Safety Impact of Introducing Automated Road Debris Removal System in South Korea

Abstract

Road debris should be removed quickly because it can cause road damage and traffic accidents. An automated road debris removal system—called ROBOS—was developed in South Korea to remove road debris automatically and quickly. It can be mounted and used on vehicles over 3.5 tons. The effects of ROBOS on the traffic flow need to be studied before it can be used on actual roads. In this study, these effects were investigated by performing driving and microscopic traffic simulations. The former can be used to analyze drivers’ reactions to new road equipment and driving behavior when detecting road debris. Driving behavior results that met the research purpose were employed in the microscopic traffic simulation to imitate reality. Finally, the number of conflicts and the conflict rate were compared between manual and ROBOS vehicle road debris removal scenarios to analyze traffic safety. The results showed that the overall traffic safety was low when road debris was present in curved sections, where it was generally difficult for the driver to easily detect it, and in the first lane and third lane, where lane changes were limited. Assuming that a ROBOS vehicle was used in these sections, the number of conflicts and the conflict rate can be significantly reduced. This study evaluated traffic safety after the use of a ROBOS vehicle input and obtained basic information needed for decision-making in future road debris removal work, such as the priority of input when choosing between manual removal and removal by a ROBOS vehicle.

Keywords

Automated Road Debris Remover System (ROBOS)road debris driving behavior traffic simulation conflict

Drivers often face various dangerous situations while driving on the road. In particular, various debris on the road can instantly cause unusual driver behavior. Road debris usually refers to objects lying on the road and includes objects that have fallen from vehicles, the remains of traffic accidents, and roadkill. In South Korea, it is not mandatory to install container boxes for cargo loaded on freight vehicles; therefore, improperly loaded cargo often falls from vehicles onto the road, causing accidents. In fact, over the past 5 years (2017–2021), 1.19 million pieces of road debris were found on South Korean highways, causing 199 traffic accidents and injuring 17 people ( 1 ). In addition, statistics on all roads in the country also revealed an average of more than 25 traffic accidents caused by road debris removal and approximately 40 debris-related traffic accidents per year. Of the accidents caused by road debris, collisions between debris and vehicles accounted for the largest proportion (54%), followed by lane departures that occurred while avoiding road debris ( 2 ). Road debris can cause both road damage and human casualties. For example, if construction materials and heavy objects are improperly loaded onto heavy vehicles and fall onto the road during travel, they can damage the road surface. Therefore, the need for quick and safe removal of road debris has gained attention. As such, traffic agencies have employed various methods, including reward systems for reporting road debris and the use of artificial intelligence technology to crack down on illegally loaded vehicles and detect road debris quickly. In addition to these efforts, it is important to remove road debris as soon as it is detected.

Currently, road workers are responsible for manually removing road debris. However, this practice can cause traffic accidents to occur when drivers fail to see the workers or the safety equipment designed to protect them. The fatality rate for traffic accidents that occur during road debris removal is approximately 31%, more than three times the rate of general traffic accidents (9.9%) ( 1 ). To address this problem, many efforts have been made to introduce automated road debris removal systems to replace manual removal by road workers. Road authorities such as the Korea Expressway Corporation and some of local governments have developed their own road debris removal equipment, which involves simply attaching a small removal box to the front of a vehicle and changing the speed of the vehicle to collect road debris. However, this equipment is no longer in use because of its poor results in practice, caused by damage to the removal box during road debris collection and work restrictions. Similarly, in the U.S., the Gator Getter was introduced, which is a cylindrical collection tray attached to the front of a vehicle (where the length of the tray is equal to the width of the vehicle) ( 3 ). The method of attaching and detaching removal equipment to the front of the vehicle is a removal method that is more manual than automated. This approach does not meet the goal of automating the removal process. In addition, the size of the removal tray must be adjusted to meet the specifications of domestic vehicles. In this case, checking the structural safety of the vehicle should be a priority. Especially for domestic use, there are problems with changing the structure of the vehicle and not meeting legal standards such as the Korean Road Traffic Act.

To address these issues, an automated road debris remover system—called ROBOS—that allows workers to remove debris more safely, was developed through national research and development in South Korea in 2019. ROBOS was developed to be mounted and used on vehicles over 3.5 tons. Since the first ROBOS vehicle was designed to be suitable for domestic roads by following the domestic Road Traffic Act and Vehicle Safety Act, its removal method is different from that of manual removal ( 4 ). To commercialize ROBOS vehicles, many studies have been conducted to advance the equipment and improve performance. Yang et al. compared the working efficiency between manual and ROBOS vehicle removal of road debris ( 5 ). They also derived the appropriate safe deceleration distance of a ROBOS vehicle approaching road debris for removal. Previous studies have simulated manual and ROBOS vehicle removal scenarios to analyze the impact of road debris removal work. However, such studies did not sufficiently consider real-world situations because the types of road debris and the driving behaviors of drivers who noticed either road debris or a ROBOS vehicle were not considered in sufficient depth.

This study was conducted to evaluate the safety before and after the introduction of a ROBOS vehicle, in a laboratory environment. For this purpose, a ROBOS vehicle was assumed to be used for road debris removal, and the effects on traffic safety could be analyzed. A driving simulator and microscopic traffic simulation were employed as the main analysis tools. It is expected that the results of this study will be used to develop an operational manual for the ROBOS vehicle and efficient management plans for road debris removal.

