1 Introduction

In recent years, many cities in China have seen accelerated levels of pollution. The quality of life and the health of general public in these cities have been impacted negatively. One of the major sources of pollution comes from vehicle traffic. Motor vehicles account for a significant portion of energy consumption. According to the government statistics reports for Beijing, the energy consumption in transportation field accounted for 25% of the city’s total energy consumption in 2011, up 7% from 2005. During the same period, the percentages of PM2.5, nitric oxide (NOx) and volatile organic compound (VOC) from motor vehicles were up to 31.1, 58 and 40%, respectively (Beijing Municipal Statistics Bureau 2011). To conserve natural resources, protect our environment and enhance the well-being of the general public, energy conservation and emissions reduction in vehicle traffic are of great and immediate importance.

Eco-driving, firstly proposed by Finnish Government at the end of the 20th century (Lin 2011), aims at changing driving behavior and vehicle maintenance practices to improve fuel efficiency and reduce greenhouse (GHG) emissions of existing vehicles. Eco-driving programs can take on many forms, including providing basic advice on anticipated changes in traffic, exercising smoother acceleration and braking, and conducting proper vehicle maintenance (Martin et al. 2012). Eco-driving is a preferred way of driving that is best suited for modern engine technology, and it has been reported to have the potential to reduce fuel consumption by about 10–20% (Zhu et al. 2005; Martin et al. 2012). Besides less fuel consumption and reduced emissions, eco-driving also offers benefits such as cost savings, greater safety and riding comfort, and less noise pollution (Barth and Boriboonsomsin 2009).

With an increasing awareness of the economic and environmental benefits from eco-driving, more and more developed countries including Japan and Australia, as well as many European union countries and North America, have launched campaigns to educate and train drivers to drive more economically and efficiently (AA Research Foundation 2011; Lin 2011). Eco-driving interventions mainly include static and dynamic types. A static approach aims at urging drivers to apply general eco-driving techniques after learning through brochures, Web sites and other information sources. Dynamic way mainly refers to provide direct eco-driving feedback to drivers through driving assistance system and devices including visual, auditory and haptic modalities (Azzi et al. 2011; Martin et al. 2012). The main types of eco-driving feedback include conventional dashboards, hybrid vehicle dashboards, smart phone applications, offline feedback systems, dedicated aftermarket feedback devices and haptic pedal feedback (Gonder et al. 2011). Specifically, with the fast development of GPS and 3G communication technology and virtual reality technology, vehicle-mounted terminal, smart phone applications and driving simulator are widely used to train drivers’ eco-driving behaviors (Ayyildiza et al. 2017; Sullman et al. 2015; Zhao et al. 2015a, b). Using these eco-driving systems and devices, drivers could know their relative levels of driving with respect to fuel consumption and emissions, and gain tips to improve driving skills (Satou et al. 2010).

There is a quickly growing body of the literature on the effectiveness of static eco-driving intervention. As early as 1987, Syme et al. (1987) evaluated the effects of implementing a television campaign practices on encouraging viewers to conserve petroleum. The results showed that the driver’s beliefs of saving petroleum in the future were not much higher but were significantly affected statistically. Van Mierlo et al. (2004) studied the effects of following Dutch eco-driving tips on fuel consumption and emissions. The result showed that the saving achieved could range from 5 to 25%. In addition, using a questionnaire, Harvey et al. (2013) discussed drivers’ attitudes and motivation for eco-driving behaviors. The results showed that convenience was rated as more important than saving money from fuel efficiency, and the concern of environment was not a high enough priority to increase fuel efficiency.

In a dynamic eco-driving intervention study, based on a real-world driving experiment with a sample of 20 drivers in Southern California, Boriboonsomsin et al. (2011) evaluated how an in-vehicle eco-driving device that provided instantaneous fuel economy feedback affected driving behaviors. The results showed that, on city streets, the fuel economy could be improved by 6% while the improvement was 1% on highways. In a follow-on study, Martin et al. (2013) also evaluated the performance of an aftermarket real-time feedback device on driving behaviors, and this real-time feedback could report instantaneous fuel economy to drivers while driving. However, the results demonstrated that not all drivers might change their behaviors according to the real-time feedback from the equipment.

While there are abundant evidences on the short-term effect of eco-driving training on fuel consumption, the long-term effect is not as clear. Both the studies of Wahlberg (2007) and Zarkadoula et al. (2007) indicated that, compared to the fuel consumption right after taking the eco-driving course, drivers partially regressed back to less environmentally friendly driving behaviors, thus resulting in lower fuel savings after sometime.

