis. Resource consumption is analyzed using simulation and statistical analysis methods. Route efficiency is mainly estimated using graph theory methods. More details on other data processing techniques applied to efficient vehicle driving analysis can be found in

Table A14. Other methods for efficient vehicle driving analysis.

Reference	Processing Methods	Data Type	Features	Objective
Driver’s Efficiency
[90]	Dempster–Shafer’s theory (cellular automaton)	sensor data	speed, acceleration, position on the road, position in the lane, highway, number of lanes	Perceiving the environment
[88]	Nagel–Schreckenberg model, Kerner–Klenov–Wolf model (cellular automaton)	simulation data	speed, acceleration, position, trajectory	Perceiving the environment
[99]	visual, numerical analysis (fuzzy logic)	vehicle data, questionnaires	traffic, travel distances, temperatu- re, trip average road grade, local wind speed	Perceiving the environment
[97]	fuzzy-logic scoring (mathematical modeling)	sensor data, topographic information	road congestion, recharging points on the path, road slopes, position, speed, instantaneous power consumption	Perceiving the environment
[32]	visual, numerical analysis (probability theory)	sensor data	speed, acceleration, driving mode duration, fuel consumption	Perceiving the environment
[35]	clustering (simulation)	vehicle data, OBD data	position, speed, acceleration	Perceiving the environment
[37]	mean time assessment (simulation)	sensor data, CAN bus data	speed, acceleration, braking, fuel consumption, driving operations	Perceiving the environment
[95]	visual, numerical analysis (simulation)	vehicle data	longitudinal and lateral acceleration, type of road	Perceiving the environment
[38]	microscopic traffic flow simulation model (simulation)	simulation data	vehicle parameters, baseline traffic model, speed profiles, roundabout specifications	Perceiving the environment
[75]	numerical simulation (simulation)	vehicle data, traffic data	distance between all drivers, signals (e.g., gas pedal, brake pedal, speed, steering wheel angle, route deviation, etc.)	Perceiving the environment
[81]	digital twin (simulation)	simulation data	vehicle body parameters, engine, drivetrain, battery	Vehicle Control
[87]	game theory (statistical analysis)	vehicle data, environmental data	drivers’ characteristics, car attribu- tes, environmental parameters	Perceiving the environment
[94]	comprehensive modal emissions model (statistical analysis)	simulation data	fuel consumption, emission of CO and NOx, average deceleration	Vehicle Control
[106]	CRITIC weighting model, cluster analysis, analysis of variance (statistical analysis)	simulation data, eye tracker data	eye movement, visual characteris- tics indicators, traffic conditions, driving style	Perceiving the environment
[86]	Wilcoxon signed-rank test (statistical analysis)	simulation data	road data, vehicle data, navigation	Perceiving the environment
Resource Consumption Efficiency
[40]	vehicle simulation model (simulation)	vehicle data	vehicle parameters	Vehicle Control
[121]	traffic micro-simulation, vehicle emission model (simulation)	generated data	emissions, trajectory	Vehicle Control
[120]	statistical methods (statistical analysis)	sensor data, OBD data	fuel consumption, emissions, moving speed, acceleration, brake usage, position	Perceiving the environment
[7]	agglomerative hierarchical clustering combined with an L-term heuristic (statistical analysis)	sensor data, GPS logs	GPS trajectories	Perceiving the environment
[116]	mathematical models of energy flow, discharging, charging (statistical analysis)	vehicle data, road data	speed, engine, battery state of charge	Perceiving the environment
[145]	vehicle specific power distributions (statistical analysis)	vehicle data, OBD data	location, operation parameters	Vehicle Control
[24]	analysis of variance (statistical analysis)	sensor data	acceleration, speed, engine, consumption	Perceiving the environment
Route Efficiency
[23]	graph theory (simulation)	network data, trip data, vehicle data	nodes, links, scenarios, properties, trip time, properties, energy/ fuel maps	Route Planning

, where processing methods, data types, main features, and target variables (objectives) are organized and detailed.

5. Future Trends and Perspectives

This section is devoted to the future trends and perspectives identified. They are classified into two main categories: the creation of datasets to test new approaches (Section 5.1) and the development of solutions to support drivers (Section 5.2).

