Energy Economy Analysis on the Vehicle Driving Decision: Eco-driving Behavior Assessment of Vehicle

2 Methodology

The research on energy economy framework of vehicles can be traced back to 1895, John A. Michon proposed the three-layer structure of the driver behavior model, which are respectively the planning level, the tactical level and the control level. Then Sivak and Schoett founded that the relationship of correlation between these three levels of decision-making and the driving energy consumption. For example, there are strategic choices made with a long- term perspective, such as choosing to drive as little as possible, buying a “green car”, or maintaining the vehicle in top condition. Posteriorly, there are tactical decisions for each trip, such as choosing which route to avoid congestion or rideshare, thereby reducing excess emissions. Finally, there are operational decisions of driver’s on-road behaviors, such as skipping gears or engine braking. These driving behaviors correspond to different levels of driving decisions.

In order to understand each driving decision associated with the energy consumption and analyze the energy economy of vehicles comprehensively, it is necessary to have an overall frame and targeted single indicator to assess the eco-driving behavior quantitatively. Therefore, based on the "driving decision model", this paper selects the strategic decision, the tactical decision and the operational decision, and extracts the most relevant characteristic indicators affecting eco-driving behavior of these three driving decisions. Through the integration of qualitative and quantitative methods, it includes three parts: measurement dimension, index weight and data integration, so as to form a complete assessment of driver's eco-driving behavior. Table 1 summarizes the full range of factors and organizes them into the three levels of driving decisions. How exactly does each factor affect a vehicle's energy economy will be elaborated in the next section

Table 1

Comprehensive evaluation system of eco-driving.
First-level indicators	Second-level indicators	Third-level indicators
Strategic level (A₁)	Vehicle type (B₁)	Power source
	Vehicle class (B₂)	Body shape (E₁)
	Vehicle class (B₂)	Body size (E₂)
	Vehicle maintenance (B₃)	Maintenance period
Tactical level (A₂)	Traffic conditions (C₁)	Traffic light (F₁)
	Traffic conditions (C₁)	Traffic flow (F₂)
	Road conditions (C₂)	Road type (G₁)
		Longitudinal slope (G₂)
		Grade profile (G₃)
Operational level (A₃)	Driving cognition (D₁)	Driving style (H₁)
		Knowledge (H₂)
		Attitudes (H₃)
	Driving skill (D₂)	Smooth accelerate (I₁)
		Reasonable distance (I₂)
		Reducing idle time (I₃)
		Coasting deceleration (I₄)
	Driving behavior (D₃)	Speed (J₁)
		Rapid acceleration (J₂)
		Rapid deceleration (J₃)
		Long-time idling (J₄)
	Driving setting (D₄)	Reminder function (K₁)
	Driving setting (D₄)	Support system (K₂)

2.1. Strategic decisions

At the strategic level, vehicle selection has the largest effect on vehicle energy economy [17]. Generally, giving priority to electric cars rather than fuel cars is an environmentally friendly way. Choosing the suitable vehicle size for needs also helps to reduce unnecessary emissions. And the regular car maintenance can further reduce losses in the energy conversion process. The specific factors are as follow, including the vehicle selection of type, class and maintenance.

2.2.1. Vehicle type

According to the source form of automobile power, it is mainly divided into fuel, hybrid and electric [18]. Fuel cars, where the gasoline or diesel fuel is an internal combustion engine to drive a vehicle. On the one hand, the compression combustion diesel engine has a higher energy conversion rate than the ignited gasoline engine, on the other hand, the diesel has the characteristics of non-volatile and larger power, so the diesel car is more environmental friendly than the gasoline car.

Affected by the increase of emissions standards, hybrid cars have emerged. The engine and the electric motor drive the wheels together, and the combination of these two power sources achieves minimal fuel consumption. According to the hybrid mode of energy and the efficiency of energy transmission, common hybrid vehicles are Hybrid Electric Vehicle (HEV), Plug-in Hybrid Electric Vehicle (PHEV) and Range Extend Electric Vehicle (REEV) in sequence. With the development of battery-driven technology, pure electric vehicle is what we should focus on in the future.

2.2.2. Vehicle class

From the perspective of vehicle size, the German automobile classification standard clearly proposes that vehicles are classified from A00 to D, according to the wheelbase, displacement, weight and other parameters. The vehicle class is higher in the alphabetical order is, its wheelbase is longer and its weight is larger, leading to more energy consumption during driving process.

From the perspective of vehicle shape, U.S. market has presented the mean rated energy economy of all available light-duty vehicles. On average, a car has 38% better energy economy than a pickup truck. Specifically, the mean mpg of the cars is 23.7, while the pickup trucks is 17.2.

2.2.3. Vehicle maintenance.

With the increase of vehicle running time, the consumable and abrasive parts of the vehicle will be easily deformed. Cars in good condition is a prerequisite for energy saving, so drivers must use the car in a regulated manner and take it in a reasonable maintenance.

