Development of a Real-World Eco-Driving Cycle for Motorcycles

Abstract

Climate change is a major issue all around the world. The transportation industry currently accounts for most CO₂ emissions. The goal of this research is to develop a real-world eco-driving cycle for internal combustion engine motorcycles that can reduce fuel consumption and CO₂ emissions. This study developed onboard measuring equipment to measure the speed profile and fuel consumption of a motorcycle driving in real time. A total of 78 motorcycle riders rode a test motorcycle with the onboard equipment along a road network to collect real-world data. All of the collected real-world data were analyzed by cluster analysis based on fuel consumption (km/L) to divide riders into two groups, high-fuel-consumption riders and low-fuel-consumption riders. The collected real-world data of the low-fuel-consumption riders were used to develop a real-world eco-driving cycle, whereas the collected real-world data from the high-fuel-consumption riders were used to develop a real-world non-eco-driving cycle. The CO₂ emissions were calculated by the speed profiles of the developed driving cycles. The findings reveal that the real-world eco-driving cycle provided a fuel consumption rate 39.3% lower than the real-world non-eco-driving cycle. In addition, the real-world eco-driving cycle provided a CO₂ emission rate 17.4% lower than the real-world non-eco-driving cycle. The application of the developed real-world eco-driving cycle for motorcycles is proposed.

1. Introduction

Global greenhouse gas (GHG) emissions reached a new high in 2019. Although 2020 GHG emissions were lower than those in 2019 due to the COVID-19 crisis and associated actions, GHG concentrations in the atmosphere are continuing to rise. CO₂ emissions account for 65% of the total greenhouse gas emissions, resulting in increased GHG emissions. To reduce global warming, net-zero CO₂ emission reductions must be sustained. The transport sector has contributed to roughly 14% of global GHG emissions during the last decade [1]. The road transport sector is principally responsible because its growth is increasing, particularly in Asia due to rapid economic growth. More than half of global CO₂ emissions are emitted in Asia [2]. Therefore, the development of transportation must be based on environmental sustainability.

Hayashi et al. (2012) [3] suggested the CUTE matrix for establishing a low carbon society. The CUTE matrix introduced three strategies, including AVOID, SHIFT, and IMPROVE, to minimize fuel consumption and emissions in the transportation sector. Four measurements, including technology, regulatory, information, and economics, were also introduced. Fukuda et al. (2014) [4] proposed three strategies for establishing a low-carbon society in Asia according to the CUTE matrix. They were AVOID (e.g., transit-oriented development, TOD), SHIFT (e.g., shift to public transit) and IMPROVE (e.g., improving driving behavior using eco-driving cycles).

Among ASEAN countries, Indonesia, Vietnam, and Thailand have the highest accumulated number of motorcycles, as displayed in Figure 1 [5]. The number of newly registered motorcycles in Thailand has risen to 21.4 million [6], leading to an increase in fuel consumption and CO₂ emissions.

An improvement in on-road driving behavior that would reduce fuel consumption would benefit a huge number of motorcycle riders, resulting in significant reductions in fuel consumption and CO₂ emissions. The term “eco-driving” refers to actions that the driver can perform while driving to improve fuel efficiency or reduce CO₂ emissions. These actions include reducing deceleration, avoiding frequent use of the brakes, maintaining a suitable distance, using proper acceleration, maintaining a constant speed, and reducing the idling time [7]. Many studies have revealed that eco-driving training in passenger car drivers could improve fuel efficiency and reduce CO₂ emissions [8,9], but eco-driving training for motorcycle riders does not yet exist. Previously, eco-driving cycles for motorcycles have been developed by an optimal controller in laboratory tests [10,11]. However, these engine-controlled eco-driving cycles may not ensure that motorcycle riders can be effectively trained for eco-driving behavior because they were not developed using representative data from the real-world driving behavior of motorcycle riders. In addition, fuel consumption and air pollutant emissions of real-world driving were reported differently from those of laboratory-tested driving cycles because drivers’ behavior and traffic and road conditions were not considered. Therefore, it is essential to aggregate sufficient realistic data to generate representative real-world driving cycles that can be employed reliably for fuel consumption and exhaust emission assessment in the future [12]. Nevertheless, the previous study did not reveal the development of a real-world eco-driving cycle for motorcycles, which is necessary for reducing fuel consumption and CO₂ emissions.

