1. Introduction
In the fight against the novel coronavirus disease (COVID-19) in early 2020, drones have been widely used as scientific and technological prevention and control tools. Their high efficiency, convenience, “zero-touch” delivery capability, and other features give drones important advantages in supplying medicines and other supplies. This delivery method can effectively reduce personal contact and reduce infection risks [
1]. In fact, drones are not limited to medical aid [
2]. In recent years, UAVs have gradually been used in various civilian fields, such as communication detection, delivery, environmental detection, and disaster relief. Traditional ground transportation is vulnerable to obstacles, such as hills and oceans that cannot be crossed. In fact, a courier company (DHL) uses drones called Parcelcopter to provide services to customers in islands or mountains [
3]. In this way, drones can save time, effort and cost. In addition, UAV delivery can also effectively avoid traffic congestion and provide fast and accurate delivery services in urban environments. JD, Antwork, and SF Express have all carried out delivery activities in cities. Amazon has developed Amazon Prime Air, which can provide fast delivery services within 30 min. At the same time, the complex environment has produced many new scenarios and new applications for drones. Their high efficiency and high adaptability can meet the diversified needs of urban scenarios and gradually win urban development space.
Statistics indicate that more than 80% of light and small unmanned aerial vehicle (UAV) activities occur below 120 m. UAVs flying at ultra-low altitudes have caused public concern over noise pollution and safety threats [
4]. In contrast to high-altitude route planning, low-altitude airspace planning is closely connected with surface information. UAV delivery is very closely related to third parties. Third parties can even prevent the operation of drones through boycotts and complaints. An assessment of the risks from drones operating in the low-altitude airspace of cities and towns will not only provide effective support for reasonable low-altitude route planning, but will also positively guide urban residents’ awareness of drone deliveries. It will also support the important and rapid development of the UAV urban logistics industry.
2. Literature Review
In recent years, the theoretical research on UAVs has mainly focused on path planning and site selection [
5,
6], urban air mobility (UAM) [
7,
8,
9], and the various constraints faced by the construction of UAM systems [
10,
11] and route planning in a low-altitude airspace [
12]. For example: Venkatesh [
5] proposed a new method based on the linear programming problem model, which aims to find the best route for vehicles carrying UAVs. Hong [
13] considered the drone’s limited range and obstacles within the required range, and proposed a new location model, which used optimization methods to design the network of site locations and delivery routes to help drones extend their coverage to a greater extent. Parker [
9] developed an exploratory case study to evaluate hypothetical On-Demand Mobility (ODM) Aviation services in Los Angeles and the greater Southern California region, and identified five key operational constraints that may inhibit the recent or long-term application of ODM Aviation operations. Both Aydin [
10] and Yoo [
14] are concerned about the impact of different roles in the urban air transportation system on the operation of drone cities. The understanding and attitudes of pilots, controllers, and the public on UAV urban delivery will affect the development prospects of UAV urban delivery. However, UAV city operation systems are getting increasingly closer to the public, which has raised public concerns [
10,
15]. Therefore, UAV urban operation systems have proposed higher requirements for the operation of UAVs in low-altitude urban environments [
16,
17,
18].
Scholars from many countries have discussed the risks of civilian drones from different disciplines and attempted to find solutions. Some scholars are concerned about the research on the law-making of civilian drones. They are committed to standardizing the legal nature and characteristics of civilian drones and clarifying the responsibilities of the legal regulatory agencies and departments for drones [
19]. This is an unavoidable step to maintain public safety and prevent social risks. Park [
20] and Choon [
21] studied the regression formula for a drone’s onboard camera and the damage caused by the crashes of drones with different heights and weights. Their research improved the UAV risk-related standards and provided favorable data support for the formulation and improvement of the UAV legal system. In terms of identifying the risk sources, it has been found that human factors, airborne equipment, and uncertainties in the complex airspace environment are multiple sources of risk for drone operations [
22]. Starting from the airborne equipment, He Qiang [
23] designed and perfected the drone flight safety assessment and control system. Xu Chenchen [
17] designed and studied the theoretical system and technical path for the iterative construction of a low-altitude air route network, with the goal of finding a safe and efficient flight method for UAVs in a low-altitude airspace. Bizhao et al. [
24] identified and classified the UAVs risks in the urban environment, constructed a general risk cost model, and applied the obtained risk cost information into the network to assist path planning. Wen et al. [
25] proposed a new search algorithm to solve the path planning problem of UAVs in low-altitude dangerous environments with dense obstacles. Koh’s [
21] work covered drone free drop modeling, FEM (finite element method)-based impact modeling, and comparable drop impact experiments, and obtained relevant data of impactor injury levels associated with HIC and AIS obtained for drones with different weights falling from various heights. The existing UAV risk research was mostly conducted from the perspective of UAV fuselage design and the risk adaptive avoidance system. At present, only some law-making studies have mentioned the social risks of drones, and these studies mostly focused on the direction and development prospects for the formulation of laws and regulations. There have been few studies on quantifying the public risk from drones, although risk quantification provides necessary support in UAV planning and flight supervision standards.