This is conducive to ensure the normal work of each component, so that the tuned engine and transmission train could maintain a low friction state in order to improve mechanical efficiency, thus the driving resistance and energy consumption of vehicles can be reduced [19]. Furthermore, rolling resistance of tires varies among different tires of the same size. EPA estimates that tire inflation will result in the rolling resistance, with a 1 psi drop reducing fuel economy by about 0.3% [20].

2.2. Tactical decisions

At the tactical level, traffic and road dominate energy consumption on a macro level. Knowing the itinerary and anticipating traffic condition in congested areas are important to mitigate unnecessary energy consumption. In addition, drivers can plan and choose the appropriate travel time and route in advance to improve vehicle energy efficiency on-road [21].

2.2.1. Traffic

Traffic light

Frequent starting, stopping and idling of vehicles at signalized intersections will increase additional energy consumption[22]. Therefore, when drivers choose the road equipped with the reasonable traffic lights, it could improve the overall traffic efficiency of road effectively, thereby reducing the queuing time of vehicles in complex road sections [23–24].
Traffic flow

Congestion can be considered within the context of route selection as well because drivers in some situations can avoid congested routes. Shankar et al. found that energy consumption could be significantly reduced with the decreasing level of traffic congestion [25].

2.2.2. Road

Road type;

Different road types result in different vehicle average speeds and the change of acceleration and deceleration[26]. Wang et al. studied the road types (expressways, urban roads and local streets) how influences the road energy-saving[27]. While because of complex terrain and topography, local streets require higher energy consumption.
Grade profile;

Grade profile has a strong effect on energy economy, which is mainly reflected in the difference of rolling resistance coefficient under different road conditions. Boriboonsomsin and Barth found that, in a particular scenario with the same origin and destination but two alternative routes, a flat route yielded 15–20% better energy economy than a hilly route. A planar road provides a stable driving environment for the vehicle, it can make full use of the high speed block driving, and the wheel rolling resistance is small so that it can reduce the vehicle running resistance.
Longitudinal slope.

Driving on the high slope road will lead to the increase of energy consumption. When the vehicle goes uphill, the slope resistance is generated by gravity factors, and the motor needs to consume more energy to overcome the resistance. And when the vehicle goes downhill, the braking energy of the motor is recovered. Through the regenerative braking system, the gravitational potential energy and kinetic energy could be converted and stored.

2.3. Operational decisions

Driving operation is the decisive factor of vehicle energy consumption in the actual driving process. Even if the “economic and environmental” vehicle operates in good road conditions, the low-level driver’s operation is difficult to maximize the energy effect economically. Studies have found that the impact of driving cognition and behavior on vehicle energy economy can reach 30%, and a reckless driver can wipe out all the energy-saving benefits brought by various vehicle technology progress[28]. Even for professional drivers, different driving operations would lead to 2%-12% difference in vehicle consumption.

2.3.1. Driving cognition

Driving style;

Driving style refers to the specific mode of thinking and behavior of drivers in the face of driving tasks [29]. There are many research have proved that driving style plays an important role in vehicle energy consumption and management[30–31]. According to the intensity of driving behaviors, driving style is generally divided into three categories: mild, normal and aggressive[32]. Devlieger et al. defined these three driving style by the change of acceleration: mild driving (a_s=1.00-1.45 m·s-1), normal driving (a_s=1.45–1.90 m·s-1) and aggressive driving (a_s=1.90–2.45 m·s-1).

Gonder et al. experimented with light vehicles and found that driving style is closely related to the vehicle's energy economy, which can cause about 30% variation on urban roads and about 20% in fuel consumption on highways in energy consumption [33]. Meseguer et al. [34] further verified through experiments that aggressive driving style leads to more energy consumption and CO2 emissions. This is because aggressive drivers usually make some sudden and unpredictable maneuvers, with the accelerator pedal and brake pedal changing within a large range frequently[35], which not only contravening the eco-driving principle, but also easily inducing dangerous situations in the process of driving [36]. While calm drivers tend to anticipate more on other road users’ movements and traffic lights, so in the face of abrupt situations, they only operate deceleration smoothly and emergency brake in low frequency[37].
Knowledge of eco-driving;

Driver training plays an important role in promoting eco-driving operation, because it disseminates eco-driving knowledge to students in the driving school. It can demonstrate clearly the benefits of implementing operational eco-driving behaviors and explain why they have an effect. The sufficient education actually encourages drivers to adopt ecological driving behaviors and helps them establish and maintain eco-driving habits [38].
Attitudes towards eco-driving

A great number of surveys show that 89% of consumers are willing to choose more energy economy behaviors because of the awareness of environmental protection. Dunlap pointed out that people who have a positive attitude towards the environment may develop the desire to engage in the implementation of the activity. That is to say, the more positive a person's attitude towards eco-driving is, the stronger his or her behavior on driving behavior and social responsibility.

2.3.2. Driving skill

Driving skill is usually defined as the driver’s ability to maintain the vehicle control and is generally used to differentiate between experienced and average ordinary drivers[39]. Experienced drivers could analyze the four mainly operation of vehicle in the actual driving process (idle, start/accelerate, cruise, decelerate/brake), and summarize qualitatively the corresponding energy-saving operation principle of the above four working conditions.

