Exploration of Riding Behaviors of Food Delivery Riders: A Naturalistic Cycling Study in Changsha, China

Abstract

Aimed at the riding safety issue of food delivery riders in China who mainly travel by electric bikes, a naturalistic cycling study was conducted by collecting the naturalistic cycling data of dozens of food delivery riders in Changsha, China, to identify their riding characteristics. It was found that the participating food delivery riders are mainly undereducated young male adults, and the primary reason for them to take the job is the flexible working hours. Furthermore, they frequently work overtime and admit to often committing risky riding behaviors to deliver food on time. The analysis of their riding trajectories indicates that they delivered orders all day long, rather than just at mealtimes. They mainly work within 3 km of the delivery station, and the average riding radius was 2.39 km. Male riders, riders working less than one year, and riders with high school education had a relatively fast riding speed. These findings provide valuable new insights for agencies to understand the riding characteristics of food delivery riders and to formulate the appropriate countermeasures to improve their occupational safety.

1. Introduction

With the popularization of the internet in the past decade, ordering food online has become an essential part of the lives of many Chinese people. As of the end of December, 2021, there were up to 544 million food delivery consumers in China [1]. Due to the low cost, maneuverability, and other advantages of e-bikes (EBs), food delivery workers mainly deliver foods by riding EBs in China (Figure 1). With the fast development of the food delivery industry, the number of food delivery riders (FDRs) has increased to 13 million [2]. However, the riding safety issue of FDRs is also increasingly serious and has become a hot social topic [3]; the intent of delivering more food to make more money and the strict delivery time limit force FDRs to ride faster. On the one hand, since the incomes of FDRs are determined by delivery numbers, they need to deliver more orders to make more money. However, food orders always concentrate around lunchtime and dinnertime. Therefore, FDRs face huge delivery tasks in short time periods. On the other hand, since delayed deliveries cause penalties, rather than rewards, FDRs must deliver orders on time. However, platforms always set very strict delivery time limits, and FDRs have to wait for food production, look for the destinations, etc. Therefore, the actual riding time is very limited. When huge delivery orders and an insufficient delivery time are combined, they are known to be willing to commit to risky riding behaviors to save on the riding time [4], leading to many collisions. A questionnaire sent to hundreds of FDRs in Tianjin showed that 76.5% of them had at least one crash record [5]. Due to the lack of protection, FDRs are easily hurt in collisions. According to the Shanghai Traffic Police Department, there is one FDR injured or killed every 2.5 days in Shanghai [6]. To formulate effective countermeasures to prevent such crashes, it is critical to identify the riding behaviors of FDRs first.

Aimed at the riding behaviors of FDRs, many studies have been conducted. At one intersection in Beijing, China, Wei and Qin [7] observed that FDRs had different risky riding behaviors (e.g., running red lights, reverse riding) at different time periods, with most occurring during lunchtime and dinnertime. Wang et al. [5] observed hundreds of FDRs in Tianjin, China, and found the speeding rate was up to 91.3%. Wang et al. [8] analyzed the helmet usage of 6941 delivery riders through an online survey in Shanghai, China, reporting that improving the riders’ familiarity with traffic regulations could increase helmet use and reduce crashes. Gao et al. [9] explored the red-light running behavior of delivery workers by checking video cameras in Shanghai, China, reporting that waiting position, traffic violations of other cyclists, and group size played an important role in impacting red-light-running behaviors. Liu and Jiang [10] explored three typical risky riding behaviors, i.e., trespassing motorized lanes, reverse riding, and distracted riding with using a cellphone in Wuhu, China, reporting that males and older FDRs were more likely to demonstrate these behaviors. In terms of countermeasures, Dong et al. [11] studied FDRs and ordinary e-bikers through questionnaire surveys in Tianjin, finding that compared with ordinary riders, traffic enforcement had a more significant effect on preventing the risky behaviors of FDRs. Wu et al. [12] used the theory of planned behavior to analyze 150 questionnaires of delivery workers in Guangzhou, China, finding that they tended to engage more in technology-based non-riding tasks, such as manually operating their phone.