However, in China, eco-driving was just introduced in recent years and this fuel-efficient driving concept was not known well by many drivers. According to a recent survey about drivers’ understanding of eco-driving carried out by Beijing Transport Energy and Environment Centre (2013), more than 75% of the drivers did not have a good understanding of what eco-driving is, and around a quarter of the drivers had never heard about eco-driving. In the past several years, there has been limited research devoted to understand its benefits and effectiveness on fuel consumption and emissions reduction. Bai (2011) studied energy-saving technologies based on theoretical calculation of vehicle fuel consumption and analyzed the influences of some unsafe driving behaviors on fuel consumption. To investigate drivers’ misunderstanding of some common energy-saving driving operation methods in China, Zeng et al. (2010) analyzed fuel consumption levels corresponding to different driving operation methods through engine bench tests. Based on the results, some fuel-saving driving methods were proposed. Overall, most previous researches focused only on fuel-saving technologies of engines and the developed fuel-efficient driving methods were not validated because they were mainly based on theoretical calculation. The impacts of different driving behaviors on vehicle emissions were largely ignored. To our knowledge, no systematic and practical rules of fuel-efficient and environment-friendly driving and training course have been developed in China.

In fact, a traditional driver training program always existed in China and anyone has to go through this training before getting his or her license. This traditional training program first presented trainees with static information (EDUCATION) before giving them practical training (COACHING). Specifically, static EDUCATION about the driving regulations, methods, skills and much other relative driving knowledge was the first class every trainee must take. Afterward, dynamic COACHING is conducted and is the reinforcement for EDUCATION. This two-stage driver training program was carried out in all driving schools throughout China. In this current structure of the training program, COACHING is always preceded by EDUCATION. This current practice of driver training is also adopted in our experiment design for eco-driving training in this study.

To promote eco-driving in China, the Chinese Government has started a campaign to promote the understanding of the potential benefits of eco-driving on improving fuel efficiency and reducing emissions, and to develop an effective strategic implementation for the country. Funded by the Beijing Transport Energy and Environment Centre, we have developed an eco-driving program to comprehensively analyze and promote eco-driving in China. The current research focused on examining whether the eco-driving training course can promote more fuel-efficient driving behaviors and boost drivers’ comprehensibility with eco-driving.

Considering the structure of a traditional driver training program, we developed an eco-driving training course with a similar two-phase training process—EDUCATION alone and COACHING after EDUCATION—to analyze and assess the drivers’ behavioral changes and subjective understanding of eco-driving. The main objectives of this study are as follows:

  • Examine the effectiveness of eco-driving training course on driving behavioral changes and on drivers’ comprehensibility with eco-driving;

  • Analyze the effects of two different eco-driving training courses (i.e., EDUCATION alone and COACHING after EDUCATION) targeting different eco-driving behaviors.

We carried out this research using a driving simulator. On the one hand, the driving simulator technology was widely applied in driver training in previous studies and it has been shown to be effective in improving driving skills (e.g., Birrell et al. 2013; Muttart et al. 2014; Neville et al. 2005). On the other hand, we considered the fact that a driver’s behavior is influenced by a wide variety of environmental and roadway factors and it is very difficult to control those factors in an on-road experiment, and more importantly, it is safer to carry out COACHING in a driving simulator than in a real-life driving environment. This study is a pioneer effort on a thorough evaluation of eco-driving and the development of an executable eco-driving training program in China.

2 Methods

2.1 Participants

In the process of participants recruiting and screening, each candidate was asked to drive the simulator at least 30 min. Only the candidate with no unconformable reactions was employed. Based on this screening procedure, participant was much familiar with our driving simulator. This procedure also contributes to avoid the learning or practice effects from drivers’ unfamiliar with driving simulator for a driver training experiment.

After the screening procedure, twenty-two drivers participated in this study. All participants were licensed drivers with an average (AVG) age of 30 [between the age of 22 and 50 years, with a standard deviation (SD) of 9.5 years] and 3–20 years of driving experience (AVG = 7.6 years and SD = 6.1 years). These participants composed of taxi drivers, full-time drivers employed by companies, undergraduate students, staff members and nearby residents of Beijing University of Technology. Among these subjects, ten were female.