The assessment of the risk of drones operating in the low-altitude airspace of cities and towns not only provides effective support for the reasonable planning of low-altitude route planning, but also positively guides urban residents’ awareness of drone deliveries. The quantification of the operational risks of operating drones in cities can also support the important and rapid development of the drone urban logistics industry. This study assessed the drone risks in urban logistics distribution from the perspective of the public acceptance of drones. The study defined UAV operation as the risk issue object, and ground personnel as the risk bearer. A third-party risk assessment model for the urban logistics distribution of drones was constructed considering three aspects: the privacy infringement risk brought by airborne cameras, hidden safety hazards brought by low-altitude flight, and noise risk brought by propellers. This study selected the southern district of the Civil Aviation University of China (CAUC) as an example application scenario, collected data at this location, and applied a risk assessment model for calculations. A risk assessment analysis was conducted of UAV operations in the airspace 30–60 m above the CAUC to verify the feasibility of the model.
5. Case Study
5.1. UAV Selection
Distribution drones have developed rapidly, and their performance has continued to improve. This study selected the RA3 logistics distribution drone application model proposed by Antwork in 2018. The UAV has a long flight time, and the cargo compartment can hold up to 5 kg of items. It is equipped with multiple sensors, comprehensive safety mechanisms, and upgraded low-decibel propellers. Therefore, RA3 is very suitable for low-altitude operations in urban areas with a high population density and complex surface. The basic parameters are listed in
Table 2.
5.2. Example Scenario Description and Grid Ddivision
5.2.1. Example Scenario Description
A complex surface, high population density, and highly dynamic changes are typical characteristics of an urban logistics environment. The complex surface environment can test the risk changes of drones in a variety of environments. The high-density and regular population flow can clearly describe the impact of population density variables on the risk of drone operations. This study selected the southern district of the CAUC as an example location. The surface environment in this area is complex, including lakes, woods, roads, buildings, open spaces, and many other types of terrain. At the same time, this area is a typical population cluster, with a dense school population and obvious mobility. According to the survey, the number of people living in the southern district of the CAUC for a long period was 20,000, the average daily express delivery volume was 5000, and the average daily catering demand was 2000. The demand for drone delivery in the southern district of the CAUC is very high, which is one of the potential scenarios for building a drone delivery system. The tallest building in the southern district of the CAUC is 22 m, and the RA3 operating height in Hangzhou is approximately 50 m. The feasibility of actual operations was considered to ensure that UAVs were above all the buildings and did not interfere with the take-off and landing of civil aircraft. The final example of the UAV operation area selected in this study was the airspace 30–60 m above the southern district of the CAUC, and the UAV operation risk in this airspace was evaluated.
5.2.2. Example Scenario Grid Division
According to the previously described three-dimensional grid division rules, the “observation” image level was selected as the privacy violation criterion. The distance and height between the corresponding UAV and the observation object were calculated using formula (1) to be 20 m and 10 m, respectively. Accordingly, the airspace above the CAUC was divided into 780 × 4 three-dimensional grids. The plane grid division of this area is shown in
Figure 3.
5.3. Example Scenario (Southern District of CAUC) Surface Data
The ground environment variables of the UAV city operation regional risk index assessment model include the barrier factor, population density, and ground noise energy. This study used a field survey to obtain surface data for the southern district of the CAUC.
5.3.1. Barrier Factor
The barrier factor refers to the safety barrier factor and sound barrier factor. The surface barrier factor data were obtained using GIS data and a field comparison, We first preliminarily determined the vegetation type of each grid according to the map, and then went to the campus for field comparison and correction according to the map coordinate calibration results. When there were two vegetation types in a grid, we chose the one with a large footprint as the result of the grid vegetation type. For example, when 80% of the area in a grid was lakes and 20% of the area was exposed, the vegetation type of the grid was defined as lakes. Finally, we assigned barrier factors to each unit according to the allocation rules shown in
Table 3 [
32]. The results are shown in
Figure 4.
The highest value of the barrier factor is distributed in the grid covered by the building. The second highest value is for the tall trees around Millennium Lake and the chemical laboratory building, and the third highest value is for the sparse trees and low shrubs scattered throughout the school. The minimum grids are distributed across the bare surface of the school.
5.3.2. Population Density
The population density indicates the density of people on the surface space. The risk-taking object for the UAV operation is people on the ground. It can be seen that the population density is an important variable in risk assessment. The spatial distribution of the school population has a strong regularity, and the main characteristics are directionality and repetition. The characteristics of the flows of people in the teaching building, dormitory building, and cafeteria at the same time of the day are basically the same.
When measuring the population density of the example environment, this study utilized relevant traffic flow measurement and calculation methods:
In this formula, represents the pedestrian flow of the measured road section, represents the pedestrian flow density, and represents the average pace.