For example, drivers should be light on the accelerator pedal to accelerate smoothly. In the cruise process, it should maintain a reasonable distance between vehicles to avoid frequent braking acceleration. When the vehicle is idling, reduce idle time as much as possible. And during the deceleration, try to use coasting to reduce the braking frequency. If drivers can improve these driving skills and avoid bad driving behaviors, 5%-15% of energy consumption and exhaust emissions can be saved in general.

2.3.3. Driving behavior

Driving behavior is one of the main factors affecting vehicle energy consumption. Research has shown that the driving habits have an impact on vehicle fuel consumption of up to 30%[40]. This study maturity the evaluation dimension of eco-driving by Chinese scholar Wang Ping, taking four types of driving behaviors, namely, speed, rapid acceleration, rapid deceleration and long-term idling, as the characteristic parameters affecting energy consumption [41].

Speed;

If driver keeps a moderate speed of driving, for the traditional internal combustion engine vehicle, it could reduce the fuel consumption of at least 10% and the vehicle exhaust emissions of 20%~30%; while for the electric vehicles, on the basis of not changing the vehicle structure, it could save about 30% driving mileage[42]. Wang et al. found that the energy consumption per unit time is positively correlated with the cruise speed. Moreover, the optimal cruise speed is different for each type, for diesel vehicle, gasoline vehicle and electric vehicle is 40 ~ 50km/h-1, 60 ~ 80km/h and 50 ~ 60km/h respectively.
Rapid acceleration;

At the same speed, slamming on the oil port will not only bring higher instantaneous engine fuel injection, but also cause damage to the cylinder wall and piston ring, resulting in a large amount of energy waste. According to the existing researches of scholars, the threshold of rapid acceleration is 4 km/h, and the continuous rapid acceleration events are combined as one rapid acceleration.
Rapid deceleration;

The actual driving investigation shows that the rapid deceleration wastes the inertial force of the vehicle due to the use of clutch down pressure, resulting in unnecessary energy consumption loss. Combined with the existing researches of scholars, the threshold of rapid deceleration is -5 km/h, and the consecutive sudden braking events are combined as one rapid deceleration.
Long-time idling;

Idling refers to the speed (v_t) = 0 and acceleration (as_t) = 0, that is, the vehicle has stopped completely and has no tendency to accelerate. When the vehicle is idling, it needs to keep the fuel burning to keep the engine running. According to the summary of international energy-saving techniques, the long-time idling (t > 60s) has a great influence on fuel consumption.

2.3.4. Driving setting

An in-vehicle eco-driving system setting has been demonstrated to be effective in improving ecological driving behavior and reducing of energy consumption[43]. When drivers following eco-driving guidelines and recommendations, it will be more critical to the effectiveness of eco-driving technology than other factors, such as vehicle specifications or eco-method algorithms.

Eco-driving reminder function;

The vehicle intelligent network monitors the vehicle's instantaneous energy consumption and other relevant data. When the vehicle has non-ecological driving behaviors, such as too high rpm speed, too long time idling, successional rapid acceleration and sharp braking, the vehicle's electrical configuration would remind the driver by visual reminder, voice reminder, tactile reminder and etc.) is used to remind the driver.
Eco-driving support system.

According to the driving data and environmental conditions collected by vehicle infrastructure equipment (ACC, AEBS, LKA, road sign recognition, pedestrian recognition and etc), the vehicle intelligent network provides the optimized driving options and reminds drivers of the ecological driving method. It makes appropriate intervention on bad operating habits when necessary, and ecological driving behaviors can be gradually improved.

In the whole driving decision model, driving behaviors (such as driving skill, driving behavior and eco-driving setting), play one of the most significant roles in promoting a wide range of energy savings [44–45]. Driving smoothly and accelerating slowly have been shown to be more effective in reducing carbon emissions than strategic and tactical eco-driving decisions. Besides, the vehicle’s economy can also be improved by other aspect, such as maintaining a constant speed, changing gear smoothly, avoiding the vehicle idling and not using air conditioning reasonably[46], which could contribute about 45% reduction in total for the vehicle on-road energy economy per driver.

3 Results

3.1. Modeling based on AHP

3.1.1. Analytical Hierarchy Process (AHP)

The analytic hierarchy process (AHP) is a multi-criteria decision analysis method combining qualitative and quantitative analysis, which is mainly used for solving problems with several evaluation criteria layers proposed by T.L. Seaty [47]. Its basic idea is to transform the overall judgment of multiple elements into a “pairwise comparison”, and then evaluate their importance proportion and order of all elements by integrating the synthetical framework.

3.1.2. Weight calculated by AHP

We used AHP to structure the evaluation indexes of ecological driving strategy of automobile users (as shown in Table 1). We set up a pairwise comparison table for comparing the evaluation indicators in pairs. By consulting experts, we asked them to quantitatively describe the relative importance of the indicators, and constructed a judgment matrix according to the expert scores, so as to further calculate and obtain the relatively accurate weight of each indicator.