Research Design

Overview of ROBOS

ROBOS was developed through prior surveys and discussions with traffic experts, engineers, and traffic agencies concerning the essential functions of road debris removal, the types and shapes of debris to be removed, the appropriate vehicles for system installation, and the related laws and systems. Table 1 shows the final specifications of the ROBOS development.

Table 1.

Considerations and Design Specifications of ROBOS

Consideration	Design specification
Collection target	Plastic boxes, aluminum cans, flat tires, roadkill, etc.
Vehicle type	Over 3.5 ton long-axis truck
Operating speed	Max. 40 km/h
Allowable debris weight by type	General road debris: max. 5 kg; roadkill: max. 10 kg
Allowable debris size	120 cm (width) x 20 cm (height)
Debris loaded capacity per one time	Approximately 2 ton

Note: max. = maximum.

ROBOS is designed to be mounted on vehicles over 3.5 tons. Figure 1 shows the basic concept of road debris removal using a ROBOS vehicle.

Figure 1.

Example of a ROBOS vehicle prototype.

A ROBOS vehicle travels at 30–35 km/h. Road debris is first collected by the front guide installed at the lower front of the vehicle and sent to the lower center of the vehicle. It is then moved to the collection box in the lower part of the vehicle and sent to the cargo box. The front guide is designed to collect all debris near the shoulder, within the lanes, and near the median. The overall height of a ROBOS vehicle is 30 cm, and the debris to be collected is 30 cm high and 120 cm wide ( 5 , 6 ).

Analysis Framework

To evaluate the effect of employing a ROBOS vehicle on the road, situations were created for removal work by a ROBOS vehicle and by road workers. The results of the analysis were compared. For this purpose, a three-step analysis procedure and its corresponding scenarios were established, as shown in Figure 2.

Figure 2.

Overview analysis procedure.

The use of a ROBOS vehicle on roads without specific operating manuals and pilot studies could be dangerous for road users. This study used simulation to analyze traffic safety related to the introduction of a ROBOS vehicle. Previous studies have often used driving simulators to analyze lane and speed changes of drivers near work zones. These simulations were used to evaluate safety, operational efficiency, and the effect of hazard warnings in work zones (7 –15). The reactions of the participants in the driving simulator experiments to ROBOS vehicles can be analyzed, but it is difficult to determine the effect of ROBOS vehicles on the traffic flow of the entire road section from the perspective of the traffic agencies. To this end, a microscopic traffic simulation was also used in this study. The basic parameters of the simulation model were set to perform lane changes while maintaining an appropriate safety distance. However, drivers on the real road may exhibit different driving behaviors from those in the simulations because avoidance actions, such as aggressive lane changes and rapid deceleration, may be performed to avoid road debris. Therefore, in this study, the driving behavior data of individual drivers detecting road debris and the ROBOS vehicle in a virtual road environment were collected using a driving simulator, and key information from the data was employed in the simulation. It was assumed that this approach would provide a more realistic simulation. To evaluate the safety of the traffic flow following ROBOS vehicles, various assumptions were made about the road geometry and the location of road debris.

Traffic safety may depend on the time at which drivers detect road debris and how they react to new vehicles. In addition, traffic flow conditions must be defined for microscopic traffic simulation analysis. Accordingly, the level of service with the road geometry and debris location was considered. Table 2 shows the driving simulator and microscopic traffic simulation scenarios.

Table 2.

Details of Analysis Scenario

Main consideration	Step 1: Scenario for driving simulator	Step 2: Scenario for traffic simulation
Road geometry	Straight and curve section	Straight and curve section
Debris location	Lanes 1, 2, and 3	Lanes 1, 2, and 3
Level of service	NA	From A to C

Note: NA = Not available.

Driving Simulator Experiments

Drivers encounter road debris or notice a ROBOS vehicle traveling at a relatively low speed. The latter can be interpreted as driver reactions to the introduction of new equipment. In general, uninterrupted flow has higher average driving speeds and more severe crash rates than interrupted flow ( 16 ). Interrupted flow is connected to surrounding roads, allowing drivers to quickly detour to other routes in the event of an unexpected situation, and it is possible to install worker safety facilities and provide sufficient workspace. Considering these characteristics, a ROBOS vehicle was considered to be more suitable for uninterrupted flow than for interrupted flow. Therefore, uninterrupted flow was selected as the simulation network.

The driving simulator network was composed of three lanes with a total length of 5 km, including straight and curved sections, and an average speed of 80 km/h, considering the typical national highway type in South Korea. The curved section was constructed based on the minimum curve radius (280 m), curve length (450 m), and maximum superelevation (6%) in urban areas at a design speed of 80 km/h according to the Rules on Road Structure and Facility Standards ( 17 ). As shown in Figure 3, the final environment was assumed to have road debris in the first, second, or third lane, 2.5 km from the starting point along the 5 km section (assuming only one piece of debris per experiment). The debris had the maximum size (30 cm [height] x 120 cm [width]) that could be collected by the ROBOS vehicle.

Figure 3.