While it would have been desirable to have a higher number of participants, it is common for simulator studies to have a small sample size of 20–30 drivers, due to the high resource demands. Examples include: research into the effects of an eco-driving message on driver distraction, also with 22 participants (Rouzikhah et al. 2013); research into speech-based interaction with in-vehicle computers with 24 participants (Lee et al. 2001); a study aiming at examining the effects of arousing attention on distracted driver’s following behavior under uncertainty with 18 participants (Makoto et al. 2014); and an examination of the effects of driver assistance system on driver behavior at traffic light intersections using 12 participants (Rittger et al. 2015).

2.2 Tasks

According to the common rules of eco-driving in developed countries (Lin 2011; Beusen et al. 2009), we defined four eco-driving rules in this experiment, which reflected typical eco-driving behaviors, and these behaviors were also easy to measure in the simulator. Following these four rules, we designed four tasks as follows:

  • START Avoid rapid starts and accelerate to 20 km/h using 5 s;

  • STOP Decelerate smoothly by releasing the accelerator in time while leaving the car in gear;

  • SPEED CHOICE Try to maintain a steady speed of 100 km/h on freeways;

  • NO-IDLING Shut down the engine when stopping time is, or expected to be, more than 10 s.

2.3 Apparatus

This study used a fixed-base driving simulator in the Key Laboratory of Traffic Engineering at Beijing University of Technology (see Fig. 1). The hardware of this simulator consists of a renovated real car, computers, and video and audio equipment. The road scenarios are projected onto three big screens providing a total of 130° of the driver’s visual field. There were two side mirrors and one rear view mirror to show the traffic behind the simulated vehicle. This simulator records the actions of drivers and vehicles, and the parameters that describe vehicle operations are saved 30 times per second. The data of drivers’ actions collected by this simulator mainly include drivers’ braking, accelerating and turning of the steering wheel, while the data of vehicle operations recorded include the engine revolution as well as speed, acceleration and deceleration.

Fig. 1
figure 1

Fixed-base driving simulator in Beijing University of technology

Full size image

Moreover, the validation of this simulator has been conducted in previous studies. Till now, more than 200 drivers in each driving experiment were used to evaluate the validation of this driving simulator through questionnaires. The evaluation items included the realistic feeling of the accelerator, brake and speed perception. The ratings results are shown in Table 1 (1—not real at all to 10—very real). Furthermore, as speed validation is a key issue for a fixed driving simulator, we had also validated the effectiveness of driving simulator on travel speeds in curves, compared to the results of travel speeds in real roads. Si-fang Interchange of the Fourth Ring Road in Beijing was selected as study site. The results indicated that the speed variation trend in the curves was highly similar to the trend in the real road environment (Ding et al. 2014). Thus, this driving simulator appears to be a valuable research tool for driver behavior studies.

Table 1 Ratings of subjective evaluation of the driving simulator
Full size table

2.4 Driving scenarios

The scenarios used in this experiment mainly include the freeways and urban roads. Corresponding to the four tasks of eco-driving listed above, four different events (START, STOP, SPEED CHOICE and NO-IDLING) are designed and controlled through the scenarios, as shown in Fig. 2. For every drive by each participant, the driver experienced three times of STOP–NO-IDLING–START (with different stopping duration) on urban roads and three times of SPEED CHOICE on freeways. The stopping events were uniformly designed in three signal intersections with digital display. Corresponding to the designed stopping time (see Fig. 2), the red time of traffic signal when drivers arrived at intersections was controlled as 30, 15 and 60 s, respectively. During the red time for a given direction at signal intersections, traffic in this direction is prohibited to move. Between each controlled events, no other events were designed and participants were asked to drive routinely and obey the traffic laws. The scenario and the events are constant for each driver’s every test.

Fig. 2
figure 2

Test events designed in the experiment

Full size image

In order to eliminate the interference of confounding factors on the experimental data, the scenarios used for the same event in different locations are uniformed. When driving on the freeways, the other vehicle kept such a distance from the participant’s vehicle so that the participant never had to pass another vehicle because of speed difference; while driving on the urban roads, no other vehicles were visible in the participant’s driving direction and no vehicles existed in front of the target vehicle when arriving at intersections, but there were vehicles from the opposing direction.

2.5 Experimental designs and procedures

To complete the whole simulated driving task, each participant was asked to complete the experiment sessions three times, including a NORMAL test, a test after EDUCATION and a test after COACHING. According to the test order described above, the effect of COACHING was actually the combined result of EDUCATION plus COACHING. The main reason for the experimental procedure designed in this study was the inherent pattern of driver training in China, which has been discussed in Introduction. In order to distinguish conveniently and express concisely, we defined EDUCATION and COACHING as two different types of eco-driving training courses through the rest of the paper.