To sum up, researchers have analyzed the riding behaviors of FDRs from different aspects, mainly by roadside observations (field or video) and questionnaires. Although these studies provide many new insights regarding the riding safety of FDRs, they still have very prominent weaknesses. Roadside observations can only be implemented at limited locations and time points due to the huge costs and may not be able to reflect the whole delivery processes of FDRs. Questionnaires can only collect the subjective ideas of FDRs and may not be able to obtain their riding details. Foremost, neither can collect the real, complete riding data of FDRs. However, without such data, it is impossible to accurately identify many basic riding characteristics of FDRs, such as riding distance, riding speed, etc., which are the preconditions for us to identify their real riding behaviors. Researchers have explored this issue using the naturalistic driving study (NDS) to conduct a more precise analysis. An NDS was designed to explore driver behaviors by analyzing the long-term high-resolution driving data of ordinary drivers in their daily lives [13]. With the Secondary Strategic Highway Research Program (SHRP2) conducted in the United States [14], the largest NDS on the planet, NDSs have been widely used in driving behavior research around the world [15,16,17,18,19]. Based on the successful implementation of NDSs, aimed at bikers, researchers have also conducted naturalistic cycling studies (NCSs) to explore their riding behaviors by analyzing similar riding data [20,21,22,23,24]. With the advent of e-bikes, researchers have also tried to explore the riding behaviors of e-bikers. For example, Bian et al. [25] explored the reverse riding behavior of shared e-bikers by analyzing their trajectory data in Changsha, China. However, to the best of our knowledge, none of them have tried to study the riding behaviors of FDRs with naturalistic cycling data.

Therefore, based on the collected naturalistic cycling data of dozens of FDRs in Changsha, this study was designed to identify their riding behaviors by accurately quantifying their riding characteristics, which is fundamental for riding safety analysis but rarely conducted before. The following paper is organized as follows: Section 2 introduces the data collection; Section 3 analyzes the demography, occupation, and risky safety attitudes of FDRs; Section 4 discusses their riding behaviors according to riding time, riding distance, and riding speed; and Section 5 provides the conclusions and discussions of the work.

2. Data Collection

To explore the riding behaviors of FDRs, a naturalistic riding study was conducted. FDRs from a food delivery station in Changsha, China, were recruited to take part in the study. The station was close to multiple universities and big communities, with high food delivery demands. This study was organized into two data collection periods as follows: the 1st ranged from 12 December 2022, to 11 January 2023, with 29 participants, while the 2 ranged from 12 January 2023, to 12 February 2023, with 21 participants. Before each data collection began, all participants filled in a questionnaire to obtain their demography and other information. Since 4 FDRs took part in both data collections, 46 valid questionnaires were collected.

During the data collection, the e-bike of each FDR was installed with a GOOME W7 GPS tracker (Figure 2a) to collect his/her riding data. As long as the e-bike was moving, the tracker would automatically record timestamp, latitude, longitude, speed, direction, battery percentage, etc., with a frequency of 1 Hz. Since each tracker was also equipped with a prepaid data card, the collected data could be transmitted to an online platform for monitoring and downloading in real time. Figure 2b shows some of the sample data. It is worth noting that the device would stop recording data once it was stationary for 30 s. It could usually work for 4 to 6 days on a full charge. During the survey period, whenever a tracker was found to have run out of battery, the FDR would be reminded to recharge it. Finally, tens of thousands of kilometers of riding trajectories were collected in this study.

The collected raw data were preprocessed before analysis. Firstly, any repeated data were removed. Then, data with timestamps out of the survey period were also removed. Finally, considering the technical specifications of e-bikes, the data with speeds greater than 60 km/h were thought to be unreasonable and removed.

3. Questionnaire Analysis

3.1. Demographic Characteristics

The questionnaire results (

Table 1. The demographic characteristics of FDRs.