Road debris situation in a driving simulator: (a) Straight section and (b) Curved section.

Participants in the experiment were selected as follows. Fifteen candidates were selected from the applicants through in-depth interviews about: 1) willingness to participate in the experiment, 2) previous experience with accidents caused by road debris, 3) driving experience, and 4) gender. Before the experiment, a preliminary survey about road debris was conducted. Ten participants were finally selected based on the results of the final survey. The main experiment was repeated after conducting pilot driving tests to show the same driving behavior as on real roads. Only a single vehicle was allowed to drive on the road, and participants entered the experiment in order. During the driving experiment, they were free to change lane to avoid road debris and the ROBOS vehicle. To identify such behaviors, three types of driving behavior data (lane-change distance, deceleration, and steering) collected from the simulator were analyzed. Lane changes caused by road debris and the ROBOS vehicle were performed when the driver recognized the traffic situation and took evasive action accordingly. Quick detection of road debris enabled the drivers to change lanes in a stable manner. This is because the lane-change distance was long, and the space needed to avoid the debris could be determined in advance. Therefore, this study used the lane-change distance to identify when the driver detected either the road debris or the ROBOS vehicle and analyzed the avoidance behavior through changes in deceleration and steering. The lane-change distance was derived by measuring the distance between the debris and the position of the vehicle in the driving lane when the changes occurred. Because deceleration and steering are immediate actions taken by a driver to control the vehicle, they can be used to quantitatively identify driver responses to road debris. Changes in deceleration and steering over time were used to analyze the rapid lane changing behavior of drivers who detected road debris. In this study, driving behavior was analyzed by comparing changes in lane-change distance (m), deceleration (m/s²), and steering (ratio) under each experimental scenario, as shown in Table 3.

Table 3.

Definition of Analysis Data

Index	Description
Lane-change distance (m)	Distance between debris and lane position when driver begins to change lanes
Deceleration (m/s²)	Change in speed reduction
Steering (ratio)	Steering wheel input value (left and right): −1 = left; 1 = right

Microscopic Traffic Simulation

The identical network environment in the driving simulator was built using VISSIM. In addition, scenarios were established with different road geometries, debris locations, and service levels of C or higher (from A to C). Service levels of D or lower did not meet the purpose of this study and were excluded because they indicate unstable traffic flow conditions where the traffic speed is low and lane changes are limited ( 18 ). For the analysis, situations in which road debris was removed manually or removed by the ROBOS vehicle were assumed, and the simulation analysis time and network analysis area were also set as shown in Figure 4. The total simulation time was set to 4,500 s, considering the time for the proper distribution of traffic in the simulation network (the so-called warm-up time), the occurrence of debris, and its removal. In addition, it was assumed that the road debris was removed 600 s after its occurrence. The situations before the start of the debris removal were the same, and only the time of the removal work was varied. The time for manual removal was set at 1,800 s by considering the average working time of actual road workers and a previous study ( 5 ).

Figure 4.

Conceptual illustration of simulation analysis time: (a) removal manually and (b) removal by the ROBOS vehicle.

The simulation network was built with a total length of 5 km (the same as in the driving experiment) and road debris was assumed to be present 2.5 km from the starting point. For the safety of the road workers, a separate work area with a lane closure was made. In accordance with national guidelines, the workspace was divided into warning (175 m), transition (30 m), work (1 m), and finish (10 m) areas ( 19 ). However, the ROBOS vehicle removal did not require separate work areas because ROBOS was developed to remove road debris while moving. Figure 5 shows the simulation method for each debris removal situation.

Figure 5.

Conceptual illustration of simulation scenario: (a) removal manually and (b) removal by the ROBOS vehicle.

A signal head was used to generate road debris in the simulation. A red signal was set to appear at the time of debris occurrence so that vehicles could make lane changes after detecting the signal. The lane change distance derived from the driving experiment was applied to the simulation. During manual removal, the red signal for the work area was on so that vehicles could not pass during the removal time. However, when the debris was removed using the ROBOS vehicle, the vehicle was set to travel at 80 km/h (the average vehicle speed in the section), decelerate slowly 100 m ahead of the debris, remove the debris while traveling at 30 km/h, and then travel again at a constant speed of 80 km/h ( 20 ). The deceleration situation of the ROBOS vehicle was implemented using the VISSIM Reduced Speed Areas function. When the ROBOS vehicle reached the debris to be removed, the red signal was changed to green to allow vehicles to pass during debris removal. As a result, as shown in Figure 6, the simulation environment was built differently according to the road debris removal situation. To improve the reliability of the simulation results, multiple simulations were run with random seeds ( 21 ).

Figure 6.

Conceptual illustration of VISSIM simulation network: (a) manual removal and (b) removal by the ROBOS vehicle.

Analysis of Safety Impact

The number of conflicts and the conflict rate were selected as indicators to evaluate the safety of traffic flow. The Surrogate Safety Assessment Model (SSAM) developed by the Federal Highway Administration in the U.S. was used. This software is capable of analyzing the occurrence of conflicts using individual vehicle data calculated from microscopic traffic simulation models based on the conflict theory. It derives conflicts from surrogate safety measures such as the time-to-collision (TTC) and post-encroachment time (PET). We set the thresholds for TTC and PET based on the SSAM manual (22 –24). For each conflict angle, the frequencies of rear-end, lane-change, and crossing conflicts were derived. In this study, crossing conflicts were not considered, so only rear-end and lane-change conflicts were defined as the number of conflicts. In addition, safety was evaluated by calculating the crash rate, which is the number of conflicts compared with the traffic volume (25 –27).