Construction of judgment matrix;

After the indicator structure is determined, the expert consultation method is adopted to design the questionnaire [51] and obtain the experts’ scores, in order to compare the relative importance of the influence of indica-tors in a certain level considering the indicator in the upper level. We used numbers 1–9 and its reciprocal as the scale to define the judgment matrix [52]. The criteria meaning of the number scale is shown in Table 2.

Table 2

Meaning of evaluation criteria in judgement matrix.
Scale	Meaning
1	The former and the latter are equally important.
3	The former is slightly more important than the latter.
5	The former is much more important than the latter.
7	The former is extremely more important than the latter.
9	The former is definitely more important than the latter.
2, 4, 6, 8	The degree is between two adjacent odd numbers.

We issued scoring scales to three experts in the fields of automobile industry and user research to judge the relative importance of indicators [53]. Eleven judgment matrices were constructed based on the geometric average of the three expert scores, which was the multiplication of the corresponding values and the power of 1/N (N represented the number of experts). Here, we take the first-level indicators as an example for analysis, and its judgment matrix of geometric mean is shown in Table 3.

Table 3

Judgment matrix of the first level indicators.
Indicators	A₁	A₂	A₃
A₁	1	5.944	1.671
A₂	0.168	1	0.531
A₃	0.598	1.882	1

The judgment matrix was expressed as: $A=\left(\begin{array}{ccc}1& 5.944& 1.671\\ 0.168& 1& 0.531\\ 0.598& 1.882& 1\end{array}\right)$.

Calculation of eigenvectors;

According to AHP, the eigenvectors were calculated as follows:

$${w}_{1}^{{\prime }}=\sqrt[3]{1\times 5.944\times 1.671}=2.150,$$

$${w}_{2}^{{\prime }}=\sqrt[3]{0.168\times 1\times 0.531}=0.447,$$

$${w}_{3}^{{\prime }}=\sqrt[3]{0.598\times 1.882\times 1}=1.040$$

To normalize them:

$${w}_{1}=\frac{{w}_{1}{\prime }}{\sum _{1}^{3}{w}_{\mathcal{i}}}=0.591$$

$${w}_{2}=\frac{{w}_{2}{\prime }}{\sum _{1}^{3}{w}_{\mathcal{i}}}=0.123,$$

$${w}_{3}=\frac{{w}_{3}{\prime }}{\sum _{1}^{3}{w}_{\mathcal{i}}}=0.286$$

According to the formula $A·w={\lambda }_{max}·w$, the largest eigenvalue ${\lambda }_{max}$ was computed as follows:

$${\lambda }_{max}=\frac{\sum ({A\bullet w)}_{i}}{n\times {w}_{i}}=\frac{1.762}{3\times 0.591}+\frac{0.374}{3\times 0.123}+\frac{0.864}{3\times 0.286}=3.046$$

Consistency test;

To examine the feasibility of the data, a consistency analysis was performed on the above results. First, we calculated $CI=\frac{{\lambda }_{max}-n}{n-1}=\frac{3.046-3}{3-1}=0.023$. Then, we could calculate $CR$. The smaller the $CR$ value is, the better the consistency of the judgment matrix is. If $CR<0.1$, the judgment matrix meets the criteria of the consistency test. If $CR>0.1$, it can be considered that the judgment matrix is not consistent, and it needs to be adjusted until it fits before next analysis. With the formula $CR=CI/RI$, and the known values: $CI=0.023$, $RI=0.520$, we could calculate $CR=0.044$, $CR<0.1$. Accordingly, the judgment matrix of the first level indicators constructed in this study met the criterion of consistency test, so the subsequent calculation results of weight were valid. Mean random consistency $RI$ can be found from the list. When $n=3$, we have $RI=0.520$, as shown in Table 4.

Table 4

List of mean random consistency.
n	RI
3	0.52
4	0.89
5	1.12
6	1.24
⡆	⡆

Complete model of ecological driving strategy;

Similarly, the above calculation method was applied to the judgment matrix of other levels of indicators, combining the geometric mean calculated from the scores of several experts. On the basis of respecting the will of its judging experts, the matrix that fails to meet the criterion of consistency test is modified appropriately until it passes the test. The final matrix eigenvectors are shown in Table 5, and the weight values shown in Table 6.

Table 5

Matrix eigenvectors of eco-driving system.
Matrix	Eigenvector	${{\lambda }}_{{m}{a}{x}}$	${C}{I}$	${R}{I}$	${C}{R}$
A	[1.762, 0.374, 0.864]	3.046	0.023	0.520	0.044
B	[2.066, 0.723, 0.211]	3.033	0.016	0.520	0.031
C	[1.609, 0.391]	2.000	0.000	0.000	-
D	[0.325, 0.970, 2.380, 0.325]	4.119	0.040	0.890	0.044
E	[1.432, 0.568]	2.000	0.000	0.000	-
F	[0.407, 1.593]	2.000	0.000	0.000	-
G	[0.589, 1.893, 0.518]	3.001	0.001	0.520	0.001
H	[1.871, 0.284, 0.845]	3.079	0.039	0.520	0.076
I	[0.701, 0.912, 2.020, 0.368]	4.125	0.042	0.890	0.047
J	[0.949, 0.762, 0.616, 1.674]	4.044	0.015	0.890	0.016
K	[0.773, 1.227]	2.000	0.000	0.000	-