In addition, as EDUCATION was less expensive and easier to implement than COACHING, another aim for the experimental design was to find which driving behaviors would be improved fuel efficiency only through EDUCATION, and whether COACHING would bring additional benefits for the same driving behaviors already improved by EDUCATION, or whether COACHING would provide new advantages for other different driving behaviors.

In every test, a participant was asked to complete the driving task in the simulator and then fill out a questionnaire about eco-driving. The procedure of the experiment was as follows:

  • NORMAL

In the first test, participants were asked to complete experiment using their normal way of driving, and there was no knowledge or training on eco-driving.

  • EDUCATION

In the second test, participants completed experimental task after receiving the eco-driving training through EDUCATION. During EDUCATION, drivers were provided static eco-driving information through a brochure and power point presentations. No feedback about eco-driving was given and drivers did not practice the eco-driving rules in the simulator.

  • COACHING

In the last test, before completing experimental task, all participants received the eco-driving training by COACHING. Different from EDUCATION, drivers could receive feedback about their eco-driving level evaluated by a driving instructor in the process of COACHING. Participants were given the information about how to apply eco-driving under the actual guidance and demonstration of a driving instructor. In addition, each participant was asked to practice the main rules of eco-driving in the simulator.

For each driving test, every participant took 3–5 min for a practice drive to become familiar with the simulator’s steering and braking dynamics before every driving experiment. The experiment session of each simulated driving took about 35–40 min to complete under the guidance of the laboratory technician. The command was sent to the participants through the simulator’s audio system. After the driving task, participants were asked to fill out a questionnaire to test their comprehensibility of eco-driving. To further reduce the learning or practicing effect on experimental data, the interval between two adjacent driving experiments was about 3 days. According to general forgetting curve (Zhang 2014), memory might be weakened effectively only more than 2 days. Combing with the time and resource cost of a driving simulator study, 3 days were selected as the time interval. No additional instructions were given to participants during this interval.

3 Results and analysis

3.1 Analysis indexes

Corresponding to the four tasks of eco-driving in our study, four types of driving behaviors were analyzed. In addition, the scores of eco-driving questionnaire (ECO-DRIVING SCORE) were also calculated. For driving behaviors of START, STOP and SPEED CHOICE, both the AVG and SD of these variables before and after eco-driving training were analyzed. The AVG is used to reflect whether the driving behavior is approached to the target value, while the SD represents whether the driving behavior is consistent. The detailed analysis indexes were described as follows:

  • START The AVG and SD of the acceleration in the process of starting were analyzed. The process of starting was defined as the vehicle speed continuously increases from zero to the moment when the diver shifted to the second gear.

  • STOP The AVG and SD of the deceleration in the process of stopping were analyzed. The process of stopping was defined from the first braking point to the point when speed is reduced to zero.

  • SPEED CHOICE The AVG and SD of speed on the freeway were analyzed. For each event, the speeds are collected in the same segment of the road every minute.

  • NO-IDLING The percentage of NO-IDLING when stopping for a longer time was calculated. NO-IDLING refers to shut down engine. The engine revolution data of the vehicle were used to indicate whether participants shut down the engine when stopping for a longer time.

  • ECO-DRIVING SCORE The scores of the questionnaire about eco-driving information were used to reflect drivers’ subjective comprehensibility with eco-driving training course. The questionnaire had some questions about how to drive fuel-efficiently. The question types included multiple-choice, filling the blank and true or false statement. The total score was 100. Due to the consideration of paper length, one example question for each of these three question types is provided below for illustration purpose:

Question 1 (multiple-choice):

For energy conservation and emissions reduction, you should shut down the engine when the stopping time is more than _.

□ 10 s

□ 30 s

□ 1 min

□ 3 min

Question 2 (filling the blank):

When traveling on the freeway, you should maintain the speed at about ____ (km/h) for saving energy.

Question 3 (true or false statement)

Braking rapidly when decelerating contributes to emissions reduction. □ True □ False

3.2 Behavioral changes

3.2.1 START

According to the first rule of eco-driving behavior, which is to accelerate to 20 km/h using 5 s (average acceleration of 1.11 m/s2 assuming a uniform acceleration process), an AVG of 1.11 m/s2 with a small SD of acceleration was defined as the optimal START in this study. Figure 3 displays the AVG and SD of the acceleration in START in each driving experiment. It was clear that both the AVG and SD of the acceleration after each eco-driving training course were reduced (compared to NORMAL), and the AVG was closer to the target value of 1.11 m/s2. The absolute difference between the AVG and the target value in NORMAL, after EDUCATION and after COACHING was 0.15, 0.11 and 0.08 m/s2, respectively.