5.1. Alternative Data Sources

The authors of 14% of the selected articles indicated that the data for the research were obtained using a vehicle simulation model [29], a dynamic car model [43], linear and network modeling [49], and simulators such as Simulink [40,56], Udacity [57], Paramics Microsimulation [58], or CarMaker [56]. So, simulation models have been applied early in research proposals but are evolving to more sophisticated approaches.

In the context of data collection based on simulation, an interesting approach to collect large datasets on energy-efficient driving using the networked game is suggested in [59,60]. The iCO₂ game belongs to the category of “serious games” or games with a purpose, aimed firstly at education and training, and secondly, at entertainment for users. The developers propose using the game to collect data on driving behavior on a large scale, taking into account the aspect of energy efficiency. The game takes place in one square kilometer, and users interact not only with each other but also with non-player characters within the game. iCO₂ provides driver behavior data: timestamps, speed, carbon emissions, remaining fuel, and vehicle location parameters.

iCO₂, as a massive multiplayer online driving game, allows us to collect data on driver behavior along with information on decisions made by users during the game [60]. The obtained data allows us to analyze the environmental impact and energy efficiency of driving, the correspondence between driving and other actions of the driver (refuelling the car, exploring the game environment, and upgrading the car), and also describe how the driver’s behavior evolves. In addition, the structure of the data obtained from the game allows the identification of four types of drivers using cluster analysis [60]: eco-accelerator, eco-braker, regular, and reckless.

In addition to specially designed platforms for collecting driving data [59,60], simulated racing video games such as Assetto Corsa (available at https://www.assettocorsa.it, accessed on 10 March 2025) can also be used for this purpose [61,62]. Assetto Corsa allows us to obtain the following data: lap characteristics, acceleration, braking and steering parameters, vehicle parameters, and location parameters [61].

Authors of [62] provides an analysis of simulated racing video games from the point of view of the use of real road infrastructure, vehicle selection, and the possibility of exporting telemetry data for the analysis of speed, trajectories, and other parameters. The following simulated racing video games were considered: WRC 9, GT Sports, F1 2020, Project Cars 2, Automobilista 2, Dirt Rally, rFactor 2, Assetto Corsa, iRacing, and RB Rally. However, the requirements for actual road infrastructure and vehicles are only met in rFactor 2, Assetto Corsa, and RB Rally. rFactor 2 and Assetto Corsa allow special plugins to collect physical parameters from the game environment. These parameters include position, speed, acceleration, vehicle tilt angles, engine parameters, tire parameters, and temperatures.

A similar approach to data collection can be found in [63], where the authors consider serious video games to address driver behavior in the incidents. Data were collected using video games, and the impact of video games on driver behavior was analyzed.

5.2. Mobility Perspective

When considering the perspectives for the development of transportation systems, it is necessary to consider trends in the development of virtual, augmented, and mixed reality technologies, as well as the digital environment (metaverse).

According to [64], given the increasing attention to technologies for creating and functioning in the digital environment, it is necessary to consider the domain of “metamobility” from the standpoint of creating tactile live maps and advanced driver assistance systems (ADAS) based on metadata. In the last decade, car manufacturers have been researching the possibilities of introducing cloud technologies into their products and services, collaborating with technology corporations to develop automotive operating systems, cloud platforms, and artificial intelligence tools for the needs of the transportation industry. Some automakers already use or plan to use the Android operating system for the car’s infotainment system. Authors of [65] provides substantial research on the topic of future multimedia consumption in autonomous vehicles, introducing the concept of a moving metaverse.

Tactile live maps contain five layers according to the frequency of information changes (dynamic or static) and the function of augmenting the perception of the real environment or enhancing interaction with the virtual environment and its content [64]. Traditional ADAS systems can be enhanced with metadata from various sources and objects. It is not only about data from sensors or infrastructure, but also historical data. Among the necessary parameters, the following can be distinguished: longitudinal/lateral position, speed, and acceleration; steering wheel, acceleration/braking pedal, and the human–machine interface; data from the in-cabin monitoring camera, seat pressure sensor, and steering wheel pressure sensor; data from neighboring vehicles, signal phase and timing (SPaT), speed limits, construction zone alerts, and traffic congestion levels [64].

Historical data on drivers and vehicle operation allow more accurate parameters to be determined for efficient driving support, predicting the dynamics of surrounding objects and the environment, and information on signal phase and timing, together with information on route congestion, allow ADAS to calculate an energy efficient speed trajectory in the green corridor of traffic lights [64].