The objective of this study was to develop a real-world eco-driving cycle for motorcycles that can reduce fuel consumption and CO₂ emissions. An onboard measurement device was developed and installed on the test motorcycle to collect real-world driving data. The rest of this paper is structured as follows. Section 2 discusses the relevant literature. Section 3 explains the research methodology. Section 4 presents the results and discussions. The conclusions and recommendations are presented in Section 5.

2. Literature Review

There have been many previous studies on real-world driving patterns, as indicated in

Table 1. Previous studies on real-world driving patterns.

Previous Studies	Data Collection	Analysis Method	Measurement and Calculation of Fuel Consumption and CO2 Emissions	Results
Tzeng and Chen (1998) [14]	Chasing vehicle technique	Statistical method for determining the driving cycle	Chassis dynamometer test	The driving cycles for motorcycles in Taipei, Taiwan
Chen et al. (2003) [21]	Onboard measurement	Repetitive algorithm for selecting micro-trips at random	Chassis dynamometer test	The driving cycles for motorcycles for cities in Taiwan
Tsai et al. (2005) [22]	Onboard measurement	Repetitive algorithm for selecting micro-trips at random	Chassis dynamometer test	The driving cycles for motorcycles in Kaohsiung, Taiwan
Tong et al. (2011) [23]	Onboard measurement	Repetitive algorithm for selecting micro-trips at random	Not considered	The driving cycles for motorcycles and light-duty vehicles in Vietnam
Seedam et al. (2015) [24]	Onboard measurement	Repetitive algorithm using principle of least total variance in target parameters from micro-trips	Not considered	The driving cycles for motorcycles in Khon Kaen city, Thailand
Tutuianu et al. (2015) [25]	Onboard measurement	Repetitive algorithm using principle of least total variance in target parameters from short trips	Not considered	The driving cycles for light-duty vehicles
Satiennam et al. (2017) [29]	Onboard measurement	Linear regression analysis	Onboard fuel consumption sensor and gas analyzer	Real-world exhaust emission and fuel consumption models for motorcycles
Mayakuntla and Verma (2018) [15]	GPS	Repetitive algorithm using target parameters from trip segment	Not considered	The driving cycles for passenger cars in Indian cities
Wang et al. (2018) [9]	Onboard measurement	Descriptive statistics	Calculation of fuel consumption and emissions	Eco-driving training efficiency according to road type
Koossalapeerom et al. (2019) [26]	Onboard measurement	Repetitive algorithm using principle of least total variance in target parameters from micro-trips	Onboard fuel consumption sensors and gas analyzer Calculation of CO2 equivalent emissions of electric motorcycle	The driving cycles for electric and gasoline motorcycles
Mahesh et al. (2019) [16]	GPS	Emission rate equations	Onboard gas analyzer	Real-world emission factors and emission models for motorcycles in India
Zhang et al. (2019) [17]	GPS	Micro-trip method and Markov Monte Carlo method	Onboard energy consumption sensor	The driving cycles for electric vehicles considering road environment
Lois et al. (2019) [30]	Onboard measurement	Multivariate analysis	Onboard energy consumption sensor	Eco-driving affected by driving behavior and fuel consumption influenced by congestion and road slope
Ma et al. (2019) [12]	GPS	Markov chain method	Calculation of fuel consumption	The driving cycles for large-sized vehicles
Desineedi et al. (2020) [18]	GPS	K-means clusters and Markov modeling method	Not considered	The driving cycles for buses in Chennai, India
Zhao et al. (2020) [13]	Chasing vehicle technique and onboard measurement	Markov Monte Carlo simulation method	Onboard energy consumption sensor and emissions not considered	The driving cycles for electric vehicles in Xi’an, China
Coloma et al. (2020) [8]	GPS	Multivariate data analysis	Calculation of fuel consumption and emissions	Eco-driving efficiency depending on city and road section
Liu et al. (2021) [19]	GPS	Combination of clustering and Markov chain algorithm	Onboard energy consumption sensor and emissions not considered	The driving cycles for plug-in hybrid electric vehicles
Ghaffarpasand et al. (2021) [20]	GPS	Repetitive algorithm using principle of least total variance in target parameters	Onboard measuring gas analyzer	The driving cycles for motorcycles in Isfahan