Fig. 3
figure 3

Effectiveness of eco-driving training on the acceleration in START. a AVG of acceleration in START. b SD of acceleration in START

Full size image

Then, the method of analysis of variance (ANOVA) was used to test the statistical significance of eco-driving training. The test results indicated that, at 95% confidence level, the AVG of acceleration was not significantly different (F (2, 42) = 3.008; p = 0.057), but the difference of the SD of acceleration was statistically significant (F (2,42) = 3.801; p = 0.028). Furthermore, Bonferroni test was applied for post hoc multiple comparisons to find the differences of the SD of accelerations among these three tests. At 95% confidence level, the result showed a significant difference between the acceleration in NORMAL and after EDUCATION (p = 0.044), and between the acceleration in NORMAL and after COACHING (p = 0.026). No significant difference was found between EDUCATION and COACHING for the SD of acceleration (p = 1.0).

The above findings imply that eco-driving training course on increasing the stability of acceleration was obviously. In addition, although the AVG of acceleration was not statistically different among these three experiment tests, EDUCATION did help drivers to learn to avoid rapid starts, and COACHING made the AVG acceleration slightly closer to the target value than EDUCATION alone. The difference in effect between EDUCATION and COACHING on START was limited, as COACHING did not bring much more benefits to urge drivers to form a more fuel-efficient START pattern.

3.2.2 STOP

One rule of eco-driving behavior defined in this study was to decelerate smoothly by releasing the accelerator in time and leave the car in gear. Thus, we reason that the smaller AVG and SD of deceleration are more fuel-efficient in STOP. In Fig. 4, it seems that both the AVG and SD of the deceleration in STOP were reduced because of eco-driving training, and this effect was more significant after COACHING.

Fig. 4
figure 4

Effectiveness of eco-driving course on the deceleration in STOP. a AVG of deceleration in STOP. b SD of deceleration in STOP

Full size image

Similarly, the method of ANOVA was also used to test the statistical significance of deceleration here. The test results showed that both the AVG and SD of deceleration were obviously different among these three experimental tests (for AVG, F (2,42) = 5.874; p = 0.005; and for SD, F (2,42) = 6.017; p = 0.004). Then, Bonferroni t-test was again performed for post hoc multiple comparisons to compare the different influence of these two training courses on the deceleration. The result showed a significant difference on both the AVG and SD of deceleration between driving in NORMAL and after COACHING (p = 0.01 for AVG and p = 0.003 for SD), and between the driving after EDUCATION and after COACHING (p = 0.016 for AVG and p = 0.046 for SD). No significant difference between driving in NORMAL and driving after EDUCATION was found (p = 1.0 for AVG and p = 0.552 for SD).

Hence, the analyses suggested that the effect of having just EDUCATION to achieve more fuel-efficient stopping behavior was rather small. The effectiveness of COACHING after EDUCATION was distinctly enhanced. This finding might imply the limited impact of EDUCATION alone on influencing some driving behaviors. After all, EDUCATION might not provide clear and practical guidelines for drivers to follow and exercise, and guided practices may be needed to decelerate a vehicle smoothly.

3.2.3 SPEED CHOICE

Maintaining a steady speed at about 100 km/h when driving on freeways was another rule of eco-driving behavior defined in this study. Therefore, an AVG of driver speed closer to 100 km/h and with a smaller SD was considered to be more fuel-efficient driving behavior. Figure 5 displays the AVG and SD of driving speeds on freeways in each driving experiment. It was obvious that, comparing these three test results, the AVG was the closest to the target value and its SD was also the smallest after COACHING. The AVG virtually did not change or even went slightly further away from the target value when only EDUCATION was applied, as compared with the NORMAL group.