Finally, according to [66], the digital environment can also be beneficial when developing vehicular networks and autonomous car applications. One of the motivations for using the metaverse for vehicular networks is proactive learning. In vehicular networks, the metaverse is introduced at the edge. Conversely, the metaverse enables the possibility of self-configuring the vehicular network.

One more trend that has been widely covered in recent publications is digital twin technology. Digital twins are described in the context of energy consumption prediction [67], smart electric vehicle charging infrastructure [68], and autonomous vehicular systems [69], to name a few.

We also find vehicular twins and emerging vehicular metaverse in [70,71,72], and it clearly defines the vector of trends and perspectives for future development of efficient vehicle driving in the digital era.

6. Practical Notes

The quality and reliability of efficient vehicle driving analysis depend in many ways not only on the availability of the datasets, but also on the completeness and variety of the data. This problem is reviewed in [73], and the authors suggest that this issue be addressed by data augmentation and synthesis. The necessary yet missing data can be generated or obtained by merging datasets.

Another option to receive the required data is to collect the data, particularly for research. This approach was chosen by the authors of [28,30,32,34,35,74,75,76].

However, this approach has several issues that should be considered and appropriately addressed. We consider the issues that must be taken into account when collecting new data (Section 6.1) or using existing datasets (Section 6.2). To do so, a practical use case for road damage detection [77] is provided to illustrate the adoption of commonly used methods to solve these issues.

6.1. Data Collection Issues

Research on road damage detection [77] revealed several issues concerning the collection of data from mobile phone sensors during vehicle movement. The key aspect of road damage monitoring is to detect anomalies on the road surface, such as potholes, cracks, and bumps, which affect driving comfort and on-road safety. So, the main goal was to develop software for analyzing road conditions based on data received from mobile phone sensors while driving. The primary phone sensors used to collect data were an accelerometer, gyroscope, magnetometer, gravity sensor, rotation sensor, and GPS. From this work, we have detected some issues that need to be considered.

Issue 1: Placing the phone. The position of the phone inside the vehicle could affect the data quality. The phone was placed in three mounting positions inside the car to investigate this problem: the windshield, the front panel, and beside the onboard computer. The last position was checked to determine whether it was important to firmly fix the device for correct data collection. The signals from the phone on the windshield and front panel were similar, while the accelerometer attached next to the onboard computer provided unexpectedly high values. So, it is preferable to attach the phone to the front panel of the vehicle, as such a location does not interfere with the driver that much, like the windshield, and produces acceptable results.

Issue 2: Binding to the area. The research showed that the permitted deviation for further analysis is 3.5–6 m. Under normal conditions, the deviation is an average of 4 m, which is sufficient for further analysis and corresponds to the typical measurement errors of modern GPS receivers. Due to GPS measurement errors, the route contains an error when reading data from the phone sensors in a moving vehicle. To eliminate this error, ref. [77] suggests approximating the user’s location to the nearest road.

Issue 3: Power consumption of the device. A significant issue with phone-based systems is the power consumption of the device. Several experiments were conducted on an iPhone XS smartphone using xCode Energy Impact tools and CPU Instruments to determine the most energy-consuming processes. Experiments showed that the most energy-consuming processes are the update frequency of motion sensors, data filtering, processing before sending to the server, and GPS accuracy. The update interval of the motion sensors was considered in the range of 0.05–0.2 s. With a higher update frequency, the device’s energy consumption was found to be too high. An update interval of 0.05 s provides very high energy consumption, while an update interval of 2 s provides only a high impact. At this point, it is essential to consider the quality of the collected data, as the update frequency directly affects this.

Data collection experiments showed that data collected at update intervals of 0.05 and 0.1 s are very similar, while the update interval of 0.2 cannot capture the required accuracy of data change. Therefore, it was not considered in further research.

When further considering the update frequency of motion sensors, a frequency of 0.1 s was chosen because the amount of energy consumed is less than that at an update frequency of 0.05 s, and the resulting values of sensor data in both cases have very slight differences. It was also decided that all possible computational operations should be transferred to the server. This approach made it possible to reduce energy consumption by another 30%. The energy consumption of the software could not be reduced to a minimum value, as this would affect the accuracy of the GPS operation. In addition, the application displays the results on the map, which requires a certain amount of resources.