. Many driving cycles have been developed according to vehicle type, road hierarchy, and the urban environment. In general, the development of a driving cycle can be divided into three steps: (1) route selection, (2) real-world data collection, and (3) driving cycle construction.

The real-world driving cycle data were collected using three methods: (1) the chasing vehicle technique—measuring the speed of the targeted vehicle, e.g., [13,14], (2) data collection using a global positioning system (GPS), e.g., [8,15,16,17,18,19,20], and (3) onboard measurements by installing a measuring device on a test vehicle, e.g., [21,22,23,24,25,26]. However, it was difficult for the driver to carry out the chasing in the circumstance of mixed and congested traffic. Furthermore, the GPS was unable to accurately measure the speed profile. On the other hand, the onboard measurement with a measuring device installed on a test vehicle was more effective.

Previously, the micro-trip method was commonly utilized to construct driving cycles, for example, [12,17,27,28]. This method divided the raw on-road driving data into many micro-trips by successive stops. The algorithm randomly and repeatedly selected a micro-trip and connected it to a previous micro-trip (if one existed) to construct a combination of micro-trips until the length of the combination of micro-trips was close to the predefined duration of the driving cycle. Recently, Mayakuntla and Verma [15] proposed an alternative trip segment method for constructing a driving cycle. This method divided raw on-road driving data into trip segments, which were more homogenous than micro-trips. This method provided a more specific indication of the driving cycle in mixed and congested traffic. However, the construction of the driving cycle requires a more complex algorithm.

Many previous studies, such as [14,21,22], used the chassis dynamometer test to measure the fuel consumption and emissions of driving cycles. In other studies, such as [8,26], equations have been used to calculate the fuel consumption and emissions. Many studies have recently developed and installed onboard sensors to measure the real-world fuel consumption and emissions, such as [13,16,17,19,20,26,29].

The eco-driving cycle was defined as a driving characteristic that consumes less gasoline fuel and emits fewer emissions. To the best of the authors’ knowledge, there is a lack of research on the development of a real-world eco-driving cycle for motorcycles. A few studies [11,31,32], have attempted to develop eco-driving cycles for two-wheel and four-wheel vehicles using mathematical models and algorithms. Coloma et al. [8] developed a real-world eco-driving cycle for passenger cars by comparing the behavior of drivers before and after an eco-driving training course. Nevertheless, no study has been done on a real-world eco-driving cycle for motorcycles.

3. Research Methodology

3.1. Development of Onboard Measurement Device

This study developed an onboard measurement device based on the authors’ previous studies [24,26,29,33], to collect the real-world speed profile and fuel consumption. It was installed on the test motorcycle so that it could collect real-world data, including the speed profile and fuel consumption, while riding on a road network in real time. The onboard measurement device was designed to include four components: (1) a data logger for processing and recording the collected data, (2) a GPS sensor for locating the test motorcycle, (3) a Hall effect sensor for measuring the instant speed, and (4) a fluid flow measuring sensor to measure the gasoline flow rate. This data logger and sensors were installed on the test motorcycle, as shown in Figure 2. The instant speed of the test motorcycle was measured by detecting the rotation of its front wheel using the Hall effect sensor. A total of 22 metal pieces installed on the motorcycle’s disc were rotated through the magnetic probe while the motorcycle was riding to measure the test motorcycle’s speed profile. The fluid flow measuring sensor was installed at the fuel tube connected to the carburetor that suctioned the fuel into the test motorcycle’s engine chamber. The Sensirion sensor, the SQL-HC60 model, was used in this study since it can measure fluid flow up to 80 mL/s with a precision of 50 mL/s, which was suitable for monitoring the fuel flow rate of the test motorcycle. This study chose a 4-stroke motorcycle with a 113 cc engine capacity as the test motorcycle because this engine size is popular among motorcycle users in Thailand.