Characteristics	Definitions	Frequency (N)	Percentage (%)
Gender	Male	44	95.7%
	Female	2	4.3%
Age	18–25	13	28.3%
	26–35	26	56.5%
	36–45	5	10.9%
	>45	2	4.3%
Education	≤Middle school	11	23.9%
	High school	21	45.7%
	Junior college/Polytechnic	12	26.1%
	Bachelor’s degree or above	2	4.3%
Reasons for being a rider (multiple choices)	Higher income than other jobs	6	13.0%
	Flexible working hours	39	84.8%
	No skill requirements	9	19.6%
Income (RMB)	<5000	7	15.2%
	5000–8000	34	73.9%
	>8000	5	10.9%

) indicate that 85% of respondents are younger than 35 years, and none of them are older than 50 years, which means this is a very young group. In terms of gender, there were only 2 female FDRs among the 46 participants. Furthermore, 70% of them had a high school diploma or less, 25% of them only had a middle school diploma or less, and only 4.3% of them had a bachelor’s degree or above. It should be noted that with the popularization of higher education in China, 7.7% of the population have a bachelor’s degree or above now [26]. Generally, FDRs mainly comprise young male undereducated adults, probably migrant workers from rural areas, who are fit for delivering food outdoors due to their strong physical condition. Traditionally, this demographic group usually works in factories, construction sites, etc. However, the development of the food delivery industry, which does not require any technical skills or diploma, provides a new choice for them to make a living in the city.

As for the income, 73.9% of the respondents made RMB 5000–8000 (about USD 690–1100) per month, and only one-tenth of them made more than RMB 8000 per month. Considering that the disposable income per capita of Changsha was RMB 58,850 in 2022, that is, RMB 4904 per month [27], it seems that these FDRs received decent payouts. However, 84.8% of the participants choose this job for the flexible working hours, whereas only 13.0% of them choose this job for the higher income than other jobs, which implies that the flexible working time might be the most attractive feature of this job. In fact, similar findings have also been proposed in many previous studies [28,29].

3.2. Occupation Characteristics

In terms of working hours, the questionnaire results indicate that up to 89.1% of respondents work more than eight hours per day (

Table 2. Occupational characteristics of FDRs (N = 46).

Characteristics	Definitions	Frequency (N)	Percentage (%)
Length of employment	<6 months	10	21.7%
	6–12 months	7	15.2%
	1–3 years	20	43.5%
	>3 years	9	19.6%
Average daily working hours	4–8 h	5	10.9%
	8–12 h	33	71.7%
	>12 h	8	17.4%
Weekly rest time	Never	20	43.5%
	1 day	17	36.9%
	≥2 days	9	19.6%
Number of traffic accidents	None	26	56.5%
	1–3 times	16	34.8%
	>3 times	4	8.7%
Electric bike source	Self-purchased	36	78.3%
	Company-provided	1	2.2%
	Renting	9	19.5%

). Furthermore, 43.5% of respondents never rest, and only 20% of them would rest for two days per week. Both indicators show that they worked very hard. This is not a surprising result. FDRs are paid based on their delivered orders. That is, the only way for them to make more money is to finish more orders. However, the long-term overtime work is expected to impair their health gradually and make them easily fatigued, increasing crash risks. Here, up to 43.5% of respondents had been involved in traffic crashes during their employment, and 8.7% of them had more than three crash records. Considering 80% of them worked less than three years, it confirms the high risk and instability of this job. In fact, many FDRs take this job as a stopgap [9].

The performance of EBs is essential for FDRs. Here, 78.3% of FDRs purchased EBs by themselves, 19.5% of FDRs rented EBs, and only 2.2% of FDRs’ EBs were provided by the company. That is, nearly all FDRs obtained their EBs by themselves. Currently, there is no specific standard for takeaway EBs in China. It is thought that many FDRs might adopt low-quality EBs to minimize costs. A possible solution is to develop a customized technical specification for food delivery EBs [30]. It is also suggested that food delivery companies and outsourcing companies may be required to provide high-quality EBs to improve riding safety.

3.3. Riding Safety Attitudes

Frequent safety training is important for worker safety. As shown in

Table 3. The attitudes of riders towards risk-taking behaviors (N = 46).