Table 6

Weight indexes of eco-driving system.
First-level indicators	Second-level indicators	Third-level indicators	Comprehensive weight
Strategic level (A₁, 58.7%)	Vehicle type (B₁, 68.9%)	Power source	40.5%
	Vehicle class (B₂, 24.1%)	Body shape (E₁, 71.6%)	10.1%
	Vehicle class (B₂, 24.1%)	Body size (E₂, 28.4%)	4.0%
	Vehicle maintenance (B₃, 7.0%)	Maintenance period	4.1%
Tactical level (A₂, 12.5%)	Traffic conditions (C₁, 80.5%)	Traffic light (F₁, 20.3%)	2.0%
	Traffic conditions (C₁, 80.5%)	Traffic flow (F₂, 79.7%)	8.0%
	Road conditions (C₂, 19.5%)	Road type (G₁, 19.6%)	0.5%
		Longitudinal slope (G₂, 63.1%)	1.5%
		Grade profile (G₃, 17.3%)	0.4%
Operational level (A₃, 28.8%)	Driving cognition (D₁, 8.1%)	Driving style (H₁, 62.4%)	1.5%
		Knowledge (H₂, 9.5%)	0.2%
		Attitudes (H₃, 28.2%)	0.7%
	Driving skill (D₂, 24.2%)	Smooth accelerate (I₁, 17.5%)	1.2%
		Reasonable distance (I₂, 22.8%)	1.6%
		Reducing idle time (I₃, 50.5%)	3.5%
		Coasting deceleration (I₄, 9.2%)	0.6%
	Driving behavior (D₃, 59.5%)	Speed (J₁, 23.7%)	4.1%
		Rapid acceleration (J₂, 19.0%)	3.3%
		Rapid deceleration (J₃, 15.4%)	2.6%
		Long-time idling (J₄, 41.8%)	7.2%
	Driving setting (D₄, 8.1%)	Reminder function (K₁, 38.6%)	0.9%
	Driving setting (D₄, 8.1%)	Support system (K₂, 61.4%)	1.4%

3.2. Empirical analysis based on FCE

3.2.1. Fuzzy Comprehensive Evaluation (FCE)

Although we have acquired the weight of the factor and sub-factor, considering that these evaluation indicators have the characteristics of uncertainty and fuzziness, the fuzzy comprehensive evaluation (FCE) method can be used to quantitatively deal with these qualitative indicators [55].

3.2.2. Questionnaire

In order to verify our index system and find out whether each index can accurately measure the ecological driving performance of users, we designed a questionnaire based on the index system above. The participants were those who owned a certain type of car and had driving experience. The type and price of the vehicle they owned were not limited in order to make the questionnaire universal. Finally, a total of 40 questionnaires were distributed, and 33 valid questionnaires were recovered after removing the missing values (including 27 males and 6 females).

The questionnaire was composed of three parts: The first part was the basic information of the participants, including age, gender, city, car type, driving mileage in the past year, energy expenditure in the past year, etc. The questions of the second part were designed according to 22 three-level indicators, and the users evaluated the 22 indicators according to their own eco-driving performance. The third part is the overall score of the respondents on the eco-driving experience of the vehicle, which is measured by a question expressed as " Comprehensively speaking, how do you evaluate the eco-environmental protection degree of this car?". Participants were asked to rate from 1 to 6.

In the second part of questionnaire, each question was offered 6 options, measured in the form of a six-scale scale. Scale degree from strong to weak, in turn, was very much (rank 6), more (rank 5), little more than general (rank 4), little less than general (rank 3), less (rank 2), very few (rank 1), the description of different indicators would be slightly different.

For the 22 three-level indicators in the user eco-driving system, most of the options are set according to the above rules, but some questions are special. Taking "vehicle speed" and "power source" as an example, there is a logical jump between two questions. When participants choose gasoline cars, the options of vehicle speed are: below 60km/h (option 1), 60-80km/h (option 2), and above 80km/h (option 3). When subjects choose diesel cars, the speed options are: below 40km/h (option 1), 40-50km/h (option 2), and above 50km/h (option 3); When subjects choose other power types, all of which belong to electric vehicles, the speed options are: below 50km/h (option 1), 50-60km/h (option 2), and above 60km/h (option 3). In order to balance the scores for the other questions, questions with only three choices were given double scores, which are 2,4, and 6 points. Based on previous studies, option 2 of "vehicle speed" is marked as 6 points for the most energy saving, option 3 is marked as 4 points for the second level, and option 1 is marked as 2 points for the maximum energy consumption. The setting of all eight special indicators is shown in Table 7. In addition, the indicators "traffic light", " traffic flow ", "driving style", "rapid acceleration", "rapid deceleration" and "long-time idling" are scored in reverse.