Analysis and Results

Driving Behaviors by Scenario

Driving behavior was analyzed for the debris-only cases and with the ROBOS vehicle. Table 4 shows the results of the analysis for each scenario.

Table 4.

Driving Behavior Results

Scenario	Debris location	Lane-change distance (m)	Deceleration (m/s²)	Deceleration change	Steering (ratio)	Steering change (ratio)
Straight section
Debris-only case	1st lane	67.344	0.000	1.306	0.013	0.014
	2nd lane	70.886	−0.155	1.191	0.005	0.013
	3rd lane	69.398	−0.187	1.023	0.006	0.010
	Average	69.209	−0.114	1.173	0.008	0.012
ROBOS case	1st lane	248.744	−0.013	0.530	0.006	0.006
	2nd lane	209.520	−0.072	0.543	0.005	0.006
	3rd lane	251.806	−0.056	0.534	0.004	0.006
	Average	236.690	−0.047	0.536	0.005	0.006
Curved section
Debris-only case	1st lane	49.185	−0.479	0.905	0.005	0.010
	2nd lane	51.809	−0.301	0.802	0.004	0.009
	3rd lane	55.685	−0.016	0.794	0.005	0.009
	Average	52.226	−0.265	0.834	0.005	0.009
ROBOS case	1st lane	185.725	−0.033	0.408	0.002	0.005
	2nd lane	187.982	−0.200	0.436	0.002	0.005
	3rd lane	208.887	−0.195	0.360	0.002	0.004
	Average	194.198	−0.143	0.401	0.002	0.005

Regardless of the road geometry, the average lane-change distance was found to be approximately three times longer in the ROBOS vehicle case (straight: 236.7 m, curved: 194.2 m) than in the debris-only case (straight: 69.2 m, curved: 52.2 m). In the debris-only case, the drivers did not allow enough time to change lanes after noticing the debris, as also indicated by changes in deceleration and steering. In this case, the average deceleration and steering change were greater than in the ROBOS vehicle case. This finding indicates that the drivers performed a faster avoidance maneuver because the debris was detected later than in the ROBOS vehicle case. It seems that the drivers changed lanes in a more stable manner when the ROBOS vehicle was in sight, because the slow-moving ROBOS vehicle could be detected at a sufficient distance, unlike the debris. The average lane-change distance was shorter and the changes in average deceleration and steering were greater in the curved section than in the straight section for the debris-only case. The change in average deceleration by lane was compared for both cases. The change in deceleration was greater in the first lane than in the other lanes in the straight section. The change in deceleration was greater in the straight section than in the curved section, regardless of the debris location, because the driver was already decelerating to safely negotiate the curved section. In addition, the driving characteristics differed by lane in the curved section because the driver’s field of view was restricted by the geometry. In both cases, the lane-change distance was the longest in the third lane in both the straight and curved sections. In the curved section, the lane-change distance was longer in the third lane (153.2 m) than in the first and second lanes (136.5 m and 136.2 m, respectively). This difference seems to exist because the driver’s field of view was the widest in the third lane at the outermost part of the curved section, and thus the driver recognized the ROBOS vehicle and changed lanes earlier than in the other lanes. In the straight section, the lane-change distances in the first and third lanes (181.4 m and 182.4 m, respectively) were found to be longer than those in the second lane (138.6 m). Thus, regardless of the road geometry and debris location, the drivers showed more appropriate lane-change distances and speed reductions in the ROBOS vehicle case than in the debris-only case.

Safety Impact of ROBOS Vehicle Employment

In microscopic traffic simulations, it is very difficult to implement a situation where individual vehicles detect road debris. Therefore, in this study, the important driving behavior information derived from the driving simulator was employed in the simulation to enhance its reliability. Previously, the driving simulator environment consisted only of road debris, a ROBOS vehicle, and participants; thus, after detecting road debris, the driver could change lanes without surrounding obstacles. However, in this simulation, because the traffic volume was set for the road section, surrounding vehicles may have interfered with the instantaneous lane changes. Arbitrarily setting changes in deceleration and steering as parameters cannot consider the natural debris avoidance behavior because it applies the same lane-change behavior to all vehicles. Therefore, among the driving behaviors studied, only the lane-change distance (which represents the debris detection distance) was considered as a simulation parameter among the driving behaviors studied. The trajectory data of individual vehicles were collected for each debris removal situation by simulating manual and ROBOS-vehicle removal. From the collected individual vehicle information, the number of conflicts was derived by SSAM. Table 5 and Figure 7 present the analysis results for the average number of conflicts according to the debris removal situation, road geometry, debris location, and service level.

Table 5.