3.2. Real-World Data Collection

This study applied the onboard measurement technique, also known as the naturalistic observational field technique [34], to measure the driving behavior of motorcycle riders. The developed measurement device was installed on the test motorcycle for collecting real-world data. This study selected a route around Khon Kaen University, Thailand, for measuring the driving behavior of motorcycle riders, as shown in Figure 3. Its size is comparable to that of a small city. This road network consists of long sections of arterials, roundabouts, and a signalized intersection.

A total of eighty motorcycle riders were chosen for riding the test motorcycle. The riders rode the test motorcycle with the installed onboard measurement device along the test route to collect real-world data. The participating riders were able to ride the test motorcycle at their desired speed. The speed profiles of the riders substantially varied depending on each of their own driving behaviors.

3.3. Analysis of Real-World Data

The collected real-world data were calculated for the driving parameters of each motorcycle rider. Pearson’s chi-squared test was used to analyze correlations between the fuel consumption and driving parameters. Hierarchical cluster analysis was applied to divide motorcycle riders into two groups based on their fuel consumption. The first group consisted of riders who consumed a high amount of gasoline, while the second group consisted of riders who consumed a low amount of gasoline. The t-test was applied to compare the mean of the driving parameters between the two groups at a 0.05 significance level. The real-world data from the high-fuel-consumption riders were used to develop the real-world non-eco-driving cycle, whereas the real-world data from the low-fuel-consumption riders were used to develop the real-world eco-driving cycle.

This study used the micro-trip method to construct the driving cycle as opposed to the trip segment method, which is appropriate for congested road networks, because the participating riders were able to ride the test motorcycle at their desired speed along the selected uncongested route. The driving cycles were constructed using the algorithm from the authors’ previous studies [24,26]. The nine targeted driving parameters were firstly identified, as shown in

Table 2. Targeted driving parameters.

Parameters	Symbol	Definitions
Average overall speed (km/h)	Vavg	Average speed in a cycle, including idling period.
Average running speed (km/h)	V1avg	Average speed in a cycle, excluding idling period.
Average acceleration (m/s2)	Accavg	Rate of speed change ≥ 0.27 m/s2 *
Average deceleration (m/s2)	Dccavg	Rate of speed change ≤ −0.27 m/s2 *
Idling time (%)	Idle	Time proportion at zero speed
Cruising time (%)	Cruise	Time proportion during absolute speed change ≤ 0.27 m/s2 *
Acceleration time (%)	Acc	Time proportion during acceleration ≥ 0.27 m/s2 *
Deceleration time (%)	Dec	Time proportion during deceleration ≤ −0.27 m/s2 *
Positive kinetic energy (m/s2)	PKE	Kinetic energy during positive acceleration

Note: * The threshold of 0.27 m/s² was applied to allow for data fluctuation according to previous studies [12,24,26].

. All of the collected real-world data were used to calculate these targeted parameters. They were utilized as the selection criteria to minimize the differences between the developed driving cycle and all of the collected data. The duration of the driving cycle was defined as approximately 1500 s. The collected real-world data were then divided into micro-trips by successive stops (≥3 s). The algorithm selected a micro-trip at random and connected it to a previous micro-trip (if one existed) to construct a combination of micro-trips. This process repeated until the length of the combination of micro-trips was close to the predefined duration of the driving cycle. The target parameters of the constructed driving cycle were calculated and compared with the target parameters of all of the collected data. The algorithm repeated until it obtained a driving cycle that did not go over the threshold (the average error was not more than 5%). The detailed procedure for the construction of the driving cycle is described in the authors’ previous studies [24,26]. Then, this study used the equation from a previous study [35], to calculate the CO₂ emissions of the developed driving cycles.