6.2. Online Data Providers

As stated in Section 3, with further details identified in Appendix A, Appendix B and Appendix C, accurate and efficient vehicle driving analysis requires the parameters related to the driver, the vehicle, and the environment. Some of these parameters can be accessed through the online services and the respective APIs.

As a practical example of the usage of online services that provide global data, the project by [78] was considered. The main goal was to develop a mobile application to form the most energy-efficient route and display it to the user by analyzing the input data and applying them to the algorithm to calculate the consumed energy. The developed software is designed to determine the most energy-efficient route from the given starting point to the destination for electric vehicles based on the power of the vehicle engine and the energy consumed, calculated by a specific formula, and the efficiency map of the electric engine for acceleration and free movement, respectively. As a result, the system determines the amount of energy needed for possible routes, determines the least energy-consuming one, and highlights it on the terrain map.

An additional function of the application is to determine the location of charging stations for electric vehicles in case the estimated amount of energy needed for the most energy-efficient route exceeds the current state of the battery charge. To apply the algorithm for determining the amount of energy required for the selected route, the following online data providers were used (all data were mainly received in JSON format):

• Real-time weather data: air temperature, atmospheric pressure, and humidity. These data are used to calculate air density. The data source is the OpenWeatherMap free API (available at https://openweathermap.org/, accessed on 10 March 2025).
• Specifications of the electric vehicle: weight and width of the vehicle, drag coefficient, and wheel radius. These data are obtained from the API (available at https://api.auto-data.net/, accessed on 10 March 2025). There are approximately 560 electric vehicles in the sample.
• Route data: obtained sets of location points, which form a pair of geographic latitude and altitude values using the Microsoft Azure Maps API (available at https://azure.microsoft.com/en-us/products/azure-maps, accessed on 10 March 2025).
• Location of traffic lights, road congestion, terrain data, and the type of road surface, obtained from the Routes API by Google (available at https://developers.google.com/maps/documentation/routes, accessed on 10 March 2025) and the OpenStreetMap API (available at URL https://openstreetmap.org/, accessed on 10 March 2025).

7. Conclusions

In conclusion, we would like to focus on the results obtained regarding the stated research questions. As to the input data which are relevant for efficient vehicle driving analysis (RQ1), we identified the specific parameters needed to analyze driver behavior and vehicle performance, determining the corresponding energy consumption and the influence of route and environmental conditions on overall efficient vehicle driving. Based on the articles analyzed, the parameters were organized into tables according to efficiency types such as driver’s efficiency, resource consumption efficiency, and route planning efficiency (Appendix A, Appendix B and Appendix C). The tables not only identify the necessary features for the analysis but also mark the dataset containing those features, making it possible to locate the corresponding dataset if freely available.

Regarding dataset availability (RQ2), our analysis showed that only 15% of the articles provide datasets that are freely available. Some of the datasets can be made available upon request to the authors. Some of the datasets are prohibited from publication.

Among the techniques which are being applied to efficient vehicle driving analysis (RQ3), we can outline such groups of methods as machine learning, optimization methods, statistical analysis, simulation, mathematical modeling methods, and cellular automaton. Appendix D, Appendix E and Appendix F contain tables with an overview of the methods applied by the authors of the articles analyzed to determine driver efficiency, resource consumption efficiency, and route planning efficiency. The tables mentioned also contain the data types and specific features that were involved in the analysis.

As to alternative data sources for efficient vehicle driving analysis (RQ4), it is also worth noting that datasets collected by the authors themselves or generated in simulation environments are often used as the basis of the analysis. Thus, there is a clear gap that requires more research in the generation of real or realistic datasets that could be used to compare different solutions to the same problems.

Furthermore, since the availability of relevant datasets for the analysis of efficient vehicle driving remains at a rather low level, the direction of networked and “serious games” can be considered as an alternative source of data. As a side effect of collecting the necessary data using games, this approach could also improve the level of efficient vehicle driving by special teaching and training of drivers while interacting with the game.

We also addressed the issues of data collection (RQ5) for efficient vehicle driving analysis, revealed by the practical use case of collecting data from mobile phone sensors. The following issues can be distinguished: the positioning of the phone during data collection, additional binding to the area due to GPS measurement errors, and the device’s power consumption.

We hope that this survey and the aggregated data provide a valuable stepping stone for researchers and practitioners to join the community investigating efficient vehicle driving analysis.