Analysis of Conflict Frequency

Removal manually				Removal by a ROBOS vehicle
Lane	A	B	C	Lane	A	B	C
Straight section
1st	37.7	149.4	396.8	1st	26.4	73.8	145.3
2nd	11.8	71.1	308.7	2nd	7.7	57.7	112.4
3rd	40.8	159.9	419.5	3rd	26.5	83.5	159.7
Curved section
1st	33.4	144.3	424.9	1st	9.2	36.9	136.0
2nd	10.5	70.5	356.9	2nd	3.1	35.4	134.4
3rd	37.0	147.3	464.1	3rd	13.0	110.7	221.5

Figure 7.

Analysis of conflict frequency: (a) straight section and (b) curved section.

The results of the analysis show that the number of conflicts increases as the service level decreases, and more conflicts occur when the service level is C. As levels A and B have less traffic than C, individual vehicles are free to change lanes, so it can be assumed that fewer conflicts will occur regardless of debris location and road geometry. In addition, one-way analysis of variance (ANOVA) was used to determine if the difference by debris location was statistically significant. Table 6 shows the ANOVA results by level of service by road geometry.

Table 6.

Analysis of Variance Results

Section	Level of service	Sum of squares	Mean squares	F-value	P-value
Removal manually
Straight	A	14,991	7,496	145.9	0.000*
	B	141,221	70,610	191.5	0.000*
	C	167,038	83,519	46.7	0.000*
Curved	A	12,376	6,188	104.4	0.000*
	B	112,684	56,342	100.2	0.000*
	C	168,461	84,230	41.0	0.000*
Removal by a ROBOS vehicle
Straight	A	2,368	1,184	4.33	0.019*
	B	4,353	2,177	0.65	**
	C	67,454	33,727	3.03	**
Curved	A	912	456	96.42	0.000*
	B	48,877	24,438	63.26	**
	C	67,821	33,910	2.92	**

P-value ≤0.1; **P-value >0.1.

There was a significant difference in the number of conflicts depending on the location of the debris. Unlike manual removal, a ROBOS vehicle can collect debris for removal more safely without the need for separate working areas for road workers. Therefore, the difference in the number of conflicts by the debris location appears to be more significant for manual removal than for ROBOS vehicle removal.

Table 7 shows the results of a post hoc analysis that was performed for the same scenarios as in the ANOVA to determine if the statistical significance was more obvious for the location of road debris. For service levels A and B, the P-value was less than 0.1. Therefore, the difference between lane 2 and lanes 1 and 3 is statistically significant. In contrast, the P-value of lanes 1 and 3 was greater than 0.1, indicating that the difference in the number of conflicts between lanes 1 and 3 is not statistically significant.

Table 7.

Post Hoc Test Results

Section	Level of service	Lane	Mean difference	Lower bound	Upper bound	P-value
Removal manually
Straight section	A	2 to 1	−25.40	−30.01	−20.79	0.000*
		3 to 1	3.60	−1.01	8.21	**
		3 to 2	29.00	24.39	33.61	0.000*
	B	2 to 1	−78.27	−90.61	−65.92	0.000*
		3 to 1	10.53	−1.81	22.88	**
		3 to 2	88.80	76.45	101.15	0.000*
	C	2 to 1	−39.94	−67.37	−12.51	0.002*
		3 to 1	67.31	39.63	94.99	0.000*
		3 to 2	107.25	79.33	135.16	0.000*
Curved section	A	2 to 1	−22.90	−27.85	−17.95	0.000*
		3 to 1	3.57	−1.39	8.52	**
		3 to 2	26.47	21.51	31.42	0.000*
	B	2 to 1	−73.77	−89.02	−58.52	0.000*
		3 to 1	2.98	−12.41	18.36	**
		3 to 2	76.44	61.36	92.12	0.000*
	C	2 to 1	−68.00	−97.42	−38.59	0.000*
		3 to 1	39.24	9.56	68.92	0.000*
		3 to 2	107.85	77.32	137.17	0.000*
Removal by a ROBOS vehicle
Straight section	A	2 to 1	−18.62	−35.22	−2.01	0.025*
		3 to 1	0.17	−13.03	13.36	**
		3 to 2	18.78	0.51	37.05	0.042*
Curved section	A	2 to 1	−6.14	−7.75	−4.54	0.000*
		3 to 1	3.79	1.65	5.93	0.000*
		3 to 2	9.93	7.94	11.93	0.000*

P-value ≤0.1; **P-value >0.1.

Although only a lane change to the second lane was possible to avoid debris in the first and third lanes, the first or third lane could be selected from the second lane. Therefore, fewer conflicts occurred in the second lane because the driver could steer the vehicle into the most appropriate lane in a more stable manner when encountering the debris. In the manual removal situation, there are more conflicts in the curved section than in the straight section at service level C. In the curved section, there are more conflicts in the first lane where it was difficult to detect debris because of the road geometry. Therefore, the introduction of a ROBOS vehicle in the sections with many conflicts (service level C, curved section, first and third lane) was considered the most appropriate. Using the road debris removal method, quantitative differences were found by the number of conflicts. A t-test was performed to determine if these differences were statistically significant. Table 8 shows the results of the t-test for each scenario.

Table 8.