4. Results and Discussions

4.1. Validation Results of Developed Onboard Measurement Device

This study validated the distance measurements of the developed onboard measurement device by comparing them to the 1 km reference distance. The validation results of the developed onboard measurement device are presented in

Table 3. Validation results of the developed onboard measurement device.

No.	Reference Distance (m)	Measured Distance (m)	Error (%)
1	1000	1000.48	+0.048
2	1000	999.97	−0.003
3	1000	1000.23	+0.023
4	1000	999.46	−0.054
5	1000	999.72	−0.028
Average	1000	999.97	−0.003

. It was found that the average error of the measured distances was −0.003%. It was concluded that the developed onboard measurement device could accurately measure the distance.

4.2. Results of Real-World Data Collection

The real-world data of 2 out of 80 participating riders were missing due to an error of the onboard measurement device during riding. In total, this study obtained the real-world data of 78 participating riders for analysis. The results of the Pearson’s chi-squared test are presented in

Table 4. Pearson correlations between fuel consumption and driving parameters.

Parameters	Pearson Correlation	Sig.
Rider weight (kg)	−0.012	0.919
Average overall speed (km/h)	−0.431 **	<0.001
Average running speed (km/h)	−0.430 **	<0.001
Acceleration (m/s2)	−0.060	0.599
Deceleration (m/s2)	0.028	0.810
Momentum1 (kg·m)/s	−0.386 **	<0.001
Momentum2 (kg·m)/s	−0.402 **	<0.001
Force1 (kg·m)/s2	−0.053	0.646
Force2 (kg·m)/s2	0.032	0.778
Accelerating time (%)	−0.094	0.414
Decelerating time (%)	−0.223	0.051
Cruising time (%)	0.123	0.283
Idling time (%)	0.146	0.202
PKE (m/s2)	−0.078	0.495

Note: 1. Total weight = rider weight + test motorcycle weight (98.8 kg). 2. Momentum1 = total weight × average overall speed, Momentum2 = total weight × average running speed. 3. Force1 = total weight × acceleration, Force2 = total weight × deceleration. 4. ** Significant at 0.01.

. It was revealed that the average overall speed, average running speed, Momentum1, and Momentum2 were significantly related to fuel consumption at a 0.01 significance level. However, the rider weight did not relate to fuel consumption because the average weight of riders, 63.7 kg, was low when compared to the weight of the test motorcycle, 98.8 kg. Therefore, the average overall speed and average running speed had significant impacts on the fuel consumption.

The results of the hierarchical cluster analysis, by determining that the rescaled distance cluster was equal to 10, revealed that 52 riders (67%) were classified as low-fuel-consumption rivers and 26 riders (33%) as high-fuel-consumption riders.

The results of the parameter comparison between the high- and low-fuel-consumption riders are presented in

Table 5. Parameter comparison between high- and low-fuel-consumption riders.