Table 7

List of special items.
Indicators	Options	Score
Power source	Gasoline (1), Diesel (2), HEV (3), PHEV (4), REEV (5), BEV (6)	1, 2, 3, 4, 5, 6
Speed	Below 60km/h (1), 60-80km/h (2), Above 80km/h (3)	2, 6, 4
	Below 40km/h (1), 40-50km/h (2), Above 50km/h (3)	2, 6, 4
	Below 50km/h (1), 50-60km/h (2), Above 60km/h (3)	2, 6, 4
Body shape	Sedan (1), Minivan (2), SUV (3), Pickup truck (4), Off-road (5), Sports car (6)	6, 5, 4, 3, 2, 1
Body size	Class A00 minicar (1), Class A0 minicar (2), Class A compact car (3), Class B midsize car (4), Class C midsize car (5), Class D large car (6).	6, 5, 4, 3, 2, 1
Road type	Highway (1), City trunk (2), Local street (3)	6, 4, 2
Longitudinal slope	Downhill section (1), No slope section (2), Uphill section (3)	6, 4, 2
Grade profile	Flat roads (1), General roads (2), Hilly roads (3)	6, 4, 2
Driving style	Very moderate (1), More moderate (2), Slightly moderate (3), Slightly aggressive (4), More aggressive (5), Very aggressive (6)	6, 5, 4, 3, 2, 1

Next, we further obtained the comprehensive evaluation results by calculating the weighted average score of each participant, basing on the score of the questionnaire and the weight of each index. The weighted average is calculated as the comprehensive evaluation score ${X}_{j}$ of eco-driving for each subject, and the formula is as follows: ${X}_{j}=\sum _{i=1}^{20}{w}_{i}{a}_{ij}$ (j=1,2 ..., n), ${w}_{i}$ is the weight of the $i$th index determined by AHP as calculated above, ${a}_{ij}$ is the 6-point scoring value of the $i$th index by the $j$th subject, and n is the total number of subjects.

3.2.3. Reliability and validity tests

Reliability test - Internal consistency;

Internal consistency test was conducted on the original scores of 22 questions of 33 participants. As a result, Cronbach's Alpha value was 0.707, while Cronbach's Alpha value based on standardized terms was 0.709, both of which were above 70% (as shown in Table 8). Therefore, we considered the 22 questions of eco-driving evaluation system have high internal consistency, which means the results were reliable.

Table 8

Reliability test of 22 third-level indicators.
Cronbach's Alpha	Cronbach's Alpha Based on Standardized Items	N of Items	N of Cases
0.707	0.709	22	33
• Validity test - Criterion correlation;

Comprehensive evaluation (weighted average) and preliminary evaluation (arithmetic average) were calculated for the questionnaires of 33 participants. Then, correlation analysis was conducted with the single rating of ecological degree (as a subjective criterion) and the mean of mileage and fuel and electricity consumption (as objective criteria) to test the degree of correlation between the comprehensive rating and the criterion. It is a critical step for examining the accuracy of the comprehensive evaluation system measuring the objective construct of ecological degree.

The correlation coefficient between comprehensive evaluation and single rating was 0.480, while the correlation coefficient between comprehensive evaluation and consumption was 0.443, as shown in the Pearson correlation analysis results. Both coefficients were significant at the 0.01 level. With respect to the correlation degree of the two criteria, the correlation coefficient of the comprehensive evaluation calculated by the weighted average is significantly higher than the preliminary evaluation calculated by the arithmetic average, indicating that the score multiplied by the weights is closer to the subjective and objective criteria of ecological driving. Therefore, the comprehensive score calculated by weight based on user eco-driving index system had high criteri-on-related validity, accurately measures the ecological degree of vehicle drivers. The results are shown in Table 9.

Table 9

Criterion correlation test.
Indicators	Comprehensive evaluation	Preliminary evaluation	Single rating	Consumption
Comprehensive evaluation	1
Preliminary evaluation	0.424*	1
Single rating	0.480**	0.339	1
Consumption	0.443**	0.270	0.363**	1
**. Correlation is significant at the 0.01 level (2-tailed)
*. Correlation is significant at the 0.05 level (2-tailed)