Characteristics	Definitions	Frequency (N)	Percentage (%)
Frequency of safety training	Only training for new employees	1	2.2%
	Monthly	27	58.7%
	Weekly	16	34.8%
	Daily	2	4.3%
Familiarity with traffic rules	Low	11	23.9%
	Moderate	16	34.8%
	High	19	41.3%
Most dangerous riding behavior at work (multiple choices)	Reverse riding	29	63.0%
	Riding in motorized lanes	14	30.4%
	Running red lights	27	58.7%
	Riding without helmets	23	50.0%
	Using mobile phones while riding	20	43.5%
Helmet source	Self-purchased	14	30.4%
	Company-provided	32	69.6%
Frequency of helmet use	Always	40	87%
	Often	4	8.7%
	Occasional	2	4.3%
Reasons for wearing helmets (multiple choices)	Company policy	23	50%
	Fear of fine from traffic police	12	26.1%
	Self-protection	43	93.5%
Ways of using mobile phones	Directly using	5	10.9%
	Vehicle bracket	25	54.3%
	Using while waiting at traffic lights	16	34.8%
Reasons for using mobile phones while cycling (multiple choices)	Checking order information	32	69.6%
	Contacting customers to pick up orders	12	26.1%
	Navigation	21	45.7%
	Other reasons	2	4.3%
Reasons for breaking traffic rules (multiple choices)	Personal habits	2	4.3%
	Overdue orders	18	39.1%
	Non-motorized lanes are occupied	27	58.7%
	No non-motorized lane	22	47.8%
	Bad weather	5	10.9%
	Following others	3	6.5%
	Not willing to take a detour	12	26.1%
	Traffic light lasts too long	3	6.5%

, up to 97.8% of respondents receive at least one safety training course per month, which indicates that their employer places great emphasis on rider safety. However, nearly 25% of them were still unfamiliar with traffic rules, which implies that trainings might still need to be improved.

Reverse riding (63.0%), running red lights (58.7%), riding without a helmet (50%), and using a phone while riding (43.5%) were ranked as the top four dangerous risky riding behaviors by the FDRs. Reverse riding is the most dangerous one, probably as it easily leads to head-on collisions. This is not a surprising finding. In fact, after analyzing bicycle–motor vehicle collision records collected in the 2012 California Statewide Integrated Traffic Records System, Stimpson et al. [31] found that approximately 12% of them were caused by reverse riding of bikers. Another analysis of the trajectory data of shared e-bikes in Changsha, China, indicates that riders exhibited frequent reverse riding [25]. Running red lights is thought to be the second most dangerous. The biggest risk of running red lights is colliding with the intersecting motor vehicles, leading to angle collisions, which are extremely destructive. The red-light-running behaviors of e-bikes, including both food delivery e-bikes and ordinary e-bikes, have become the main cause of traffic accidents at signalized intersections in China [32]. Zhang et al. [33] collected 3335 samples at four signalized intersections in Xi’an, China, and found that FDRs were more likely to run red lights than ordinary e-bike riders. Wearing a helmet is one of the most important measures to protect riders and is mandated by law. Here, up to 95.7% of respondents always or often wore helmets during work, mainly for self-protection (93.5%). This indicates that most riders are aware of the importance of wearing helmets. Furthermore, as shown in Table 3, since 50.0% and 26.1% of FDRs also chose “Company policy” and “Fear of fine from traffic police” as the reasons for wearing helmets, respectively, the employer’s requirements and strict law enforcement are thought to also play essential roles in promoting helmet use. Here, nearly 70% of respondents received their helmets from the employer, which implies that the employer thinks it highly important. Furthermore, although using a phone while riding is also thought to be very dangerous, all respondents still did it and two thirds of them always did it. They mainly used cellphones for checking orders (69.6%), navigation (45.7%), and contacting customers (26.1%), which are important for delivering food smoothly. Therefore, it is nearly impossible to not use cellphones while riding for FDRs. Promoting smart helmets to reduce distracted phone usages while riding might be more practical [34,35].