Parameters	High-Fuel-Consumption Riders		Low-Fuel-Consumption Riders		Diff (%)	p Value
	Mean	SD	Mean	SD
Fuel consumption (km/L)	35.86	4.10	50.29	5.17	14.43 (40.2%)	0.000 *
Average overall speed (km/h)	35.08	4.22	30.52	2.81	−4.56 (−13.0%)	0.000 *
Average running speed (km/h)	36.76	3.82	32.11	2.69	−4.65 (−12.7%)	0.000 *
Acceleration (m/s2)	0.85	0.15	0.86	0.20	0.01 (1.2%)	0.415
Deceleration (m/s2)	0.88	0.14	0.90	0.19	0.02 (2.3%)	0.373
Momentum1 (kg·m)/s	1709.67	184.53	1462.77	145.60	−246.90 (−14.4%)	0.000 *
Momentum2 (kg·m)/s	1631.77	202.33	1389.74	140.73	−242.03 (−14.8%)	0.000 *
Force1 (kg·m)/s2	141.62	24.15	141.55	35.70	−0.07 (−0.1%)	0.276
Force2 (kg·m)/s2	146.91	22.02	147.42	33.86	0.51 (0.4%)	0.237
Accelerating time (%)	30.44	6.42	28.75	5.18	−1.69 (−5.6%)	0.226
Decelerating time (%)	30.20	6.00	28.03	4.85	−2.17 (−7.2%)	0.136
Cruising time (%)	34.40	10.82	38.26	10.53	3.86 (11.2%)	0.183
Idling time (%)	4.81	3.08	4.83	2.77	0.02 (0.4%)	0.778
PKE (m/s2)	0.60	0.16	0.55	0.20	−0.05 (−8.3%)	0.932

Note: 1. Total weight = rider weight + test motorcycle weight. 2. Momentum1 = total weight × average overall speed, Momentum2 = total weight × average running speed. 3. Force1 = total weight × acceleration, Force2 = total weight × deceleration. 4. * Significant at 0.05.

. It was found that the fuel consumption (km/L) of the high-fuel-consumption riders was significantly lower than that of low-fuel-consumption riders. In other words, the high-fuel-consumption riders consumed significantly more gasoline than the low-fuel-consumption riders. The average overall speed and average running speed of the high-fuel-consumption riders were significantly higher than those of the low-fuel-consumption riders. Momentum1 and Momentum2 of the high-fuel-consumption riders were significantly higher than those of the low-fuel-consumption riders.

4.3. Results of Developed Real-World Driving Cycles

The results of the development of the real-world non-eco-driving cycle and real-world eco-driving cycle are presented in

Table 6. Parameters of developed real-world driving cycles.

Parameters	Real-World Non-Eco-Driving Cycle	Real-World Eco-Driving Cycle	Diff.
Average overall speed (km/h)	33.90	30.39	−3.51 (−10.4%)
Average running speed (km/h)	35.72	31.96	−3.76 (−10.5%)
Average acceleration (m/s2)	0.88	0.86	−0.02 (−2.3%)
Average deceleration (m/s2)	−0.93	−0.91	0.02 (−2.2%)
Idling time (%)	4.95	4.92	−0.03 (−0.6%)
Cruising time (%)	32.95	36.73	3.78 (11.5%)
Acceleration time (%)	31.62	29.69	−1.93 (−6.1%)
Deceleration time (%)	30.35	28.66	−1.69 (−5.6%)
PKE (m/s2)	0.61	0.53	−0.08 (−13.1%)

. It was revealed that the absolute average errors of the parameters of the developed real-world non-eco-driving cycle and the real-world eco-driving cycle were 3.2% and 1.9%, respectively, when compared to the parameters of all of the collected real-world data. These values are within a 5% error threshold. It was found that the idling time of the real-world non-eco-driving cycle was similar to that of the real-world eco-driving cycle. This indicates that the riders in both groups spent the same amount of stopping time at the signalized intersection along the test route. This implies that they rode in similar circumstances.

The motorcycle’s real-world eco-driving cycle obtained an average overall speed that was 10.4% lower and an average running speed that was 10.5% lower than those of the non-eco-driving cycle. These reduction values are consistent with the findings of several previous studies. Nonaggressive motorcycle riders obtained an average running speed 12% lower than the aggressive motorcycle riders [33]. Passenger car drivers, after completing the eco-driving course, could drive their vehicles at an average overall speed that was 3% lower on local streets [36], 13% lower on uncongested arterials [8], and 3.5% lower on all road types [9].