References

Ramanathan, V.; Feng, Y., 2009. Air pollution, greenhouse gases and climate change: Global and regional perspectives. Atmos. Environ. 2009, 43, 37–50.
Emberger, G, 2015. Low carbon transport strategy in Europe—A critical review. Int. J. Sustain. Transp. 2015, doi:10.1080/15568318.2015.1106246.
Oliver H HGallagher K STian Det al.China’s fuel economy standards for passenger vehicles: Rationale,policy processand impacts[J].EnergyPolicy20037(11):4720–4729
Johansson H.Impact of Eco Driving on emissions and fuel consumption[R].1999
Riener A.Subliminal persuasion and its potential for driver behavior adaptation[J]IEEE Trans. Intelligent Transportation Systems.201213(1):71–80
ReitzR DDirections in internal combustion engine research[J].Combustion and Flame2013,160(1): 1–8.
Ericsson E.Independent driving pattern factors and their influence on fuel-use and exhaust emission factors [J].Transp Res PartD:Transpand Envir20016(5):325–345
Boriboonsomsin K.Barth MVu AEvaluation of driving behavior and attitude towards Eco-Driving: A southern California limited case study[C].90th Annual Meeting of the Transportation Research Board.WashingtonD.CUSA2011: 1–14.
Barkenbus,J.N., 2010.Eco-driving: an overlooked climate change initiative. Energy Policy 38 (2), 762–769.
Ando, R., and Y. Nishihori. 2011. “How Does Driving Behavior Change When following an Eco-driving Car?” Procedia – Social and Behavioral Sciences 20: 577–587.
M. Sivak and B. Schoettle, ‘‘Eco-driving: Strategic, tactical, and operational decisions of the driver that influence vehicle fuel economy,’’ Transp. Policy, vol. 22, pp. 96–99, Jul. 2012.
Stillwater, T., Kurani, K., 2012. In-vehicle ecodriving interface: Theory, design and driver responses. 91st Annual Meeting of the Transportation Research Board, Washington, DC, January.
García, M., Coloma, J. F., & Wang, Y. (2018). Eco-driving in small cities. Driving performance in relation to driver’s profile. Transportation research procedia, 33, 267–274.
Mensing,F.,Bideaux, E., Trigui, R., Tattegrain, H., 2013. Trajectory optimization for eco-driving taking into account traffic constraints. Transp. Res. Part D: Transp. Environ. 18 (0),55–61.
GONDER JEARLEYWINEMSPARKS W.Analyzing Vehicle Fuel Saving Opportunities Through In telligent Driver Fee
Strömberg, H. K., and I. C. M. Karlsson. 2013. “Comparative Effects of Eco-driving Initiative Aimed at Urban Bus Drivers – Results from a Field Trial.” Transportation Research Part D: Transport and Environment 22: 28–33.
Barkenbus, J. N. 2010. “Eco-driving: An Overlooked Climate Change Initiative.” Energy Policy 38 (2): 762–769.
Leijon, J., & Boström, C. (2022). Charging Electric Vehicles Today and in the Future. World Electric Vehicle Journal, 13(8), 139.
Xu Shaobing. (2016). The strategy to identify and control laws of vehicle fuel economy driving. (Doctoral dissertation, Tsinghua university).
EPA [Environmental Protection Agency] (2011). Gas mileage tips – Keeping your car in shape. Available at: http://www.fueleconomy.gov/feg/maintain.shtml.
Young, M. S., S. A. Birrell, and N. A. Stanton. 2011. “Safe Driving in a Green World: A Review of Driver Performance Benchmarks and Technologies to Support ‘Smart’ Driving.” Applied Ergonomics 42 (4): 533–539.
Tang T, Yi Z, Lin Q. Effects of signal light on the fuel consumption and emissions under car-following model[J]. Physica A:Statistical Mechanics and its Applications, 2017 (469):200–205.
MD-ABDUS-SAMAD K, HAYAKAWA T, IMURA J. Development and evaluation of an adaptive traffic signal con- trol scheme under a mixed-automated traffic scenario[J]. IEEE Transactions on Intelligent Transportation Systems, 2020, 21(2):590–602.
Wan N, Vahidi A, Luckow A. Optimal speed advisory for connected vehicles in arterial roads and the impact on mixed traf⁃ fic[J]. Transportation Research Part C:Emerging Technologies, 2016, 69:548–563.
Shankar Rmarco J.Method for Estimating the Energy Consumption of Electric Vehicles and Plug in Hybrid Electric Vehicles Under Real-world Driving Conditions[J]. IET Intelligent Transport Systems, 2013, 7(1):138–150
YAO EYANG ZSONG Y, et al.Comparison of Electric Vehicle's Energy Consumption Factors for Different Road Types[J].Discrete Dynamics in Na ture and Society, 2013, 1:1–7.
WANG Y, BOGGIO-MARZET A. Evaluation of eco-driv- ing training for fuel efficiency and emissions reduction ac- cording to road type[J]. Sustainability, 2018, 10(11):3891.
Z Org. Eco-driving: An overlooked climate change initiative.
K Hu, J Wu, T Schwanen. Differences in Energy Consumption in Electric Vehicles: An Exploratory Real-World Study in Beijing [J]. Journal of Advanced Transportation (S0197-6729), 2017, 2017(2):1–17.