Fig. 5
figure 5

Effectiveness of eco-driving course on the speed in SPEED CHOICE. a AVG of speed in SPEED CHOICE. b SD of speed in SPEED CHOICE

Full size image

According to the results of the method of ANOVA, the effectiveness of eco-driving training was statistically significant for both the AVG (F (2,42) = 3.240; p = 0.046) and SD (F (2,42) = 4.831; p = 0.011) of speed. Bonferroni t-test was also used here for post hoc multiple comparisons to explore the effects of EDUCATION alone and COACHING on both the AVG and SD of speed on freeways. For the AVG, the result of Bonferroni t-test showed a statistical difference between the speed in NORMAL and after COACHING (p = 0.074), and between the speed after EDUCATION and COACHING (p = 0.04). No significant difference between NORMAL and EDUCATION was found (p = 1.0). For the SD, the result of Bonferroni t-test showed a significant difference between NORMAL and after EDUCTION (p = 0.025), and between NORMAL and after COACHING (p = 0.03). No significant difference between EDUCATION and COACHING was found (p = 1.0).

The significant difference between NORMAL and COACHING for both the AVG and SD of speed demonstrated that COACHING could promote maintaining a steady speed when driving on freeways and make drivers to select a most fuel-efficient speed. For EDUCATION, only SD of speed between NORMAL and EDUCATION was significantly different. EDUCATION alone made drivers to maintain their speed at a steady level but failed to approach the target value. These findings again indicated that EDUCATION alone was not enough to change some driver behaviors toward more fuel-efficient driving.

3.2.4 NO-IDLING

Vehicles in an idling state not only waste fuel but certainly affect the environment and public health. One rule of eco-driving states that the engine should be turned off when stopping for a longer time. In this section, the numbers 0 and 1 were used to indicate the engine turned off and not turned off, respectively. Thus, the proportion of drivers to turn off the engine could be calculated to analyze the effectiveness of eco-driving training on vehicle NO-IDLING. The larger proportion of drivers to turn off the engine when stopping for a longer time the more effective the eco-driving training is. The percentages of NO-IDLING when stopping for 15, 30 and 60 s in NORMAL, after EDUCATION and after COACHING were calculated, respectively (see Fig. 6).

Fig. 6
figure 6

Percentages of NO-IDLING when stopping for longer time in each driving experiment

Full size image

Figure 6 clearly shows that, on the one hand, the percentages of NO-IDLING after eco-driving training were markedly higher than their counterparts in NORMAL, regardless of the stopping time. What is more, the percentages of NO-IDLING after COACHING were a little higher than when only EDUCATION is applied. On the other hand, after eco-driving training, the percentage of NO-IDLING was the highest when stopping for 60 s and the lowest for only 15 s. Before receiving eco-driving information, the percentage of NO-IDLING was also the highest when the stopping time was the longest. These results are rather intuitive.

As the independent variable was discontinuous, the method of nonparametric equivalents of the ANOVA was analyzed statistically. In this section, Friedman test was used to explore the effectiveness of eco-driving training course on NO-IDLING. The results indicated that eco-driving for promoting NO-IDLING was statistically significant for each stopping time (χ 2 = 14.286, 19.857 and 20.462 when stopping 15, 30 and 60 s; p < 0.001 in all cases). According to the mean ranks of Friedman test, the effect of COACHING on reducing vehicle idling was slightly better than EDUCATION alone only when stopping time was 30 and 60 s. For stopping 15 s, the mean rank was 1.47 in NORMAL, while both of them were 2.26 in EDUCATION and COACHING. For stopping 30 s, the mean rank was 1.36 in NORMAL, 2.19 in EDUCATION and 2.44 in COACHING. For stopping 60 s, the mean rank was 1.43, 2.25 and 2.33, respectively.

Overall, the evident increasing of NO-IDLING after training indicated that EDUCATION alone could significantly reduce the percentages of vehicle idling, and COACHING was more effective for drivers to shut down the engine when parking for a longer time. In addition, these results also demonstrated that stopping time was an important factor considered by drivers when determining whether to shut down the engine.

3.3 ECO-DRIVING SCORE

The ECO-DRIVING SCORE was the result of eco-driving questionnaire, which reflected drivers’ subjective comprehensibility with eco-driving information. In this section, participants’ AVG score of eco-driving questionnaire after each driving task was calculated, as shown in Fig. 7.

Fig. 7
figure 7

Average score of eco-driving questionnaire after each driving task

Full size image

It is apparent that the participants’ AVG score was very low in NORMAL, and the AVG score was markedly improved after exposure to the eco-driving training course. Additionally, it seems that the effectiveness between EDUCATION alone and COACHING on understanding the eco-driving information was nearly the same. The results of ANOVA verified that the effectiveness of eco-driving training on drivers’ eco-driving score was significant (F (2,42) = 38.762; p < 0.001). Furthermore, Bonferroni t-test of post hoc multiple comparisons also stated that the average score was significantly different between NORMAL and EDUCATION (p < 0.001), and between NORMAL and COACHING (p < 0.001), but not statistically different between EDUCATION and COACHING (p = 1.0).