The motorcycle’s real-world eco-driving cycle provided less aggressive driving behavior, as evidenced by a 6.1% reduction in the acceleration time and a 5.6% reduction in the deceleration time. This reduction value is in line with the findings of previous studies. The nonaggressive motorcycle riders obtained an acceleration time that was 12% lower and a deceleration time that was 9% lower than the aggressive motorcycle riders [33]. In addition, passenger car drivers, after passing an eco-driving course, drove more smoothly, with 50% less acceleration time and 45% less deceleration time on uncongested arterials [8] and 33% less acceleration time and 44% less deceleration time on all road types [9].

The developed real-world non-eco-driving cycle and real-world eco-driving cycle are displayed in Figure 4. It was revealed that the lengths of the real-world non-eco-driving cycle and real-world eco-driving cycle were 1496 s and 1462 s, respectively. The maximum speeds were 66.4 km/h and 55.9 km/h, respectively. The real-world eco-driving cycle provided a maximum speed 16% lower than the real-world non-eco-driving cycle. In addition, the eco-driving training in passenger cars provided an 8% lower maximum speed [9]. The maximum accelerations were 6.98 m/s² and 4.62 m/s², respectively, and the maximum decelerations were −6.25 m/s² and −4.03 m/s², respectively. The maximum speed, acceleration, and deceleration of the real-world eco-driving cycle were less than those of the real-world non-eco-driving cycle.

4.4. Results of Comparison of Fuel Consumptions

The results of the fuel consumption calculations of the real-world non-eco-driving cycle driving cycle and the real-world eco-driving cycle are presented in

Table 7. Comparison of fuel consumptions.

Parameters	Real-World Non-Eco-Driving Cycle	Real-World Eco-Driving Cycle	Diff.
Fuel consumption (km/L)	33.95	50.44	16.49 (48.6%)
Fuel consumption rate (mL/s)	0.28	0.17	−0.11 (−39.3%)

. It was revealed that the real-world non-eco-driving cycle and the real-world eco-driving cycle consumed 33.95 and 50.44 km/L, respectively. In other words, the real-world non-eco-driving cycle consumed one liter of gasoline to travel 33.95 km, whereas the real-world eco-driving cycle consumed one liter of gasoline to travel 50.44 km. The real-world eco-driving cycle consumed 48.6% less fuel per distance than the real-world non-eco-driving cycle. In addition, the fuel consumption rates of the real-world non-eco-driving cycle and the real-world eco-driving cycle were 0.28 and 0.17 mL/s, respectively. The real-world eco-driving cycle consumed 39.3% less fuel per travel time than the real-world non-eco-driving cycle. This reduction value for the motorcycle is higher than the findings of other previous studies. The nonaggressive motorcycle riders consumed 12% less gasoline than the aggressive motorcycle riders [33]. Driving passenger cars with eco-driving behavior along uncongested arterials could reduce fuel consumption by 16% [8]. Eco-driving training could reduce fuel consumption by 7% in passenger cars [9,30].

The fuel consumption rate profiles of the real-world non-eco-driving cycle and the real-world eco-driving cycle are comparatively displayed in Figure 5. The fuel consumption rate profile of the real-world eco-driving cycle was significantly lower than that of the real-world non-eco-driving cycle.

4.5. Results of Comparison of CO₂ Emissions

The results of the calculation of the CO₂ emissions of the real-world non-eco-driving cycle and the real-world eco-driving cycle are presented in

Table 8. Comparison of CO₂ emissions.

Parameters	Real-World Non-Eco-Driving Cycle	Real-World Eco-Driving Cycle	Diff.
CO2 emissions (g⋅CO2/km)	23.64	22.81	−0.83 (−3.5%)
CO2 emission rate (g⋅CO2/s)	0.23	0.19	−0.04 (−17.4%)