Romm, J. J., and A. A. Frank. 2006. “Hybrid Vehicles Gain Traction.” Scientific American 294 (4): 72–79.
Martinez, C. M., Heucke, M., Wang, F. Y., Gao, B., & Cao, D. (2017). Driving style recognition for intelligent vehicle control and advanced driver assistance: A survey. IEEE Transactions on Intelligent Transportation Systems, 19(3), 666–676.
Ting Xu, Xiang Zhang, Yakun Zhang, Jian Wang. Vehicle Driver Safety Propensity Classification based on AdaBoost Algorithm [J]. Journal of Safety and Environment, 2019,19(4): 1273–1281.
GONDER JEARLEYWINEMSPARKS W.Analyzing Vehicle Fuel Saving Opportunities Through In telligent Driver FeedbackJ7SAE Paper2012-01-0494.
MESEGUER J E, TOH C K, CALAFATE C, et al. Driving- Styles:A mobile platform for driving styles and fuel con- sumption characterization[J]. Journal of Communications And Networks, 2017, 19(2):162–168.
L.Xu, J.Hu H. Jiang,and WMeng“Establishing style-oriented driver models by imitating human driving behaviorsIEEE Trans.Intell Transp.Systvol16no5pp2522-2530Oct2015.
O.Taubman-Ben-Ari,M.Mikulincer,and O.GillathThe multidimen-sional driving style inventory-Scale construct and validationAccident Anal. Preventionvol36no3pp323-3322004
Murphey, Y. L., Milton, R., & Kiliaris, L. (2009). Driver’s style classification using jerk analysis. 2009 IEEE Workshop on Computational Intelligence in Vehicles and Vehicular Systems, CIVVS 2009 – Proceedings, 23–28. http://doi.org/10.1109/CIVVS.2009.4938719.
MADLEN G, KACPERSKI C, KREMS J. Can electric vehi- cle drivers be persuaded to eco-drive? A field study of feed- back, gamification and financial rewards in Germany[J]. En- ergy Research & Social Science, 2020, 63101407.
T.Qu,H.Chen,D.Cao H.Guo and B.Gao(2015)Switching-based stochastic model predictive control approach for modelling driver steering skillIEEE Transactions on Intelligent Transportation SystemsVol16 No.1,pp.365–375.
Martin E,Chan N,Shaheen s.Understanding how ecodriving public education can result in reduced fuel use and greenhouse gas emissions [J]. Journal of the Transportation Research Board, 20122287: 163–173.
Wang Ping, Wan Wei, Zhang Ke, et al. Evaluation of ecological driving Behavior of taxi drivers [J]. Traffic Engineering, 2018, (in Chinese) 18(6):41–44.WANG Ping, WAN WeiZHANG Ke et al.Evaluation of low energy consumption driving behavior of taxi drivers[J].Journal of TransportationEngineer-ing,201818(6):41–44.
‘‘Internal project on ’Eco-driving’ in category B driver training and the driving test,’’ CIECA, Brussels, WI, USA, Final Rep. 05-11-07, 2007.
Magana, V. C., and M. Munoz-Organero. 2016. “Artemisa: A Personal Driving Assistant for Fuel Saving.” IEEE Transactions on Mobile Computing 15 (10): 2437–2451.
Y. Zhao, A. Wagh, Y. Hou, K. Hulme, C. Qiao, and A. W. Sadek, ‘‘Integrated traffic-driving-networking simulator for the design of connected vehicle applications: Eco-signal case study,’’ J. Intell. Transp. Syst., vol. 20, no. 1, pp. 75–87, 2014.
L.CuiandB.Park,‘‘Assessingeco-drivingbehaviorsusingdrivingsimu- lator,’’ in Proc. 96th Annu. Meeting Transp. Res. Board. Washington, DC, USA, Jan. 2017, pp. 1–16.
Barié D, Zovak GPerisa M.Effects of Eco-Drive Education on the Reduction of Fuel Consumption and CO2 Emissions [J].PROMET-Traffic &Transportation, 2013, 25(3):265–272.
Saaty, T. L. (1979). The US-OPEC energy conflict the payoff matrix by the Analytic Hierarchy Process. International Journal of Game Theory, 8(4), 225–234.
Wind, Y., & Saaty, T. L. (1980). Marketing applications of the analytic hierarchy process. Management science, 26(7), 641–658.
Crostack, H. A., Hackenbroich, I., Refflinghaus, R., & Winter, D. (2007). Investigations into more exact weightings of customer demands in QFD. Asian Journal on Quality. 2008, 8 (3): 71–80.
Jin-yu, G., Zhong-bin, Z., Qing-yun, S.: Study and Applications of Analytic Hierarchy Process. China Safety Science Journal (5), 148–153 (2008).
Xue, D., Jia-ming, L., Hao-jian, Z, Jun-yang, C., Jun-feng, Z.: Research on Computation Methods of AHP Weight Vector and Its Applications. Journal of Mathematics in Practice and Theory (7), 93–100 (2012).
Ying-luo, W.: System Engineering. China Machine Press, Beijing (2010).
Yan, W., Keying, L. H., Yi-wei, F., Jingyi, Y., Chang, Z., Junyi, W.: Construction of Evaluation Index System of Agriculture and Tourism Coupling Degree of Rural Complex Based on Production-living-ecological Space. Agricultural Outlook 16(6), 8–14 (2020).
Y.Y. Chen, Fuzzy Mathematics, Huazhong University of Science and Technology Press, Wuhan, 1984.
Chen, J. F., Hsieh, H. N., & Do, Q. H. (2015). Evaluating teaching performance based on fuzzy AHP and comprehensive evaluation approach. Applied Soft Computing, 28, 100–108.
omepage, http://www.springer.com/lncs, last accessed 2016/11/21.