These lower score in NORMAL indicated that drivers’ eco-driving knowledge before getting eco-driving training was every lacking. This conclusion might be a highly relevant explanation for drivers’ poor fuel-efficient driving behaviors before receiving eco-driving course. More importantly, combing drivers’ behavioral changes and subjective understanding improvements after eco-driving training, it is suggested that the effectiveness of EDUCATION on improving drivers’ eco-driving comprehensibility was acceptable. But the effect of EDUCATION alone on changing some driving behaviors was very limited, especially for these behaviors that are difficult to operate to achieve more fuel-efficient driving pattern (e.g., STOP and SPEED CHOICE). Drivers just realized that they should drive more fuel-efficiently after EDUCATION, but EDUCATION did not give them clear guidance for proper action for certain driving tasks.

4 Discussion

This study focused on investigating the effectiveness of a recently developed eco-driving training course, using a driving simulator with a sample of twenty-two drivers. Through the whole eco-driving training course, only two participants’ subjective comprehensibility and driving behavior were almost unchanged. More than 90% of drivers improved their comprehensibility and driving skills and conformed closer to the guidance of eco-driving with varying degrees, after receiving the eco-driving training course. These results provided good evidences of the effectiveness of the eco-driving training course in this study.

The main findings of this study was that drivers’ eco-driving behavior and comprehensibility improvement was quantitatively analyzed after eco-driving training, which was different with many other study results aiming at finding the effectiveness of eco-driving on fuel consumption and emissions reduction (e.g., Zhu et al. 2005; Martin et al. 2012; Van Mierlo et al. 2004). The current study results laid a foundation for exploring the relationship between driving behavior improvement and saving vehicle fuel consumption and reducing emissions, and also contributing to explicit the influence of eco-driving on driving safety and efficiency from the aspect of driving behavior change.

For eco-driving rules defined in this study, we mainly referenced the guidance of eco-driving in Japan (Lin 2011) and in America (Beusen et al. 2009). For measuring easily and analyzing quantitatively, four types of driving behaviors in the process of driving (i.e., START, STOP, SPEED CHOICE and NO-IDLING) were discussed. As drivers’ attitudes and motivation were much important for implementing eco-driving (Harvey et al. 2013), the ECO-DRIVING SCORE was also used to reflect drivers’ comprehensibility with eco-driving by questionnaires.

Similarly to the typical driver training methods existed (AA Research Foundation 2011), both static and dynamic eco-driving training patterns were introduced in the current study. The static approach (i.e., EDUCATION) was applied through a brochure and power point presentations, while the dynamic way (i.e., COACHING) was performed in driving simulator under the actual guidance and demonstration of a driving instructor. In this study, driving simulator was selected as a key factor influencing both the effectiveness of eco-driving training and driving behavior collecting. Based on our current study results, it was tested and verified that driving simulator was an effective approach for driving behaviors representation and eco-driving training, which was consistent to our previous study results (e.g., Zhao et al. 2015a, b) and researches aboard (e.g., Rouzikhah et al. 2013; Sullman et al. 2015).

In this study, the effectiveness of eco-driving EDUCATION and COACHING after EDUCATION was analyzed. The more obvious benefit of COACHING in improving fuel-efficient driving was likely the cooperative effect from EDUCATION. Although the effectiveness of EDUCATION on changing some driving behaviors was limited, it did offer a concept of eco-driving for drivers, and thus, drivers tend to drive more fuel-efficiently even before receiving COACHING. Furthermore, as EDUCATION was easier and less expensive to implement than COACHING, we also wanted to further explore whether EDUCATION alone could transform some driver behaviors to meet the requirements of eco-driving. Our study results indicated that EDUCATION was effective to improve drivers’ comprehensibility with eco-driving, reduce the percentages of vehicle idling and help drivers to avoid rapid starts. EDUCATION alone, however, did not significantly improve driver’s fuel-efficient driving abilities in the aspect of STOP and SPEED CHOICE. Besides, COACHING could bring more benefits to improve driver’s fuel-efficient driving abilities in STOP, SPEED CHOICE and NO-IDLING. These study results also implicated that the effectiveness of eco-driving training on various driving behaviors was dependent on both training methods (e.g., levels of practicing provided) and the driving behavior itself (e.g., levels of practicing required).