T-Test Results

Level of service	Lane	Method	Standard deviation	T-value	P-value
Straight section
A	1	Removal manually	8.72	3.05	0.003
	1	Removal by a ROBOS vehicle	17.02	3.05	0.003
	2	Removal manually	4.26	2.18	0.035*
	2	Removal by a ROBOS vehicle	5.43	2.18	0.035*
	3	Removal manually	7.40	2.28	0.010*
	3	Removal by a ROBOS vehicle	17.79	2.28	0.010*
B	1	Removal manually	17.19	7.51	0.000*
	1	Removal by a ROBOS vehicle	51.40	7.51	0.000*
	2	Removal manually	18.03	1.89	0.066*
	2	Removal by a ROBOS vehicle	19.01	1.89	0.066*
	3	Removal manually	21.18	4.731	0.000*
	3	Removal by a ROBOS vehicle	71.75	4.731	0.000*
C	1	Removal manually	25.52	21.72	0.000*
	1	Removal by a ROBOS vehicle	43.19	21.72	0.000*
	2	Removal manually	36.26	13.85	0.000*
	2	Removal by a ROBOS vehicle	35.71	13.85	0.000*
	3	Removal manually	36.04	22.27	0.000*
	3	Removal by a ROBOS vehicle	44.42	22.27	0.000*
Curved section
A	1	Removal manually	8.31	14.58	0.000*
	1	Removal by a ROBOS vehicle	2.62	14.58	0.000*
	2	Removal manually	4.39	8.68	0.000*
	2	Removal by a ROBOS vehicle	1.48	8.68	0.000*
	3	Removal manually	9.15	12.57	0.000*
	3	Removal by a ROBOS vehicle	2.61	12.57	0.000*
B	1	Removal manually	20.34	22.34	0.000*
	1	Removal by a ROBOS vehicle	10.36	22.34	0.000*
	2	Removal manually	18.27	6.05	0.000*
	2	Removal by a ROBOS vehicle	13.92	6.05	0.000*
	3	Removal manually	29.90	3.64	0.000*
	3	Removal by a ROBOS vehicle	28.10	3.64	0.000*
C	1	Removal manually	37.42	19.31	0.000*
	1	Removal by a ROBOS vehicle	18.72	19.31	0.000*
	2	Removal manually	48.93	11.28	0.000*
	2	Removal by a ROBOS vehicle	26.95	11.28	0.000*
	3	Removal manually	46.73	8.98	0.000*
	3	Removal by a ROBOS vehicle	127.35	8.98	0.000*

Note:*P-value ≤ 0.1

Since all the P-values were less than 0.1 in all scenarios, the number of conflicts according to the road debris removal method can be considered statistically significant. The level of traffic safety is statistically different for the removal methods; however, removal by a ROBOS vehicle was found to have a more positive impact on traffic safety.

When the number of conflicts was analyzed by debris removal method, there were fewer conflicts with removal by the ROBOS vehicle than with manual removal. The use of the ROBOS vehicle reduced the number of conflicts more significantly for the removal of debris in the first and third lanes than in the second lane. In particular, when a ROBOS vehicle is placed in the first lane of a curved section, the highest reduction rate in the scenario of approximately 70% or more is achieved. Since the vehicle in the second lane can perform selective lane changes, the use of a ROBOS vehicle in the more dangerous first and third lanes was considered to be the most effective. Regardless of the level of service and the location of road debris, the use of the ROBOS vehicle has a positive effect on safety. In particular, it was highly effective in improving safety in the curved section.

These results indicate that the ROBOS vehicle should be employed for sections where debris removal is the most dangerous for road workers. The input and operating systems of the ROBOS vehicle can be improved by a safety analysis of the road geometry and lane position. The conflict rate was estimated for each scenario by calculating the number of conflicts compared with the traffic volume. Table 9 and Figure 8 show the results of the conflict rate analysis according to the debris removal method. Similar to the number of conflicts, the conflict rate for ROBOS vehicle removal is lower than for manual removal.

Table 9.

Analysis of Conflict Rate

Removal manually				Removal by a ROBOS vehicle
Lane	A	B	C	Lane	A	B	C
Straight section
1st	0.021	0.050	0.091	1^st	0.015	0.026	0.035
2nd	0.006	0.024	0.071	2^nd	0.004	0.020	0.027
3rd	0.023	0.055	0.098	3^rd	0.015	0.028	0.039
Curved section
1st	0.019	0.048	0.115	1^st	0.005	0.013	0.033
2nd	0.005	0.023	0.091	2^nd	0.002	0.012	0.033
3rd	0.021	0.053	0.110	3^rd	0.007	0.039	0.054

Figure 8.

Results of conflict rate by scenario.

For manual removal, the conflict rate differs depending on the road geometry for service level C. The conflict rate is high in the first and third lanes of the curved and straight sections. In particular, the first lane of the curved section shows the most negative traffic safety. It is expected that the use of a ROBOS vehicle in the most negative sections will, relatively, improve traffic safety. The ROBOS vehicle appears to have a relatively low conflict rate because it immediately removes the debris without a separate working space, and thus no risk is caused by this space. Drivers also seemed to find it easier to change lanes when encountering a slow-moving ROBOS vehicle removing debris, as there was no work area to interrupt the flow of traffic. Therefore, from the perspective of traffic agencies, the use of a ROBOS vehicle is expected to probably result in safer work.