In terms of traffic violations, the most common reasons were occupied non-motorized lanes (58.7%), the lack of non-motorized vehicle lanes (47.8%), overdue orders (39.1%), and unwillingness to take a detour (26.1%). Theoretically, EBs should run on non-motorized lanes. However, it seems FDRs are often forced to run on motorized lanes. For example, non-motorized lanes are often occupied by parked vehicles. Zhang et al. [36] collected video records of non-motorized vehicles in Hefei, China, and found that poor riding conditions on non-motorized lanes can provoke electric bikers to travel at higher speed to drive on motorways. Agencies can make road designs friendly to non-motorized road users [37]. Only 26.1% of the respondents would violate traffic laws for avoiding detours. Therefore, traffic management departments may improve non-motorized lane usages by optimizing traffic facilities, such as increasing the width of non-motorized lanes [38], which may help reduce risky riding behaviors.

4. Riding Characteristics Analysis

4.1. Riding Time

Since the trajectory points were recorded periodically, their temporal distributions can reflect the working time of the FDRs. Figure 3a shows the histogram of the trajectory points by the time of day. Considering the technical specifications of EBs, only data with speed below 60 km/h were thought to be reasonable and analyzed in this section. It can be found that food deliveries start at 08:00, i.e., breakfast time, and last until midnight with two peaks at 12:00 (lunchtime) and 18:00 (dinnertime), respectively, which were thought to be reasonable. Furthermore, there were many deliveries between mealtimes and at nighttime. Further, Figure 3b shows a histogram of the trajectory points by the day of the week. It can be seen that deliveries are very consistent over the week, implying consistent food delivery demands on weekdays and weekends. Food delivery was initially designed to provide lunches to busy workers on working days, but our findings indicate that it has been greatly expanded to provide many more services. In fact, currently, FDRs can deliver foods, flowers, medicines, and anything needed in life in China. The diverse delivery services are also believed to stimulate more demand in turn. This may also explain why 80.4% of participating FDRs never rested or rested for just one day per week, and often worked overtime as mentioned above.

4.2. Riding Distance

Riding distance could directly reflect the workload of the FDRs. Figure 4 shows the histogram of their individual daily riding distances. It can be found that most FDRs rode 40 km to 90 km per day, with the mean being 65.4 km and the 85th percentile being 99.3 km. Since most EBs can only run 40–60 km per charge, it means that most FDRs might need to charge at least once per day.

It should be noted that for education, junior college/polytechnic and bachelor’s degree or above were merged as the college type, as they are both taken as higher education in China. Lengths of employment less than 12 months were also taken as one type. Furthermore, depending on whether they had crash records, FDRs were redivided into two types.

In Figure 5, the median daily riding distance of male FDRs is 62.4 km, much shorter than that of females (85.2 km). The t-test (male vs. female) indicates that their difference is significant with the 95% confidence interval being (−29.5 km, −8.3 km). Generally, a longer riding distance means more deliveries and higher incomes, which implies the stronger motivations of females to make money. This needs further investigations to identify what leads to the big difference between females and males in terms of riding distance. Meanwhile, it also should be noted that since there were only two female participants in this study, further studies are needed to identify whether this finding is general or just a special case by investigating more female FDRs. In terms of length of employment, the median daily riding distances of FDRs working less than 1 year, between 1 year and 3 years, and more than 3 years were 69.6 km, 58.9 km, and 61.9 km, respectively. It can be observed that FDRs who had been working for less than one year had longer riding distances. The t-test (<1 year vs. ≥1 year) indicates that the difference is significant with the 95% confidence interval being (4.9 km, 11.5 km). However, the differences between the other two groups of FDRs were not as pronounced. As shown in Figure 5c, crash experiences did not have a major influence on the riding distances of FDRs. Although the t-test (no crash vs. experienced crash) indicates that the difference is significant with the 95% confidence interval being (0.3 km, 6.9 km), it is quite limited. With regard to education, FDRs with a high school education had the highest median daily riding distances (71.9 km), while riders with a college education had the lowest (52.4 km).