Additionally, considering that the learning effect might be existed in a driving simulator study, besides the adaption test in participants screening procedure, two measurements were also implemented in this study. Firstly, every participant was asked to take 3–5 min for a practice drive to become familiar with the simulator’s performance before every driving experiment (e.g., normal test, test after EDUCATION and test after COACHING). This measurement was used to make drivers’ familiarity with driving simulator before experiment test to be as same as possible. Secondly, the interval between two adjacent driving experiments was set as 3 days. Based on these controlling methods, the learning effect on experimental data was effectively decreased.

5 Summary and conclusions

In this study, we tested the effectiveness of eco-driving training (first EDUCATION and then COACHING) on four typical driving behaviors (i.e., START, STOP, SPEED CHOICE and NO-IDLING) and drivers’ comprehensibility with eco-driving information. The following findings were obtained in this research.

  1. 1.

    EDUCATION could promote a better comprehensibility with eco-driving for drivers, and additional benefits from COACHING on improving drivers’ comprehensibility with eco-driving may be limited.

  2. 2.

    The function of EDUCATION alone and COACHING (EDUCATION followed by COACHING) was not always equally effective for different driving behaviors:

    • START EDUCATION could help drivers avoid rapid starts slightly, and COACHING did not introduce much additional benefit for more fuel-efficient starting pattern compared to EDUCATION alone.

    • STOP COACHING was helpful for drivers to decelerate smoothly, but EDUCATION alone did not appear to be effective.

    • NO-IDLING EDUCATION alone could promote drivers to turn off the engine when stopping for an extended period of time, and COACHING further enhanced this effectiveness slightly.

    • SPEED CHOICE EDUCATION alone did not help to maintain the speed at the most fuel-efficient value when driving on freeways, but COACHING worked.

In conclusion, the developed eco-driving training course did have a positive effect on promoting drivers to drive more fuel-efficiently and also helped drivers to better understand the concept and master the practices of eco-driving. In addition, eco-driving training led to more consistent eco-driving driving behavior as evidenced by smaller standard deviation values on all measures related to fuel efficiency. The findings of this study can be used as a foundation for governments to evaluate the potentials of eco-driving, and develop and implement strategies to maximize the benefits of eco-driving.

6 Limitation and future research needs

All conclusions in this study were based on driving simulator experiments. The limitation of simulator experiments should not have affected the relative validity of these results; nevertheless, all findings and results should be tested through experiments in a real driving environment in the future. To assess eco-driving comprehensively, surveys of eco-driving attitudes and on-road assessments should be conducted in future. Specifically, with the fast development of GPS and 3G communication, vehicle-mounted terminal and smart phone applications with real-time feedback would be beneficial to apply eco-driving training in the field.

In the current study, we only focused on four rules of eco-driving behaviors and we did not consider different vehicle types and performance. In future studies, more comprehensive eco-driving practices should be evaluated based on driving characteristics, traffic features and vehicle types to develop more comprehensive eco-driving strategies and training programs. Besides, scores of the questionnaire about eco-driving information were used to reflect drivers’ subjective comprehensibility with eco-driving training course, further efforts will be applied to explore drivers’ comprehensibility with eco-driving more comprehensively and accurately, and the relationship between driver comprehensibility and driving behavior will be established to explore the mechanism of eco-driving training on fuel-efficient driving behaviors improvement.

Additionally, each participant was designed to complete the performance task three times in the current study. The study results might be influenced by the practice effects from the repeated measurements for the same driver. So a control group would be set to add significant value to the results in the following research. Besides, the driving skills and habits between professional and non-professional drivers might be not always the same, and these differences might influence the effectiveness of eco-driving training course on different driver groups. The current study results did not distinguish the effectiveness of eco-driving training between professional and non-professional drivers, and we have planned to discuss the different driving habits and eco-driving behavior improving potential between these two driver groups with a bigger data sample sizes for each. In addition, just as stated in Introduction, drivers might regress back to less environmentally friendly driving behaviors; after sometime, it is necessary to further explore the long-term effectiveness of eco-driving training in future.

In this study, we only examined the behavioral changes and the subjective understanding of the improvements in eco-driving after the eco-driving training course, the changes in fuel consumption and GHG emissions were not evaluated explicitly. Therefore, an evaluation of the potential of eco-driving on improving fuel efficiency and reducing GHG emissions using air quality models such as MOVES developed by the EPA should be beneficial (EPA 2011).