This study applied the formula to calculate the average speed of the interval to calculate the average walking speed of the pedestrian flow:
where
is the length of the observation section,
is the number of people passing the section, and
is the time taken by the
-th person walking across the section.
The population density data in this study were obtained between 11:00 a.m. and 12:30 p.m. on a certain working day. Drawing lessons from the traffic flow density survey method, the survey results are shown in
Figure 5. The figure shows that the densely populated grid during this time period is within the flow range from the teaching building and dormitory building to the canteen, and the entrances of the south no. 1 and south no. 3 canteens have the highest density.
5.3.3. Noise Data
This study used a noise meter to measure the environmental noise on the ground. The monitoring period was between 11:00 a.m. and 12:30 p.m. on working days. The experiment selected 24 measurement points on the campus. The A right (simulating human ear) slow mode was selected on the sound level meter. In the test, the instantaneous A sound level data were read 5 times at 10-s intervals. The arithmetic average of these five data points was used as the noise data of the measurement point, and the result is shown in
Figure 6. The figure shows that the noise hot spots mainly appear in the cafeteria and roads.
5.4. Analysis of Risk Index of UAV Southern Flight Area
The environmental data and UAV performance data were used in formula (8) for calculations. The propeller noise in the formula is 80 decibels, and the probability of a UAV accident is 10–5 [
11]. This study calculated the third-party risk assessment index for drone operations over the CAUC, and compared and analyzed the results at the same altitude and same vertical plane.
The study randomly selected different types of grids and compared the risk indexes of different levels, as listed in
Table 4:
There were two types of high-level risk index distributions. The first involved raster risks such as for the forests, canteen entrances, and roads, which gradually increased with an increase in height. The second type was the risk index of buildings such as dormitory buildings and teaching buildings, which gradually decreased as the height increased. This was due to the difference in barrier factors. Because of the protection provided by a building, the only risk from drones to the people in a building was the noise risk. According to the noise reduction principle, when the drone was further away, the noise risk to objects inside a building was lower.
In order to evaluate a suitable flight situation in the southern district of the CAUC, a 4-level evaluation of the UAV operation risk index for different grids at a flight altitude of 40 m was conducted. This evaluation not only considered the surface conditions, but also considered the extreme values, differences, and sudden values of the risk index. The results are listed in
Table 5.
As shown in
Table 5 and
Figure 7, high-risk areas are concentrated at the entrance of the canteen, where the population density is high and there is no building protection. The middle-risk area is the road around the canteen and dormitory. The teaching buildings, playgrounds, and intersections are low-risk areas. Millennium Lake, the surrounding woods, and the woods northwest of the school are micro-risk areas. In addition, the teaching building is the area with the highest building risk. The important influencing factors are the population density and ground noise. The teaching building has a high population density and quiet environment.
An examination of the whole airspace showed that a lower drone operating altitude was associated with a lower risk. When compared at the same height, the lowest risk area was Millennium Lake and the woods to the northwest, followed by the sky above the buildings; the highest risk area was the passage in front of the two canteens.
It can be seen from this that the lowest risk heights above different surface features were different. A higher drone operating area above a building was associated with a lower risk, whereas a lower height was associated with a lower risk in other areas. When drones flew at the same altitude, the risk of flying over lakes and greenery was the lowest, followed by buildings. The risk of flying over densely populated areas was the highest, and drones should try to avoid flying over such areas.
6. Conclusions
The efficiency of UAV logistics can meet the flexible and high-frequency mid- and short-distance terminal distribution needs in cities, and gradually win the development space of the urban market. Currently, UAVs mainly fly at low altitudes or even ultra-low altitudes. Especially in an urban area with a complex surface and high population density, the hidden dangers of drones urgently need to be solved. This study developed and verified the feasibility of a third-party risk assessment process for UAV urban logistics. The first step was to use the pixel regression mode of the drone camera to select different privacy infringement standards according to different scenarios to output the three-dimensional grid division size. In the second step, the third-party risk model of UAV city logistics was used to evaluate the regional risk index. Finally, based on the results of the regional risk index, classification was performed to identify suitable flying grids in the region.
The verification results of the study example showed that at the same airspace height over the school, the third-party risk of the grid above the lake and greenery was the lowest, followed by the grid above the buildings and square with low density. The largest third-party risk area was above the canteen passage. In the same vertical plane of the school airspace, the third-party risk index of the lowest level was the smallest in the entire airspace. However, when the level above the building increased, the risk decreased, whereas the 30-m level above the other areas had the lowest risk.
This study constructed a third-party risk index for drone city operations from the perspective of ground activists’ acceptance of such operations. However it needs to be admitted that when the model is further applied to actual scenarios to make decisions, more rigorous verification is still needed. In future research, we will study the change of risk index in the time dimension. In addition, other roles in the UAV operation system such as those of the government and operating departments can be considered to construct first-party and second-party risk index models. The model can consider the UAV’s own performance factors and urban planning factors such as endurance, wind resistance, the air environment, and urban low-altitude airspace division.