There were two special events occurring during the data collections: the official lift of lockdown policies in China on 8 January 2023, and the Chinese New Year holidays from 21 January 2023, to 27 January 2023. They both had huge impacts on society. To check whether they also impacted food delivery activities, the data were divided into the following four periods according to the date: Period 1 from 12 December 2022 to 7 January 2023, i.e., before the lifting of the pandemic restrictions; Period 2 from 8 January 2023 to 20 January 2023, i.e., after restrictions were lifted but before Chinese New Year; Period 3 from 21 January 2023 to 27 January 2023, i.e., the Chinese New Year holiday; and Period 4 from 28 January 2023 to 12 February 2023, i.e., after the Chinese New Year holiday. Figure 6 shows the boxplots of the individual daily riding distances during the different periods. It can be found that, from Period 1 to Period 2, there was a decrease in the riding distances of FDRs. The t-test (Period 1 vs. Period 2) indicates that the decrease was significant with the 95% confidence interval being (2.7 km, 11.6 km). Period 3 saw the shortest riding distances out of all the periods, at only 50.7 km. The t-test (Period 3 vs. other periods) indicates that this was significantly smaller than other periods with the 95% confidence interval being (−19.0 km, −9.7 km). This was thought to be reasonable. As most Chinese people would go back to their hometowns to reunite with families in the Chinese New Year holidays, the population would greatly decline, leading to much fewer food delivery demands. Furthermore, it is a tradition to cook food at home during the Chinese New Year holiday. After the end of the holiday, with people returning to the city, Period 4 showed that the riding distances increased to normal.

Furthermore, the riding range is another point of interest. It was found that the average distance of the riding trajectory points to the delivery station was only 1.08 km. Figure 7 shows the spatial distribution of the furthest individual daily trajectory points. Their average distance to the delivery station was 2.39 km, and FDRs mainly work within 3 km of the delivery station. This might be attributed to the fact that the service radius of food delivery stations is usually 3 km. That is, the FDRs generally work in very small areas around the food delivery station. The small riding range means that FDRs could become familiar with the service area very quickly, which would help them to find the most efficient delivery paths at work.

4.3. Riding Speed

Riding speed could be directly used as an indicator of riding safety, as speeding is the consequence of many risky riding behaviors. Figure 8 shows the histogram of the riding speed. It was found that the riding speed concentrates around 15 km/h, with the mean being 17.1 km/h and the 85th percentile being 27.0 km/h. Although EBs could run much faster in theory, their real running speeds were determined by traffic environments, including traffic congestion, law enforcement, and so on. As shown as Figure 9, male FDRs rode much faster than females, with their average riding speeds being 17.3 km/h and 14.1 km/h, respectively. The t-test (male vs. female) indicates that their speed difference was significant with the 95% confidence interval being (3.18 km/h, 3.21 km/h). This is not a surprising result, as male riders are known to be more adventurous [9,12]. In terms of length of employment, FDRs who had worked for less than 1 year had a mean speed of 18.1 km/h, while that of those with 1–3 years’ experience was 16.4 km/h, and that of those with more than 3 years’ experience was 16.5 km/h. The t-test (<1 year vs. ≥1 year) indicates that the speed difference was significant with the 95% confidence interval being (1.60 km/h, 1.62 km/h). Perhaps 1 year of employment is a turning point for FDRs, as they may better control their riding speed as their length of employment increases. In Figure 9c, when dividing the FDRs into two groups—with accident experience and without accident experience—the average riding speeds did not change much, which were 16.9 km/h and 17.1 km/h, respectively. With regard to education, FDRs with different education levels had slightly different mean speeds. The mean speeds corresponding to education levels from low to high were 16.6 km/h, 17.8 km/h, and 16.0 km/h, respectively. It was observed that FDRs with a college education had the lowest mean speed. The t-test (≤high school vs. college) indicates that the speed difference was significant with the 95% confidence interval being (1.45 km/h, 1.47 km/h).

Figure 10 shows the boxplots of the riding speeds during the different time periods. It can be found that the riding speed showed an increasing trend from Period 1 to Period 3, with Period 3 having the highest value, but then decreased from Period 3 to Period 4. It is thought that riding speeds of FDRs are mainly determined by the traffic on the road. When approaching the holiday, people are expected to gradually leave to go back to their hometowns; thus, traffic would also decrease. The less congested traffic means that FDRs could ride faster without worrying about interruptions from other road users. The t-test (Period 3 vs. other periods) indicates that FDRs did ride much faster in the holidays with the 95% confidence interval being (2.25 km/h, 3.47 km/h). After the holiday, with the increase in road traffic, FDRs could only ride slower. The findings also imply that risky riding behaviors might be much more frequent during the Chinese New Year holidays.