. It was revealed that the CO₂ emissions per distance of the real-world non-eco-driving cycle and the real-world eco-driving cycle were 23.64 and 22.81 g⋅CO₂/km, respectively. The real-world eco-driving cycle emitted 3.5% less CO₂ per distance 3.5% less than that of the real-world non-eco-driving cycle. In addition, the CO₂ emission rates of the real-world non-eco-driving cycle and the real-world eco-driving cycle were 0.23 and 0.19 g⋅CO₂/s, respectively. The real-world eco-driving cycle for the motorcycle emitted 17.4% less CO₂ per travel time than that of the real-world non-eco-driving cycle. These reduction values are in line with the findings of the previous studies. The nonaggressive motorcycle riders emitted 21% fewer CO₂ emissions than the aggressive motorcycle riders [33]. Driving passenger cars with eco-driving behavior along arterial roads in a small city could reduce CO₂ emissions by 13% [8]. Eco-driving training could reduce CO₂ emissions by 6.3% in passenger cars [9].

The CO₂ emission rate profiles of the real-world non-eco-driving cycle and the real-world eco-driving cycle are comparatively displayed in Figure 6. The CO₂ emission rate profile of the real-world eco-driving cycle was substantially lower than that of the real-world non-eco-driving cycle, which is similar to the trend in fuel consumption.

5. Conclusions and Recommendations

The objective of this study was to develop a real-world eco-driving cycle for internal combustion engine motorcycles for reducing fuel consumption and CO₂ emissions. This study developed an onboard measuring device, which was installed in the test motorcycle to measure the speed profile and fuel consumption of the driving motorcycle in real time. A total of 78 motorcycle riders rode the test motorcycle with the onboard equipment along a road network to collect real-world data. Cluster analysis was applied to analyze all of the collected real-world data to divide the riders into two groups: high-fuel-consumption riders and low-fuel-consumption riders. The collected real-world data from the low-fuel-consumption riders were used to develop the real-world eco-driving cycle, whereas the collected real-world data from high-fuel-consumption riders were used to develop the real-world non-eco-driving cycle. The CO₂ emissions of the developed real-world driving cycles were calculated.

The findings show that the real-world eco-driving cycle provided a fuel consumption rate 39.3% lower than the real-world non-eco-driving cycle. The real-world eco-driving cycle achieved a CO₂ emission rate 17.4% lower than the real-world non-eco-driving cycle. The driving characteristics, which influenced reductions in the fuel consumption and CO₂ emissions, included the average speed, average running speed, cruise time, acceleration time, deceleration time, and PKE. The motorcycle riders consumed less gasoline and emitted less CO₂ when they spent longer times at a cruise speed. In contrast, the motorcycle riders consumed more gasoline and emitted more CO₂ when they had longer acceleration and deceleration times. When the PKE, i.e., the average energy required for acceleration, decreased, the motorcycle consumed less gasoline and emitted less CO₂.

The real-world eco-driving cycle in this study can be applied to the strategies for enhancing a low-carbon society. An eco-driving training course for motorcycle riders and an eco-mode in an assistant device on motorcycles were developed with the same or similar engine type and capacity of the test motorcycle, 4-stroke 113 cc. This eco-driving training course is suitable for motorcycle riders with high fuel consumption to improve their driving behavior because this eco-driving cycle was developed using representative data of the actual driving behavior of motorcycle riders with low fuel consumption. It will assist high-fuel-consumption riders to reduce their fuel consumption, thereby reducing CO₂ emissions. To develop real-world eco-driving cycles for different engine types and capacities, such as an internal combustion engine with a capacity greater than 125 cc or an electric engine, the innovative approach of this study can be applied.

This study has the limitation that the developed onboard measurement device was installed on the test motorcycle. The participants were asked to ride the test motorcycle to collect their real-world driving behavior. They may have ridden slightly more carefully than usual because they were unfamiliar with the test motorcycle.

Future studies should focus on improving the onboard measurement device that will be installed on the participants’ motorcycles without requiring any modifications. It might be able to capture the more natural driving behavior of motorcycle riders. This innovative approach should extend to develop the real-world eco-driving cycle for the electric motorcycle. The eco-driving training course and the eco-mode in an assistant device on electric motorcycles can help electric motorcycle riders reduce their electric power consumption.