Conclusions

A technology that automatically removes road debris is needed to significantly improve road safety for both road workers and the general public. Accordingly, ROBOS was developed in South Korea, considering the domestic road environment, laws, and policies. The road debris collection and loading performance of the first vehicle equipped with ROBOS has been previously evaluated through various tests on actual roads; however, although investigations on the debris removal capability of the ROBOS vehicle have been conducted, there are no studies existing on its effect on traffic safety. Therefore, this study analyzed the effect of introducing a ROBOS vehicle for debris removal in various road environments. To this end, a driving simulator and microscopic traffic simulation were utilized as the main analysis tools, and, finally, traffic safety according to the debris removal method was evaluated based on SSAM.

Using a driving simulator, we analyzed the driving behavior of drivers when there is road debris on the road and when the ROBOS vehicle is performing road debris removal. The analysis of the driving behavior of the participants according to the scenarios showed that, when there is road debris on the road, the distance to change lanes is shorter and the amount of deceleration and steering is greater than when the ROBOS vehicle is working. This indicates that the driver did not detect the road debris while driving relatively fast and made a sharp lane change to avoid it. The analysis results from the simulator were used as parameters for VISSIM. In other words, by including the lane-change distance as a parameter in VISSIM, the driving behavior of real drivers was realistically represented. In addition, the traffic simulation environment is the same as the driving simulator, and additional traffic conditions (e.g., service level) are considered.

The first and third lanes of the curved sections had more conflicts than in the other sections. The presence of debris on the curved section negatively affected traffic safety because it was difficult to quickly detect the debris because of the driver’s limited visibility. The number of conflicts and the conflict rate were higher in the first and third lanes than in the second lane because the lane change was limited to only one lane (the second lane). Using a ROBOS vehicle in the first lane of the curve (least safe) reduced the number of conflicts to approximately 70% of the manual removal. Therefore, ANOVA and a t-test were performed to determine the statistical significance of the quantitative comparison of the results for each scenario. The results indicated that the difference between manual removal and ROBOS removal was statistically significant with reference to traffic safety.

However, additional research is needed before a ROBOS vehicle can be used in the field. Specific driving behaviors need to be subdivided (e.g., older drivers need to be considered) for more objective verification. In addition, this study only conducted experiments on a single section without access roads; therefore, access roads, weaving sections, and traffic accident hot spots must also be considered when setting up the experimental network. However, the effect of introducing a ROBOS vehicle on the traffic flow will be different depending on the road geometry. In this study, safety was evaluated based on the number of conflicts caused by debris. If the safety level is evaluated taking into account different surrogate safety measures, road managers will be able to introduce ROBOS vehicles more strategically. In addition, future research should address not only safety, but also optimal ROBOS input and operation methods.

In conclusion, when debris is present in the first and third lanes of curved sections, removal by a ROBOS vehicle should be given higher priority than manual removal. The results of this study indicate that the use of a ROBOS vehicle is beneficial for traffic safety, and that overall traffic safety can be improved by the systematic use of ROBOS vehicles, taking into account the debris location and road geometry.

Footnotes

Author Contributions

The authors confirm contribution to the paper as follows: study conception and design: J. Kim, C. Yang; data collection: S. Park; analysis and interpretation of results: S. Park, J. Kim, C. Yang; draft manuscript preparation: S. Park, C. Yang. All authors reviewed the results and approved the final version of the manuscript.

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This work was supported by a grant from the Korea Institute for Advancement of Technology (KIAT) funded by the Government of Korea (Ministry of Trade, Industry and Energy). Grant Number: P0019326; Development of High-Standard Road Debris Remover System.

ORCID iDs

Subin Park

Choongheon Yang

Data Accessibility Statement

Some or all data, models, or code that support the findings of this study are available from the corresponding author on reasonable request.

References

Korea Road Traffic Authority. Traffic Accident Analysis System (TAAS). http://taas.koroad.or.kr/. Accessed January 16, 2023.

Hyundai Insurance Research Center. Analysis of Highway Debris Accident Characteristics, 2021. https://blog.hi.co.kr/2507. Accessed February, 21, 2023.

Valdes-Vasquez

Lewis

A. M.

Strong

K. C.

Assessment of a System for Debris Removal at High Speeds: Implications for Roadway Operations and Maintenance. Transportation Research Record: Journal of the Transportation Research Board, 2014. 2440: 9–15.

Korea Agency for Infrastructure Technology Advancement. Development of Accident Risk Mitigating Technology for Vertical Structure Collision on Road and Road Workers Annual Report, 2016.

Yang

C. H.

Kim

H. K.

Kim

J. G.

Kim

Impact of an Automated Road Debris Removal System on Traffic Flow Using VISSIM. Simulation, Vol. 96, No. 9, 2020, pp. 725–735.

Kim

J. G.

Yang

C. H.

Kim

H. K.

Simulation Analysis for ROBOS Operation on Actual Roads. International Journal of Highway Engineering, Vol. 21, No. 4, 2019, pp. 95–100.

Almallah

Hussain

Alhajyaseen

W. K. M.

Pirdavani

Brijs

Dias

Brijs

Improved Traffic Safety at Work Zones Through Animation-Based Variable Message Signs. Accident Analysis & Prevention, Vol. 159, 2021, p. 106284. https://doi.org/10.1016/j.aap.2021.106284.