5. Conclusions

With the explosive development of the online food order industry in the past decade, food delivery has created tens of millions of jobs in China, and many people make a living in these jobs. However, as FDRs travel fast on their way to deliver foods, their riding safety is also an increasingly serious issue. To identify their riding characteristics and develop effective countermeasures, we conducted a naturalistic riding study by collecting and analyzing the high-resolution trajectory data of 46 FDRs in Changsha, China. The survey found that most participating FDRs were young men with a high school or lower diploma, and their monthly income was concentrated in the range RMB 5000–8000. They mainly travelled by EBs, which are low-cost, convenient, and fast, but also lack protection in collisions. Meanwhile, 89.1% of these FDRs worked more than 8 h per day, with some even exceeding 12 h, and nearly half of them did not rest once per week, which confirms the heavy workload of this job. In terms of riding safety attitude, reverse riding, running red lights, not wearing helmets, and using cellphones while driving were taken as the top four risky riding behaviors by them. While FDRs generally adhered to wearing helmets to protect themselves, they often used cellphones while riding. They admitted that they often violated traffic rules to complete orders on time, but this was mainly due to uncontrollable factors, such as the lack of non-motorized lanes, occupied non-motorized lanes, and the harsh delivery time limit. FDRs seemed to deliver food all day long, rather than just at mealtimes. In terms of gender, male FDRs rode much faster, whereas female FDRs rode for much longer distances. Furthermore, during the Chinese New Year holidays, FDRs rode much faster, but their riding distances significantly decreased.

This study provides reliable data and theoretical foundations for agencies, companies, and FDRs to gain a deeper understanding of the characteristics of FDRs. This study provides solid data analysis and many new insights regarding the riding safety of FDRs, which can help agencies to establish more scientific and targeted countermeasures to protect both FDRs and the public. Firstly, non-motorized lanes must be built and maintained in good condition so that FDRs can ride on them to minimize risky riding behaviors. Considering the possible conflicts between low-speed bikes/pedestrians and high-speed e-bikes, when possible, non-motorized lanes should be widened to separate slow and fast lanes to eliminate these conflicts [39]. Secondly, aiming at the frequent cellphone usage while driving, food delivery companies should provide FDRs Bluetooth-based smart helmets [34,35] to make phone calls handsfree. Currently, although smart helmets have been developed, they are not widely adopted due to their high costs. Thirdly, it is also necessary to give FDRs more safety training by watching accident videos and learning traffic laws, especially for those without crash experiences.

Although this naturalistic cycling data study provides many new findings, it does have limitations. Firstly, this study focused on giving a general picture of the riding behaviors of FDRs, rather than delving into the details of specific risky riding behaviors. However, as shown in this study, there are some extremely dangerous riding behaviors. To formulate customized countermeasures, it is essential to perform exclusive analyses of each risky riding behavior. Future research can use the collected data to deeply investigate these specific risky riding behaviors, such as speeding, reverse riding, and running red lights, to identify the factors affecting their occurrences and provide refined solutions. Secondly, as is known, the riding behaviors of FDRs are heavily influenced by the delivery time limits determined by the companies. However, such information was not collected in this study, as it would have been too cumbersome for the participants to record them and the budget for this study was too limited to support this. Future studies might consider collecting the delivery time information to identify its influencing mechanism on FDRs. Thus, it can be used to evaluate whether some actions should be taken by food delivery companies to reduce the intent of FDRs to fundamentally travel faster. Thirdly, the working statuses of the FDRs were believed to vary by time of day, as they rode up to nearly 65 km per day. Therefore, it would also be interesting to explore whether and how their riding behaviors would change over time by checking their trajectory data. Some trajectory-focused methods, such as functional data analysis, might be very helpful in identifying the possible changes and trends. Finally, although female FDRs rode slower than males, they were found to ride for much longer distances per day, which is very surprising. This needs further investigations to identify why those female FDRs work so hard. In particular, considering there were only 2 females out of the 46 participants, it would also be greatly important to identify whether this is general or just a special case.