Bham

G. H.

Leu

M. C.

Vallati

Mathur

D. R.

Driving Simulator Validation of Driver Behavior with Limited Safe Vantage Points for Data Collection in Work Zones. Journal of Safety Research, Vol. 49, 2014, pp. 53–60. https://doi.org/10.1016/j.jsr.2014.02.012.

Domenichini

La Torre

Branzi

Nocentini

Speed Behaviour in Work Zone Crossovers. A Driving Simulator Study. Accident Analysis & Prevention, Vol. 98, 2017, pp. 10–24. https://doi.org/10.1016/j.aap.2016.09.018.

10.

Duan

Yan

Hang

Yang

Analysis of Driver’s Decision Distance and Merging Distance in Work Zone Area Based on Parametric Survival Models: With the Aid of a Driving Simulator Experiment. Transportation Research Part F: Traffic Psychology and Behaviour, Vol. 71, 2020, pp. 31–48. https://doi.org/10.1016/j.trf.2020.03.017.

11.

Hang

Yan

Duan

In-Vehicle Warnings for Work Zone and Related Rear-End Collisions: A Driving Simulator Experiment. Accident Analysis & Prevention, Vol. 174, 2022, p. 106768. https://doi.org/10.1016/j.aap.2022.106768.

12.

Bakhshi

A. K.

Ahmed

M. M.

Accounting for Human-Related Unobserved Heterogeneity in the Safety Performance of Connected Vehicles: An Incorporation of Bayesian Hierarchical Negative Binomial into Simulated Work Zone Warning Application. IATSS Research, Vol. 45, No. 4, 2021, pp. 539–550.

13.

Raddaoui

Ahmed

M. M.

Gaweesh

S. M.

Assessment of the Effectiveness of Connected Vehicle Weather and Work Zone Warnings in Improving Truck Driver Safety. IATSS Research, Vol. 44, No. 3, 2020, pp. 230–237.

14.

Thapa

Mishra

Using Workers Naturalistic Response to Determine and Analyze Work Zone Crashes in the Presence of Work Zone Intrusion Alert Systems. Accident Analysis & Prevention, Vol. 156, 2021, p. 106125. https://doi.org/10.1016/j.aap.2021.106125.

15.

Zhao

Chen

Chang

Feng

Chen

Development and Application of Connected Vehicle Technology Test Platform Based on Driving Simulator: Case Study. Accident Analysis & Prevention, Vol. 161, 2021, p. 106330. https://doi.org/10.1016/j.aap.2021.106330.

16.

Kwon

Y. M.

Jang

K. T.

Son

S. H.

Risk Factors Affecting the Injury Severity of Rental Car Accidents in South Korea: An Application of Ordered Probit Model. The Journal of The Korea Institute of Intelligent Transport Systems, Vol. 17, No. 3, 2018, pp. 1–17.

17.

Korean Ministry of Land, Infrastructure and Transport. Rules about the Road Structure and Facilities Standards, 2020.

18.

Korean Ministry of Land, Infrastructure and Transport. Transport and Maritime Affairs, Korea Highway Capacity Manual2013, 2013.

19.

Korean Ministry of Land, Infrastructure and Transport. Guidelines on Road Construction Site Traffic Management, 2018.

20.

Kim

H. J.

Jin

E. J.

S. J.

Park

J. J.

A Study on the Analysis of Falling Objects from Vehicle Accidents Through AI Learning-Based Keyword Analysis. Journal of Korean Society of Transportation, Vol. 39, No. 1, 2021, pp. 78–86.

21.

J. E.

Kim

Y. J.

Park

J. W.

A Methodology for Evaluating Real-time Crash Risks in Driving Big Data Era. Journal of Korean Society of Transportation, Vol. 37, No. 4, 2019, pp. 350–364.

22.

Bared

Surrogate Safety Assessment Model (SSAM). FHWA-HRT-08-049. Federal Highway Administration Research and Technology, Washington, D.C., 2008.

23.

Hayward

J. C.

Near-Miss Determination through Use of a Scale of Danger. Highway Research Board, Vol. 384, 1972, pp. 24–35.

24.

Sayed

Brown

Navis

Simulation of Traffic Conflicts at Unsignalized Intersections with TSC-Sim. Accident Analysis & Prevention, Vol. 26, No. 5, 1994, pp. 593–607.

25.

Hwang

D. H.

Park

H. J.

Jeong

B. J.

Safety Evaluation of Traffic Merging Control Methods for Freeway Work Zone. Journal of Korean Society of Transportation, Vol. 34, No. 3, 2016, pp. 263–277.

26.

Lee

S. Y.

A Methodology to Establish Operational Strategies for Truck Platoonings on Freeway On-ramp Areas. Journal of Korean Society of Transportation, Vol. 36, No. 2, 2018, pp. 67–85.

27.

Yun

I. S.

Lee

C. K.

Choi

J. H.

S. J.

Safety Assessment of Signalized Intersection Using SSAM: A Case of Actuated Signal Control. The Journal of The Korea Institute of Intelligent Transport Systems, Vol. 10, No. 6, 2011, pp. 1–14.